radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
/*
|
|
|
|
* Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
|
|
|
|
*
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
|
|
|
* copy of this software and associated documentation files (the "Software"),
|
|
|
|
* to deal in the Software without restriction, including without limitation
|
|
|
|
* on the rights to use, copy, modify, merge, publish, distribute, sub
|
|
|
|
* license, and/or sell copies of the Software, and to permit persons to whom
|
|
|
|
* the Software is furnished to do so, subject to the following conditions:
|
|
|
|
*
|
|
|
|
* The above copyright notice and this permission notice (including the next
|
|
|
|
* paragraph) shall be included in all copies or substantial portions of the
|
|
|
|
* Software.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
* THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
|
|
|
|
* DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
|
|
|
|
* OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
|
|
|
|
* USE OR OTHER DEALINGS IN THE SOFTWARE.
|
|
|
|
*/
|
2014-01-22 17:50:36 +00:00
|
|
|
|
|
|
|
#include "si_pipe.h"
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
#include "util/u_format.h"
|
2014-09-06 16:07:50 +01:00
|
|
|
#include "util/u_surface.h"
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
enum si_blitter_op /* bitmask */
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-07 01:51:35 +00:00
|
|
|
SI_SAVE_TEXTURES = 1,
|
|
|
|
SI_SAVE_FRAMEBUFFER = 2,
|
2015-11-06 22:16:11 +00:00
|
|
|
SI_SAVE_FRAGMENT_STATE = 4,
|
|
|
|
SI_DISABLE_RENDER_COND = 8,
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2015-11-06 22:16:11 +00:00
|
|
|
SI_CLEAR = SI_SAVE_FRAGMENT_STATE,
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2015-11-06 22:16:11 +00:00
|
|
|
SI_CLEAR_SURFACE = SI_SAVE_FRAMEBUFFER | SI_SAVE_FRAGMENT_STATE,
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
SI_COPY = SI_SAVE_FRAMEBUFFER | SI_SAVE_TEXTURES |
|
2015-11-06 22:16:11 +00:00
|
|
|
SI_SAVE_FRAGMENT_STATE | SI_DISABLE_RENDER_COND,
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2015-11-06 22:16:11 +00:00
|
|
|
SI_BLIT = SI_SAVE_FRAMEBUFFER | SI_SAVE_TEXTURES |
|
|
|
|
SI_SAVE_FRAGMENT_STATE,
|
2012-09-13 17:48:51 +01:00
|
|
|
|
2015-11-06 22:16:11 +00:00
|
|
|
SI_DECOMPRESS = SI_SAVE_FRAMEBUFFER | SI_SAVE_FRAGMENT_STATE |
|
|
|
|
SI_DISABLE_RENDER_COND,
|
2013-07-30 21:29:28 +01:00
|
|
|
|
2015-11-06 22:16:11 +00:00
|
|
|
SI_COLOR_RESOLVE = SI_SAVE_FRAMEBUFFER | SI_SAVE_FRAGMENT_STATE
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
};
|
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
static void si_blitter_begin(struct pipe_context *ctx, enum si_blitter_op op)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
|
|
|
|
2014-01-21 23:06:32 +00:00
|
|
|
r600_suspend_nontimer_queries(&sctx->b);
|
2014-01-11 15:00:50 +00:00
|
|
|
|
2015-11-06 22:16:11 +00:00
|
|
|
util_blitter_save_vertex_buffer_slot(sctx->blitter, sctx->vertex_buffer);
|
|
|
|
util_blitter_save_vertex_elements(sctx->blitter, sctx->vertex_elements);
|
|
|
|
util_blitter_save_vertex_shader(sctx->blitter, sctx->vs_shader.cso);
|
2015-10-07 00:48:18 +01:00
|
|
|
util_blitter_save_tessctrl_shader(sctx->blitter, sctx->tcs_shader.cso);
|
|
|
|
util_blitter_save_tesseval_shader(sctx->blitter, sctx->tes_shader.cso);
|
2015-11-06 22:16:11 +00:00
|
|
|
util_blitter_save_geometry_shader(sctx->blitter, sctx->gs_shader.cso);
|
2014-01-11 15:00:50 +00:00
|
|
|
util_blitter_save_so_targets(sctx->blitter, sctx->b.streamout.num_targets,
|
|
|
|
(struct pipe_stream_output_target**)sctx->b.streamout.targets);
|
2015-11-06 22:16:11 +00:00
|
|
|
util_blitter_save_rasterizer(sctx->blitter, sctx->queued.named.rasterizer);
|
|
|
|
|
|
|
|
if (op & SI_SAVE_FRAGMENT_STATE) {
|
|
|
|
util_blitter_save_blend(sctx->blitter, sctx->queued.named.blend);
|
|
|
|
util_blitter_save_depth_stencil_alpha(sctx->blitter, sctx->queued.named.dsa);
|
|
|
|
util_blitter_save_stencil_ref(sctx->blitter, &sctx->stencil_ref.state);
|
|
|
|
util_blitter_save_fragment_shader(sctx->blitter, sctx->ps_shader.cso);
|
|
|
|
util_blitter_save_sample_mask(sctx->blitter, sctx->sample_mask.sample_mask);
|
|
|
|
util_blitter_save_viewport(sctx->blitter, &sctx->viewports.states[0]);
|
|
|
|
util_blitter_save_scissor(sctx->blitter, &sctx->scissors.states[0]);
|
|
|
|
}
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
if (op & SI_SAVE_FRAMEBUFFER)
|
2014-03-04 16:49:39 +00:00
|
|
|
util_blitter_save_framebuffer(sctx->blitter, &sctx->framebuffer.state);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
if (op & SI_SAVE_TEXTURES) {
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
util_blitter_save_fragment_sampler_states(
|
2014-06-18 01:46:49 +01:00
|
|
|
sctx->blitter, 2,
|
2016-02-06 21:09:45 +00:00
|
|
|
sctx->samplers[PIPE_SHADER_FRAGMENT].views.sampler_states);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-06-18 02:23:46 +01:00
|
|
|
util_blitter_save_fragment_sampler_views(sctx->blitter, 2,
|
2014-01-11 15:00:50 +00:00
|
|
|
sctx->samplers[PIPE_SHADER_FRAGMENT].views.views);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2015-11-07 13:45:58 +00:00
|
|
|
if (op & SI_DISABLE_RENDER_COND)
|
|
|
|
sctx->b.render_cond_force_off = true;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
static void si_blitter_end(struct pipe_context *ctx)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
2015-11-07 13:45:58 +00:00
|
|
|
|
|
|
|
sctx->b.render_cond_force_off = false;
|
2014-01-21 23:06:32 +00:00
|
|
|
r600_resume_nontimer_queries(&sctx->b);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2013-08-06 20:30:59 +01:00
|
|
|
static unsigned u_max_sample(struct pipe_resource *r)
|
|
|
|
{
|
|
|
|
return r->nr_samples ? r->nr_samples - 1 : 0;
|
|
|
|
}
|
|
|
|
|
2014-01-07 01:40:22 +00:00
|
|
|
static void si_blit_decompress_depth(struct pipe_context *ctx,
|
|
|
|
struct r600_texture *texture,
|
|
|
|
struct r600_texture *staging,
|
|
|
|
unsigned first_level, unsigned last_level,
|
|
|
|
unsigned first_layer, unsigned last_layer,
|
|
|
|
unsigned first_sample, unsigned last_sample)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
2013-08-06 20:30:59 +01:00
|
|
|
unsigned layer, level, sample, checked_last_layer, max_layer, max_sample;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
float depth = 1.0f;
|
2013-01-17 18:36:41 +00:00
|
|
|
const struct util_format_description *desc;
|
2013-08-05 02:42:11 +01:00
|
|
|
struct r600_texture *flushed_depth_texture = staging ?
|
2013-01-15 18:36:32 +00:00
|
|
|
staging : texture->flushed_depth_texture;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-08-05 13:40:43 +01:00
|
|
|
if (!staging && !texture->dirty_level_mask)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
return;
|
|
|
|
|
2013-08-06 20:30:59 +01:00
|
|
|
max_sample = u_max_sample(&texture->resource.b.b);
|
|
|
|
|
2013-01-17 18:36:41 +00:00
|
|
|
desc = util_format_description(flushed_depth_texture->resource.b.b.format);
|
2014-08-23 02:25:29 +01:00
|
|
|
|
|
|
|
if (util_format_has_depth(desc))
|
|
|
|
sctx->dbcb_depth_copy_enabled = true;
|
|
|
|
if (util_format_has_stencil(desc))
|
|
|
|
sctx->dbcb_stencil_copy_enabled = true;
|
|
|
|
|
|
|
|
assert(sctx->dbcb_depth_copy_enabled || sctx->dbcb_stencil_copy_enabled);
|
2013-01-17 18:36:41 +00:00
|
|
|
|
|
|
|
for (level = first_level; level <= last_level; level++) {
|
2013-08-05 13:40:43 +01:00
|
|
|
if (!staging && !(texture->dirty_level_mask & (1 << level)))
|
2013-01-17 18:36:41 +00:00
|
|
|
continue;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-01-17 18:36:41 +00:00
|
|
|
/* The smaller the mipmap level, the less layers there are
|
|
|
|
* as far as 3D textures are concerned. */
|
2013-01-02 19:40:24 +00:00
|
|
|
max_layer = util_max_layer(&texture->resource.b.b, level);
|
2013-01-17 18:36:41 +00:00
|
|
|
checked_last_layer = last_layer < max_layer ? last_layer : max_layer;
|
|
|
|
|
|
|
|
for (layer = first_layer; layer <= checked_last_layer; layer++) {
|
2013-08-06 20:30:59 +01:00
|
|
|
for (sample = first_sample; sample <= last_sample; sample++) {
|
|
|
|
struct pipe_surface *zsurf, *cbsurf, surf_tmpl;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-08-23 02:25:29 +01:00
|
|
|
sctx->dbcb_copy_sample = sample;
|
2015-08-09 22:42:32 +01:00
|
|
|
si_mark_atom_dirty(sctx, &sctx->db_render_state);
|
2014-08-23 02:25:29 +01:00
|
|
|
|
2013-08-06 20:30:59 +01:00
|
|
|
surf_tmpl.format = texture->resource.b.b.format;
|
|
|
|
surf_tmpl.u.tex.level = level;
|
|
|
|
surf_tmpl.u.tex.first_layer = layer;
|
|
|
|
surf_tmpl.u.tex.last_layer = layer;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-08-06 20:30:59 +01:00
|
|
|
zsurf = ctx->create_surface(ctx, &texture->resource.b.b, &surf_tmpl);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-08-06 20:30:59 +01:00
|
|
|
surf_tmpl.format = flushed_depth_texture->resource.b.b.format;
|
|
|
|
cbsurf = ctx->create_surface(ctx,
|
|
|
|
(struct pipe_resource*)flushed_depth_texture, &surf_tmpl);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_begin(ctx, SI_DECOMPRESS);
|
2014-01-11 15:00:50 +00:00
|
|
|
util_blitter_custom_depth_stencil(sctx->blitter, zsurf, cbsurf, 1 << sample,
|
2014-08-23 02:25:29 +01:00
|
|
|
sctx->custom_dsa_flush, depth);
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_end(ctx);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-08-06 20:30:59 +01:00
|
|
|
pipe_surface_reference(&zsurf, NULL);
|
|
|
|
pipe_surface_reference(&cbsurf, NULL);
|
|
|
|
}
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2013-01-17 18:36:41 +00:00
|
|
|
/* The texture will always be dirty if some layers aren't flushed.
|
|
|
|
* I don't think this case can occur though. */
|
2013-08-06 20:30:59 +01:00
|
|
|
if (!staging &&
|
|
|
|
first_layer == 0 && last_layer == max_layer &&
|
|
|
|
first_sample == 0 && last_sample == max_sample) {
|
2013-08-05 13:40:43 +01:00
|
|
|
texture->dirty_level_mask &= ~(1 << level);
|
2013-01-17 18:36:41 +00:00
|
|
|
}
|
|
|
|
}
|
2014-08-23 02:25:29 +01:00
|
|
|
|
|
|
|
sctx->dbcb_depth_copy_enabled = false;
|
|
|
|
sctx->dbcb_stencil_copy_enabled = false;
|
2015-08-09 22:42:32 +01:00
|
|
|
si_mark_atom_dirty(sctx, &sctx->db_render_state);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2014-01-11 15:00:50 +00:00
|
|
|
static void si_blit_decompress_depth_in_place(struct si_context *sctx,
|
2013-08-05 02:42:11 +01:00
|
|
|
struct r600_texture *texture,
|
2015-09-06 16:37:38 +01:00
|
|
|
bool is_stencil_sampler,
|
2013-01-17 18:36:41 +00:00
|
|
|
unsigned first_level, unsigned last_level,
|
|
|
|
unsigned first_layer, unsigned last_layer)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2013-01-17 18:36:41 +00:00
|
|
|
struct pipe_surface *zsurf, surf_tmpl = {{0}};
|
|
|
|
unsigned layer, max_layer, checked_last_layer, level;
|
2015-09-06 16:37:38 +01:00
|
|
|
unsigned *dirty_level_mask;
|
|
|
|
|
|
|
|
if (is_stencil_sampler) {
|
|
|
|
sctx->db_flush_stencil_inplace = true;
|
|
|
|
dirty_level_mask = &texture->stencil_dirty_level_mask;
|
|
|
|
} else {
|
|
|
|
sctx->db_flush_depth_inplace = true;
|
|
|
|
dirty_level_mask = &texture->dirty_level_mask;
|
|
|
|
}
|
2015-08-09 22:42:32 +01:00
|
|
|
si_mark_atom_dirty(sctx, &sctx->db_render_state);
|
2014-08-23 02:25:29 +01:00
|
|
|
|
2013-01-17 18:36:41 +00:00
|
|
|
surf_tmpl.format = texture->resource.b.b.format;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-01-17 18:36:41 +00:00
|
|
|
for (level = first_level; level <= last_level; level++) {
|
2015-09-06 16:37:38 +01:00
|
|
|
if (!(*dirty_level_mask & (1 << level)))
|
2013-01-17 18:36:41 +00:00
|
|
|
continue;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-01-17 18:36:41 +00:00
|
|
|
surf_tmpl.u.tex.level = level;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-01-17 18:36:41 +00:00
|
|
|
/* The smaller the mipmap level, the less layers there are
|
|
|
|
* as far as 3D textures are concerned. */
|
2013-01-02 19:40:24 +00:00
|
|
|
max_layer = util_max_layer(&texture->resource.b.b, level);
|
2013-01-17 18:36:41 +00:00
|
|
|
checked_last_layer = last_layer < max_layer ? last_layer : max_layer;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-01-17 18:36:41 +00:00
|
|
|
for (layer = first_layer; layer <= checked_last_layer; layer++) {
|
|
|
|
surf_tmpl.u.tex.first_layer = layer;
|
|
|
|
surf_tmpl.u.tex.last_layer = layer;
|
|
|
|
|
2014-01-11 15:00:50 +00:00
|
|
|
zsurf = sctx->b.b.create_surface(&sctx->b.b, &texture->resource.b.b, &surf_tmpl);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-11 15:00:50 +00:00
|
|
|
si_blitter_begin(&sctx->b.b, SI_DECOMPRESS);
|
|
|
|
util_blitter_custom_depth_stencil(sctx->blitter, zsurf, NULL, ~0,
|
2014-08-23 02:25:29 +01:00
|
|
|
sctx->custom_dsa_flush,
|
2013-01-17 18:36:41 +00:00
|
|
|
1.0f);
|
2014-01-11 15:00:50 +00:00
|
|
|
si_blitter_end(&sctx->b.b);
|
2013-01-17 18:36:41 +00:00
|
|
|
|
|
|
|
pipe_surface_reference(&zsurf, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* The texture will always be dirty if some layers aren't flushed.
|
|
|
|
* I don't think this case occurs often though. */
|
|
|
|
if (first_layer == 0 && last_layer == max_layer) {
|
2015-09-06 16:37:38 +01:00
|
|
|
*dirty_level_mask &= ~(1 << level);
|
2013-01-17 18:36:41 +00:00
|
|
|
}
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
2014-08-23 02:25:29 +01:00
|
|
|
|
2015-09-06 16:37:38 +01:00
|
|
|
sctx->db_flush_depth_inplace = false;
|
|
|
|
sctx->db_flush_stencil_inplace = false;
|
2015-08-09 22:42:32 +01:00
|
|
|
si_mark_atom_dirty(sctx, &sctx->db_render_state);
|
2013-01-17 18:36:41 +00:00
|
|
|
}
|
|
|
|
|
2016-02-08 03:27:17 +00:00
|
|
|
static void
|
|
|
|
si_flush_depth_textures(struct si_context *sctx,
|
|
|
|
struct si_textures_info *textures)
|
2013-01-17 18:36:41 +00:00
|
|
|
{
|
|
|
|
unsigned i;
|
2014-06-18 02:23:46 +01:00
|
|
|
unsigned mask = textures->depth_texture_mask;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-06-18 02:23:46 +01:00
|
|
|
while (mask) {
|
2013-01-17 18:36:41 +00:00
|
|
|
struct pipe_sampler_view *view;
|
2015-09-06 16:37:38 +01:00
|
|
|
struct si_sampler_view *sview;
|
2013-08-05 02:42:11 +01:00
|
|
|
struct r600_texture *tex;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-06-18 02:23:46 +01:00
|
|
|
i = u_bit_scan(&mask);
|
|
|
|
|
2013-08-06 05:42:22 +01:00
|
|
|
view = textures->views.views[i];
|
2014-06-18 02:23:46 +01:00
|
|
|
assert(view);
|
2015-09-06 16:37:38 +01:00
|
|
|
sview = (struct si_sampler_view*)view;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2013-08-05 02:42:11 +01:00
|
|
|
tex = (struct r600_texture *)view->texture;
|
2014-06-18 02:23:46 +01:00
|
|
|
assert(tex->is_depth && !tex->is_flushing_texture);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-11 15:00:50 +00:00
|
|
|
si_blit_decompress_depth_in_place(sctx, tex,
|
2015-09-06 16:37:38 +01:00
|
|
|
sview->is_stencil_sampler,
|
2013-01-17 18:36:41 +00:00
|
|
|
view->u.tex.first_level, view->u.tex.last_level,
|
2013-01-02 19:40:24 +00:00
|
|
|
0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level));
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
static void si_blit_decompress_color(struct pipe_context *ctx,
|
2013-08-06 07:48:07 +01:00
|
|
|
struct r600_texture *rtex,
|
|
|
|
unsigned first_level, unsigned last_level,
|
2015-10-20 23:10:41 +01:00
|
|
|
unsigned first_layer, unsigned last_layer,
|
|
|
|
bool need_dcc_decompress)
|
2013-08-06 07:48:07 +01:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
2013-08-06 07:48:07 +01:00
|
|
|
unsigned layer, level, checked_last_layer, max_layer;
|
|
|
|
|
2015-10-20 23:10:41 +01:00
|
|
|
if (!rtex->dirty_level_mask && !need_dcc_decompress)
|
2013-08-06 07:48:07 +01:00
|
|
|
return;
|
|
|
|
|
|
|
|
for (level = first_level; level <= last_level; level++) {
|
2015-10-20 23:10:41 +01:00
|
|
|
void* custom_blend;
|
|
|
|
|
|
|
|
if (!(rtex->dirty_level_mask & (1 << level)) && !need_dcc_decompress)
|
2013-08-06 07:48:07 +01:00
|
|
|
continue;
|
|
|
|
|
2015-10-20 23:10:41 +01:00
|
|
|
if (rtex->dcc_offset && need_dcc_decompress) {
|
|
|
|
custom_blend = sctx->custom_blend_dcc_decompress;
|
|
|
|
} else if (rtex->fmask.size) {
|
|
|
|
custom_blend = sctx->custom_blend_decompress;
|
|
|
|
} else {
|
|
|
|
custom_blend = sctx->custom_blend_fastclear;
|
|
|
|
}
|
|
|
|
|
2013-08-06 07:48:07 +01:00
|
|
|
/* The smaller the mipmap level, the less layers there are
|
|
|
|
* as far as 3D textures are concerned. */
|
|
|
|
max_layer = util_max_layer(&rtex->resource.b.b, level);
|
|
|
|
checked_last_layer = last_layer < max_layer ? last_layer : max_layer;
|
|
|
|
|
|
|
|
for (layer = first_layer; layer <= checked_last_layer; layer++) {
|
|
|
|
struct pipe_surface *cbsurf, surf_tmpl;
|
|
|
|
|
|
|
|
surf_tmpl.format = rtex->resource.b.b.format;
|
|
|
|
surf_tmpl.u.tex.level = level;
|
|
|
|
surf_tmpl.u.tex.first_layer = layer;
|
|
|
|
surf_tmpl.u.tex.last_layer = layer;
|
|
|
|
cbsurf = ctx->create_surface(ctx, &rtex->resource.b.b, &surf_tmpl);
|
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_begin(ctx, SI_DECOMPRESS);
|
2015-10-20 23:10:41 +01:00
|
|
|
util_blitter_custom_color(sctx->blitter, cbsurf, custom_blend);
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_end(ctx);
|
2013-08-06 07:48:07 +01:00
|
|
|
|
|
|
|
pipe_surface_reference(&cbsurf, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* The texture will always be dirty if some layers aren't flushed.
|
|
|
|
* I don't think this case occurs often though. */
|
|
|
|
if (first_layer == 0 && last_layer == max_layer) {
|
|
|
|
rtex->dirty_level_mask &= ~(1 << level);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-02-08 03:27:17 +00:00
|
|
|
static void
|
2016-02-08 03:30:46 +00:00
|
|
|
si_decompress_sampler_color_textures(struct si_context *sctx,
|
|
|
|
struct si_textures_info *textures)
|
2013-08-06 07:48:07 +01:00
|
|
|
{
|
|
|
|
unsigned i;
|
|
|
|
unsigned mask = textures->compressed_colortex_mask;
|
|
|
|
|
|
|
|
while (mask) {
|
|
|
|
struct pipe_sampler_view *view;
|
|
|
|
struct r600_texture *tex;
|
|
|
|
|
|
|
|
i = u_bit_scan(&mask);
|
|
|
|
|
|
|
|
view = textures->views.views[i];
|
|
|
|
assert(view);
|
|
|
|
|
|
|
|
tex = (struct r600_texture *)view->texture;
|
2016-02-21 21:49:38 +00:00
|
|
|
assert(tex->cmask.size || tex->fmask.size || tex->dcc_offset);
|
2013-08-06 07:48:07 +01:00
|
|
|
|
2014-01-11 15:00:50 +00:00
|
|
|
si_blit_decompress_color(&sctx->b.b, tex,
|
2014-01-07 01:51:35 +00:00
|
|
|
view->u.tex.first_level, view->u.tex.last_level,
|
2015-10-20 23:10:41 +01:00
|
|
|
0, util_max_layer(&tex->resource.b.b, view->u.tex.first_level),
|
|
|
|
false);
|
2013-08-06 07:48:07 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-02-08 03:30:46 +00:00
|
|
|
static void
|
|
|
|
si_decompress_image_color_textures(struct si_context *sctx,
|
|
|
|
struct si_images_info *images)
|
|
|
|
{
|
|
|
|
unsigned i;
|
|
|
|
unsigned mask = images->compressed_colortex_mask;
|
|
|
|
|
|
|
|
while (mask) {
|
|
|
|
const struct pipe_image_view *view;
|
|
|
|
struct r600_texture *tex;
|
|
|
|
|
|
|
|
i = u_bit_scan(&mask);
|
|
|
|
|
|
|
|
view = &images->views[i];
|
|
|
|
assert(view->resource->target != PIPE_BUFFER);
|
|
|
|
|
|
|
|
tex = (struct r600_texture *)view->resource;
|
|
|
|
if (!tex->cmask.size && !tex->fmask.size && !tex->dcc_offset)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
si_blit_decompress_color(&sctx->b.b, tex,
|
|
|
|
view->u.tex.level, view->u.tex.level,
|
|
|
|
0, util_max_layer(&tex->resource.b.b, view->u.tex.level),
|
|
|
|
false);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-02-08 03:27:17 +00:00
|
|
|
void si_decompress_textures(struct si_context *sctx)
|
|
|
|
{
|
2016-03-09 20:42:31 +00:00
|
|
|
unsigned compressed_colortex_counter;
|
|
|
|
|
2016-02-08 03:27:17 +00:00
|
|
|
if (sctx->blitter->running)
|
|
|
|
return;
|
|
|
|
|
2016-03-09 20:42:31 +00:00
|
|
|
/* Update the compressed_colortex_mask if necessary. */
|
|
|
|
compressed_colortex_counter = p_atomic_read(&sctx->screen->b.compressed_colortex_counter);
|
|
|
|
if (compressed_colortex_counter != sctx->b.last_compressed_colortex_counter) {
|
|
|
|
sctx->b.last_compressed_colortex_counter = compressed_colortex_counter;
|
|
|
|
si_update_compressed_colortex_masks(sctx);
|
|
|
|
}
|
|
|
|
|
2016-02-08 03:27:17 +00:00
|
|
|
/* Flush depth textures which need to be flushed. */
|
|
|
|
for (int i = 0; i < SI_NUM_SHADERS; i++) {
|
|
|
|
if (sctx->samplers[i].depth_texture_mask) {
|
|
|
|
si_flush_depth_textures(sctx, &sctx->samplers[i]);
|
|
|
|
}
|
|
|
|
if (sctx->samplers[i].compressed_colortex_mask) {
|
2016-02-08 03:30:46 +00:00
|
|
|
si_decompress_sampler_color_textures(sctx, &sctx->samplers[i]);
|
|
|
|
}
|
|
|
|
if (sctx->images[i].compressed_colortex_mask) {
|
|
|
|
si_decompress_image_color_textures(sctx, &sctx->images[i]);
|
2016-02-08 03:27:17 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
static void si_clear(struct pipe_context *ctx, unsigned buffers,
|
|
|
|
const union pipe_color_union *color,
|
|
|
|
double depth, unsigned stencil)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
2014-03-04 16:49:39 +00:00
|
|
|
struct pipe_framebuffer_state *fb = &sctx->framebuffer.state;
|
2014-08-23 02:39:08 +01:00
|
|
|
struct pipe_surface *zsbuf = fb->zsbuf;
|
|
|
|
struct r600_texture *zstex =
|
|
|
|
zsbuf ? (struct r600_texture*)zsbuf->texture : NULL;
|
2012-05-16 17:19:13 +01:00
|
|
|
|
2014-03-06 01:44:07 +00:00
|
|
|
if (buffers & PIPE_CLEAR_COLOR) {
|
2015-08-29 01:32:13 +01:00
|
|
|
evergreen_do_fast_color_clear(&sctx->b, fb,
|
|
|
|
&sctx->framebuffer.atom, &buffers,
|
|
|
|
&sctx->framebuffer.dirty_cbufs,
|
|
|
|
color);
|
2015-07-28 10:39:35 +01:00
|
|
|
if (!buffers)
|
|
|
|
return; /* all buffers have been fast cleared */
|
2014-03-06 01:44:07 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (buffers & PIPE_CLEAR_COLOR) {
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* These buffers cannot use fast clear, make sure to disable expansion. */
|
|
|
|
for (i = 0; i < fb->nr_cbufs; i++) {
|
|
|
|
struct r600_texture *tex;
|
|
|
|
|
|
|
|
/* If not clearing this buffer, skip. */
|
|
|
|
if (!(buffers & (PIPE_CLEAR_COLOR0 << i)))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (!fb->cbufs[i])
|
|
|
|
continue;
|
|
|
|
|
|
|
|
tex = (struct r600_texture *)fb->cbufs[i]->texture;
|
|
|
|
if (tex->fmask.size == 0)
|
|
|
|
tex->dirty_level_mask &= ~(1 << fb->cbufs[i]->u.tex.level);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-12-10 00:37:39 +00:00
|
|
|
if (zstex && zstex->htile_buffer &&
|
2014-08-23 02:39:08 +01:00
|
|
|
zsbuf->u.tex.level == 0 &&
|
|
|
|
zsbuf->u.tex.first_layer == 0 &&
|
|
|
|
zsbuf->u.tex.last_layer == util_max_layer(&zstex->resource.b.b, 0)) {
|
2015-12-10 00:37:39 +00:00
|
|
|
if (buffers & PIPE_CLEAR_DEPTH) {
|
|
|
|
/* Need to disable EXPCLEAR temporarily if clearing
|
|
|
|
* to a new value. */
|
|
|
|
if (zstex->depth_cleared && zstex->depth_clear_value != depth) {
|
|
|
|
sctx->db_depth_disable_expclear = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
zstex->depth_clear_value = depth;
|
|
|
|
sctx->framebuffer.dirty_zsbuf = true;
|
|
|
|
si_mark_atom_dirty(sctx, &sctx->framebuffer.atom); /* updates DB_DEPTH_CLEAR */
|
|
|
|
sctx->db_depth_clear = true;
|
|
|
|
si_mark_atom_dirty(sctx, &sctx->db_render_state);
|
2014-08-23 15:46:53 +01:00
|
|
|
}
|
|
|
|
|
2015-12-10 00:37:39 +00:00
|
|
|
if (buffers & PIPE_CLEAR_STENCIL) {
|
|
|
|
stencil &= 0xff;
|
|
|
|
|
|
|
|
/* Need to disable EXPCLEAR temporarily if clearing
|
|
|
|
* to a new value. */
|
|
|
|
if (zstex->stencil_cleared && zstex->stencil_clear_value != stencil) {
|
|
|
|
sctx->db_stencil_disable_expclear = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
zstex->stencil_clear_value = stencil;
|
|
|
|
sctx->framebuffer.dirty_zsbuf = true;
|
|
|
|
si_mark_atom_dirty(sctx, &sctx->framebuffer.atom); /* updates DB_STENCIL_CLEAR */
|
|
|
|
sctx->db_stencil_clear = true;
|
|
|
|
si_mark_atom_dirty(sctx, &sctx->db_render_state);
|
|
|
|
}
|
2014-08-23 02:39:08 +01:00
|
|
|
}
|
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_begin(ctx, SI_CLEAR);
|
2014-01-11 15:00:50 +00:00
|
|
|
util_blitter_clear(sctx->blitter, fb->width, fb->height,
|
2013-11-21 14:41:36 +00:00
|
|
|
util_framebuffer_get_num_layers(fb),
|
2013-05-29 14:51:20 +01:00
|
|
|
buffers, color, depth, stencil);
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_end(ctx);
|
2014-08-23 02:39:08 +01:00
|
|
|
|
2014-08-23 15:46:53 +01:00
|
|
|
if (sctx->db_depth_clear) {
|
|
|
|
sctx->db_depth_clear = false;
|
|
|
|
sctx->db_depth_disable_expclear = false;
|
|
|
|
zstex->depth_cleared = true;
|
2015-08-09 22:42:32 +01:00
|
|
|
si_mark_atom_dirty(sctx, &sctx->db_render_state);
|
2014-08-23 15:46:53 +01:00
|
|
|
}
|
2015-12-10 00:37:39 +00:00
|
|
|
|
|
|
|
if (sctx->db_stencil_clear) {
|
|
|
|
sctx->db_stencil_clear = false;
|
|
|
|
sctx->db_stencil_disable_expclear = false;
|
|
|
|
zstex->stencil_cleared = true;
|
|
|
|
si_mark_atom_dirty(sctx, &sctx->db_render_state);
|
|
|
|
}
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
static void si_clear_render_target(struct pipe_context *ctx,
|
|
|
|
struct pipe_surface *dst,
|
|
|
|
const union pipe_color_union *color,
|
|
|
|
unsigned dstx, unsigned dsty,
|
|
|
|
unsigned width, unsigned height)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_begin(ctx, SI_CLEAR_SURFACE);
|
2014-01-11 15:00:50 +00:00
|
|
|
util_blitter_clear_render_target(sctx->blitter, dst, color,
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
dstx, dsty, width, height);
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_end(ctx);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
static void si_clear_depth_stencil(struct pipe_context *ctx,
|
|
|
|
struct pipe_surface *dst,
|
|
|
|
unsigned clear_flags,
|
|
|
|
double depth,
|
|
|
|
unsigned stencil,
|
|
|
|
unsigned dstx, unsigned dsty,
|
|
|
|
unsigned width, unsigned height)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_begin(ctx, SI_CLEAR_SURFACE);
|
2014-01-11 15:00:50 +00:00
|
|
|
util_blitter_clear_depth_stencil(sctx->blitter, dst, clear_flags, depth, stencil,
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
dstx, dsty, width, height);
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_end(ctx);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2013-08-06 07:48:07 +01:00
|
|
|
/* Helper for decompressing a portion of a color or depth resource before
|
|
|
|
* blitting if any decompression is needed.
|
|
|
|
* The driver doesn't decompress resources automatically while u_blitter is
|
|
|
|
* rendering. */
|
2014-01-07 01:51:35 +00:00
|
|
|
static void si_decompress_subresource(struct pipe_context *ctx,
|
|
|
|
struct pipe_resource *tex,
|
|
|
|
unsigned level,
|
|
|
|
unsigned first_layer, unsigned last_layer)
|
2013-08-06 07:48:07 +01:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
2013-08-06 07:48:07 +01:00
|
|
|
struct r600_texture *rtex = (struct r600_texture*)tex;
|
|
|
|
|
|
|
|
if (rtex->is_depth && !rtex->is_flushing_texture) {
|
2015-09-06 16:37:38 +01:00
|
|
|
si_blit_decompress_depth_in_place(sctx, rtex, false,
|
2013-08-06 07:48:07 +01:00
|
|
|
level, level,
|
|
|
|
first_layer, last_layer);
|
2015-09-06 16:37:38 +01:00
|
|
|
if (rtex->surface.flags & RADEON_SURF_SBUFFER)
|
|
|
|
si_blit_decompress_depth_in_place(sctx, rtex, true,
|
|
|
|
level, level,
|
|
|
|
first_layer, last_layer);
|
2016-02-21 21:49:38 +00:00
|
|
|
} else if (rtex->fmask.size || rtex->cmask.size || rtex->dcc_offset) {
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blit_decompress_color(ctx, rtex, level, level,
|
2015-10-20 23:10:41 +01:00
|
|
|
first_layer, last_layer, false);
|
2013-08-06 07:48:07 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
struct texture_orig_info {
|
|
|
|
unsigned format;
|
|
|
|
unsigned width0;
|
|
|
|
unsigned height0;
|
2012-05-16 17:19:13 +01:00
|
|
|
unsigned npix_x;
|
|
|
|
unsigned npix_y;
|
|
|
|
unsigned npix0_x;
|
|
|
|
unsigned npix0_y;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
};
|
|
|
|
|
2014-09-06 16:07:50 +01:00
|
|
|
void si_resource_copy_region(struct pipe_context *ctx,
|
|
|
|
struct pipe_resource *dst,
|
|
|
|
unsigned dst_level,
|
|
|
|
unsigned dstx, unsigned dsty, unsigned dstz,
|
|
|
|
struct pipe_resource *src,
|
|
|
|
unsigned src_level,
|
|
|
|
const struct pipe_box *src_box)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context *)ctx;
|
2014-02-11 02:02:35 +00:00
|
|
|
struct pipe_surface *dst_view, dst_templ;
|
|
|
|
struct pipe_sampler_view src_templ, *src_view;
|
2015-07-16 12:17:14 +01:00
|
|
|
unsigned dst_width, dst_height, src_width0, src_height0;
|
|
|
|
unsigned src_force_level = 0;
|
2014-02-11 02:02:35 +00:00
|
|
|
struct pipe_box sbox, dstbox;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2015-07-16 12:17:14 +01:00
|
|
|
/* Handle buffers first. */
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
if (dst->target == PIPE_BUFFER && src->target == PIPE_BUFFER) {
|
2014-12-29 13:45:49 +00:00
|
|
|
si_copy_buffer(sctx, dst, src, dstx, src_box->x, src_box->width, false);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2015-07-16 12:17:14 +01:00
|
|
|
assert(u_max_sample(dst) == u_max_sample(src));
|
2013-11-27 12:27:54 +00:00
|
|
|
|
2013-08-06 07:48:07 +01:00
|
|
|
/* The driver doesn't decompress resources automatically while
|
|
|
|
* u_blitter is rendering. */
|
2014-01-07 01:51:35 +00:00
|
|
|
si_decompress_subresource(ctx, src, src_level,
|
|
|
|
src_box->z, src_box->z + src_box->depth - 1);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2015-07-16 12:17:14 +01:00
|
|
|
dst_width = u_minify(dst->width0, dst_level);
|
|
|
|
dst_height = u_minify(dst->height0, dst_level);
|
|
|
|
src_width0 = src->width0;
|
|
|
|
src_height0 = src->height0;
|
|
|
|
|
|
|
|
util_blitter_default_dst_texture(&dst_templ, dst, dst_level, dstz);
|
|
|
|
util_blitter_default_src_texture(&src_templ, src, src_level);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2015-08-24 00:08:48 +01:00
|
|
|
if (util_format_is_compressed(src->format) ||
|
2013-01-11 11:14:28 +00:00
|
|
|
util_format_is_compressed(dst->format)) {
|
2015-07-16 12:17:14 +01:00
|
|
|
unsigned blocksize = util_format_get_blocksize(src->format);
|
|
|
|
|
|
|
|
if (blocksize == 8)
|
|
|
|
src_templ.format = PIPE_FORMAT_R16G16B16A16_UINT; /* 64-bit block */
|
|
|
|
else
|
|
|
|
src_templ.format = PIPE_FORMAT_R32G32B32A32_UINT; /* 128-bit block */
|
|
|
|
dst_templ.format = src_templ.format;
|
|
|
|
|
|
|
|
dst_width = util_format_get_nblocksx(dst->format, dst_width);
|
|
|
|
dst_height = util_format_get_nblocksy(dst->format, dst_height);
|
|
|
|
src_width0 = util_format_get_nblocksx(src->format, src_width0);
|
|
|
|
src_height0 = util_format_get_nblocksy(src->format, src_height0);
|
|
|
|
|
|
|
|
dstx = util_format_get_nblocksx(dst->format, dstx);
|
|
|
|
dsty = util_format_get_nblocksy(dst->format, dsty);
|
|
|
|
|
|
|
|
sbox.x = util_format_get_nblocksx(src->format, src_box->x);
|
|
|
|
sbox.y = util_format_get_nblocksy(src->format, src_box->y);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
sbox.z = src_box->z;
|
2015-07-16 12:17:14 +01:00
|
|
|
sbox.width = util_format_get_nblocksx(src->format, src_box->width);
|
|
|
|
sbox.height = util_format_get_nblocksy(src->format, src_box->height);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
sbox.depth = src_box->depth;
|
2014-02-11 00:50:03 +00:00
|
|
|
src_box = &sbox;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2015-07-16 12:17:14 +01:00
|
|
|
src_force_level = src_level;
|
2015-07-30 16:38:44 +01:00
|
|
|
} else if (!util_blitter_is_copy_supported(sctx->blitter, dst, src) ||
|
|
|
|
/* also *8_SNORM has precision issues, use UNORM instead */
|
2015-10-22 22:36:11 +01:00
|
|
|
util_format_is_snorm8(src->format)) {
|
2014-06-04 17:54:37 +01:00
|
|
|
if (util_format_is_subsampled_422(src->format)) {
|
2015-07-16 12:17:14 +01:00
|
|
|
src_templ.format = PIPE_FORMAT_R8G8B8A8_UINT;
|
|
|
|
dst_templ.format = PIPE_FORMAT_R8G8B8A8_UINT;
|
|
|
|
|
|
|
|
dst_width = util_format_get_nblocksx(dst->format, dst_width);
|
|
|
|
src_width0 = util_format_get_nblocksx(src->format, src_width0);
|
|
|
|
|
|
|
|
dstx = util_format_get_nblocksx(dst->format, dstx);
|
2014-06-04 17:54:37 +01:00
|
|
|
|
|
|
|
sbox = *src_box;
|
2015-07-16 12:17:14 +01:00
|
|
|
sbox.x = util_format_get_nblocksx(src->format, src_box->x);
|
|
|
|
sbox.width = util_format_get_nblocksx(src->format, src_box->width);
|
2014-06-04 17:54:37 +01:00
|
|
|
src_box = &sbox;
|
|
|
|
} else {
|
|
|
|
unsigned blocksize = util_format_get_blocksize(src->format);
|
|
|
|
|
|
|
|
switch (blocksize) {
|
|
|
|
case 1:
|
2015-07-16 12:17:14 +01:00
|
|
|
dst_templ.format = PIPE_FORMAT_R8_UNORM;
|
|
|
|
src_templ.format = PIPE_FORMAT_R8_UNORM;
|
2014-06-04 17:54:37 +01:00
|
|
|
break;
|
|
|
|
case 2:
|
2015-07-16 12:17:14 +01:00
|
|
|
dst_templ.format = PIPE_FORMAT_R8G8_UNORM;
|
|
|
|
src_templ.format = PIPE_FORMAT_R8G8_UNORM;
|
2014-06-04 17:54:37 +01:00
|
|
|
break;
|
|
|
|
case 4:
|
2015-07-16 12:17:14 +01:00
|
|
|
dst_templ.format = PIPE_FORMAT_R8G8B8A8_UNORM;
|
|
|
|
src_templ.format = PIPE_FORMAT_R8G8B8A8_UNORM;
|
2014-06-04 17:54:37 +01:00
|
|
|
break;
|
|
|
|
case 8:
|
2015-07-16 12:17:14 +01:00
|
|
|
dst_templ.format = PIPE_FORMAT_R16G16B16A16_UINT;
|
|
|
|
src_templ.format = PIPE_FORMAT_R16G16B16A16_UINT;
|
2014-06-04 17:54:37 +01:00
|
|
|
break;
|
|
|
|
case 16:
|
2015-07-16 12:17:14 +01:00
|
|
|
dst_templ.format = PIPE_FORMAT_R32G32B32A32_UINT;
|
|
|
|
src_templ.format = PIPE_FORMAT_R32G32B32A32_UINT;
|
2014-06-04 17:54:37 +01:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
fprintf(stderr, "Unhandled format %s with blocksize %u\n",
|
|
|
|
util_format_short_name(src->format), blocksize);
|
|
|
|
assert(0);
|
|
|
|
}
|
2013-01-11 11:14:28 +00:00
|
|
|
}
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2014-02-11 02:02:35 +00:00
|
|
|
/* Initialize the surface. */
|
2014-02-09 22:25:06 +00:00
|
|
|
dst_view = r600_create_surface_custom(ctx, dst, &dst_templ,
|
2015-07-16 12:17:14 +01:00
|
|
|
dst_width, dst_height);
|
2014-02-11 02:02:35 +00:00
|
|
|
|
|
|
|
/* Initialize the sampler view. */
|
2015-07-16 12:17:14 +01:00
|
|
|
src_view = si_create_sampler_view_custom(ctx, src, &src_templ,
|
|
|
|
src_width0, src_height0,
|
|
|
|
src_force_level);
|
2014-02-11 02:02:35 +00:00
|
|
|
|
|
|
|
u_box_3d(dstx, dsty, dstz, abs(src_box->width), abs(src_box->height),
|
|
|
|
abs(src_box->depth), &dstbox);
|
|
|
|
|
|
|
|
/* Copy. */
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_begin(ctx, SI_COPY);
|
2014-02-11 02:02:35 +00:00
|
|
|
util_blitter_blit_generic(sctx->blitter, dst_view, &dstbox,
|
2015-07-16 12:17:14 +01:00
|
|
|
src_view, src_box, src_width0, src_height0,
|
2014-11-08 12:02:47 +00:00
|
|
|
PIPE_MASK_RGBAZS, PIPE_TEX_FILTER_NEAREST, NULL,
|
|
|
|
FALSE);
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_end(ctx);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
|
2014-02-11 02:02:35 +00:00
|
|
|
pipe_surface_reference(&dst_view, NULL);
|
|
|
|
pipe_sampler_view_reference(&src_view, NULL);
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|
|
|
|
|
2013-07-30 21:29:28 +01:00
|
|
|
/* For MSAA integer resolving to work, we change the format to NORM using this function. */
|
|
|
|
static enum pipe_format int_to_norm_format(enum pipe_format format)
|
|
|
|
{
|
|
|
|
switch (format) {
|
|
|
|
#define REPLACE_FORMAT_SIGN(format,sign) \
|
|
|
|
case PIPE_FORMAT_##format##_##sign##INT: \
|
|
|
|
return PIPE_FORMAT_##format##_##sign##NORM
|
|
|
|
#define REPLACE_FORMAT(format) \
|
|
|
|
REPLACE_FORMAT_SIGN(format, U); \
|
|
|
|
REPLACE_FORMAT_SIGN(format, S)
|
|
|
|
|
|
|
|
REPLACE_FORMAT_SIGN(B10G10R10A2, U);
|
|
|
|
REPLACE_FORMAT(R8);
|
|
|
|
REPLACE_FORMAT(R8G8);
|
|
|
|
REPLACE_FORMAT(R8G8B8X8);
|
|
|
|
REPLACE_FORMAT(R8G8B8A8);
|
|
|
|
REPLACE_FORMAT(A8);
|
|
|
|
REPLACE_FORMAT(I8);
|
|
|
|
REPLACE_FORMAT(L8);
|
|
|
|
REPLACE_FORMAT(L8A8);
|
|
|
|
REPLACE_FORMAT(R16);
|
|
|
|
REPLACE_FORMAT(R16G16);
|
|
|
|
REPLACE_FORMAT(R16G16B16X16);
|
|
|
|
REPLACE_FORMAT(R16G16B16A16);
|
|
|
|
REPLACE_FORMAT(A16);
|
|
|
|
REPLACE_FORMAT(I16);
|
|
|
|
REPLACE_FORMAT(L16);
|
|
|
|
REPLACE_FORMAT(L16A16);
|
|
|
|
|
|
|
|
#undef REPLACE_FORMAT
|
|
|
|
#undef REPLACE_FORMAT_SIGN
|
|
|
|
default:
|
|
|
|
return format;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-12-07 16:57:10 +00:00
|
|
|
static bool do_hardware_msaa_resolve(struct pipe_context *ctx,
|
|
|
|
const struct pipe_blit_info *info)
|
2013-07-30 21:29:28 +01:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context*)ctx;
|
2013-12-07 16:57:10 +00:00
|
|
|
struct r600_texture *dst = (struct r600_texture*)info->dst.resource;
|
|
|
|
unsigned dst_width = u_minify(info->dst.resource->width0, info->dst.level);
|
|
|
|
unsigned dst_height = u_minify(info->dst.resource->height0, info->dst.level);
|
|
|
|
enum pipe_format format = int_to_norm_format(info->dst.format);
|
2013-07-30 21:29:28 +01:00
|
|
|
unsigned sample_mask = ~0;
|
|
|
|
|
2016-01-15 13:40:19 +00:00
|
|
|
/* Hardware MSAA resolve doesn't work if SPI format = NORM16_ABGR and
|
|
|
|
* the format is R16G16. Use R16A16, which does work.
|
|
|
|
*/
|
|
|
|
if (format == PIPE_FORMAT_R16G16_UNORM)
|
|
|
|
format = PIPE_FORMAT_R16A16_UNORM;
|
|
|
|
if (format == PIPE_FORMAT_R16G16_SNORM)
|
|
|
|
format = PIPE_FORMAT_R16A16_SNORM;
|
|
|
|
|
2013-12-07 16:57:10 +00:00
|
|
|
if (info->src.resource->nr_samples > 1 &&
|
|
|
|
info->dst.resource->nr_samples <= 1 &&
|
|
|
|
util_max_layer(info->src.resource, 0) == 0 &&
|
|
|
|
util_max_layer(info->dst.resource, info->dst.level) == 0 &&
|
|
|
|
info->dst.format == info->src.format &&
|
|
|
|
!util_format_is_pure_integer(format) &&
|
|
|
|
!util_format_is_depth_or_stencil(format) &&
|
|
|
|
!info->scissor_enable &&
|
|
|
|
(info->mask & PIPE_MASK_RGBA) == PIPE_MASK_RGBA &&
|
|
|
|
dst_width == info->src.resource->width0 &&
|
|
|
|
dst_height == info->src.resource->height0 &&
|
|
|
|
info->dst.box.x == 0 &&
|
|
|
|
info->dst.box.y == 0 &&
|
|
|
|
info->dst.box.width == dst_width &&
|
|
|
|
info->dst.box.height == dst_height &&
|
|
|
|
info->dst.box.depth == 1 &&
|
|
|
|
info->src.box.x == 0 &&
|
|
|
|
info->src.box.y == 0 &&
|
|
|
|
info->src.box.width == dst_width &&
|
|
|
|
info->src.box.height == dst_height &&
|
|
|
|
info->src.box.depth == 1 &&
|
|
|
|
dst->surface.level[info->dst.level].mode >= RADEON_SURF_MODE_1D &&
|
2014-06-02 12:51:29 +01:00
|
|
|
!(dst->surface.flags & RADEON_SURF_SCANOUT) &&
|
2015-10-20 23:10:37 +01:00
|
|
|
(!dst->cmask.size || !dst->dirty_level_mask) && /* dst cannot be fast-cleared */
|
2016-02-21 21:49:38 +00:00
|
|
|
!dst->dcc_offset) {
|
2014-06-02 15:14:09 +01:00
|
|
|
si_blitter_begin(ctx, SI_COLOR_RESOLVE |
|
|
|
|
(info->render_condition_enable ? 0 : SI_DISABLE_RENDER_COND));
|
2014-01-11 15:00:50 +00:00
|
|
|
util_blitter_custom_resolve_color(sctx->blitter,
|
2013-07-30 21:29:28 +01:00
|
|
|
info->dst.resource, info->dst.level,
|
|
|
|
info->dst.box.z,
|
|
|
|
info->src.resource, info->src.box.z,
|
2014-01-11 15:00:50 +00:00
|
|
|
sample_mask, sctx->custom_blend_resolve,
|
2013-12-07 16:57:10 +00:00
|
|
|
format);
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_end(ctx);
|
2013-12-07 16:57:10 +00:00
|
|
|
return true;
|
2013-07-30 21:29:28 +01:00
|
|
|
}
|
2013-12-07 16:57:10 +00:00
|
|
|
return false;
|
2013-07-30 21:29:28 +01:00
|
|
|
}
|
|
|
|
|
2012-09-13 17:48:51 +01:00
|
|
|
static void si_blit(struct pipe_context *ctx,
|
2013-12-07 16:57:10 +00:00
|
|
|
const struct pipe_blit_info *info)
|
2012-09-13 17:48:51 +01:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
struct si_context *sctx = (struct si_context*)ctx;
|
2012-09-13 17:48:51 +01:00
|
|
|
|
2013-12-07 16:57:10 +00:00
|
|
|
if (do_hardware_msaa_resolve(ctx, info)) {
|
2012-09-13 17:48:51 +01:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2014-01-11 15:00:50 +00:00
|
|
|
assert(util_blitter_is_blit_supported(sctx->blitter, info));
|
2013-07-30 21:29:28 +01:00
|
|
|
|
2013-08-06 07:48:07 +01:00
|
|
|
/* The driver doesn't decompress resources automatically while
|
|
|
|
* u_blitter is rendering. */
|
2014-01-07 01:51:35 +00:00
|
|
|
si_decompress_subresource(ctx, info->src.resource, info->src.level,
|
|
|
|
info->src.box.z,
|
|
|
|
info->src.box.z + info->src.box.depth - 1);
|
2012-09-13 17:48:51 +01:00
|
|
|
|
2014-09-06 16:07:50 +01:00
|
|
|
if (sctx->screen->b.debug_flags & DBG_FORCE_DMA &&
|
|
|
|
util_try_blit_via_copy_region(ctx, info))
|
|
|
|
return;
|
|
|
|
|
2014-06-02 15:14:09 +01:00
|
|
|
si_blitter_begin(ctx, SI_BLIT |
|
|
|
|
(info->render_condition_enable ? 0 : SI_DISABLE_RENDER_COND));
|
2014-01-11 15:00:50 +00:00
|
|
|
util_blitter_blit(sctx->blitter, info);
|
2014-01-07 01:51:35 +00:00
|
|
|
si_blitter_end(ctx);
|
2012-09-13 17:48:51 +01:00
|
|
|
}
|
|
|
|
|
2013-09-20 14:08:29 +01:00
|
|
|
static void si_flush_resource(struct pipe_context *ctx,
|
2014-03-06 01:44:07 +00:00
|
|
|
struct pipe_resource *res)
|
2013-09-20 14:08:29 +01:00
|
|
|
{
|
2014-03-06 01:44:07 +00:00
|
|
|
struct r600_texture *rtex = (struct r600_texture*)res;
|
|
|
|
|
|
|
|
assert(res->target != PIPE_BUFFER);
|
|
|
|
|
2016-02-21 21:49:38 +00:00
|
|
|
if (!rtex->is_depth && (rtex->cmask.size || rtex->dcc_offset)) {
|
2014-03-06 01:44:07 +00:00
|
|
|
si_blit_decompress_color(ctx, rtex, 0, res->last_level,
|
2015-10-20 23:10:41 +01:00
|
|
|
0, util_max_layer(res, 0), false);
|
2014-03-06 01:44:07 +00:00
|
|
|
}
|
2013-09-20 14:08:29 +01:00
|
|
|
}
|
|
|
|
|
2016-02-24 21:04:47 +00:00
|
|
|
static void si_decompress_dcc(struct pipe_context *ctx,
|
|
|
|
struct r600_texture *rtex)
|
|
|
|
{
|
|
|
|
if (!rtex->dcc_offset)
|
|
|
|
return;
|
|
|
|
|
|
|
|
si_blit_decompress_color(ctx, rtex, 0, rtex->resource.b.b.last_level,
|
|
|
|
0, util_max_layer(&rtex->resource.b.b, 0),
|
|
|
|
true);
|
|
|
|
}
|
|
|
|
|
2015-11-03 18:35:46 +00:00
|
|
|
static void si_pipe_clear_buffer(struct pipe_context *ctx,
|
|
|
|
struct pipe_resource *dst,
|
|
|
|
unsigned offset, unsigned size,
|
|
|
|
const void *clear_value_ptr,
|
|
|
|
int clear_value_size)
|
|
|
|
{
|
|
|
|
struct si_context *sctx = (struct si_context*)ctx;
|
|
|
|
uint32_t dword_value;
|
|
|
|
unsigned i;
|
|
|
|
|
|
|
|
assert(offset % clear_value_size == 0);
|
|
|
|
assert(size % clear_value_size == 0);
|
|
|
|
|
|
|
|
if (clear_value_size > 4) {
|
|
|
|
const uint32_t *u32 = clear_value_ptr;
|
|
|
|
bool clear_dword_duplicated = true;
|
|
|
|
|
|
|
|
/* See if we can lower large fills to dword fills. */
|
|
|
|
for (i = 1; i < clear_value_size / 4; i++)
|
|
|
|
if (u32[0] != u32[i]) {
|
|
|
|
clear_dword_duplicated = false;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!clear_dword_duplicated) {
|
|
|
|
/* Use transform feedback for 64-bit, 96-bit, and
|
|
|
|
* 128-bit fills.
|
|
|
|
*/
|
|
|
|
union pipe_color_union clear_value;
|
|
|
|
|
|
|
|
memcpy(&clear_value, clear_value_ptr, clear_value_size);
|
|
|
|
si_blitter_begin(ctx, SI_DISABLE_RENDER_COND);
|
|
|
|
util_blitter_clear_buffer(sctx->blitter, dst, offset,
|
|
|
|
size, clear_value_size / 4,
|
|
|
|
&clear_value);
|
|
|
|
si_blitter_end(ctx);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Expand the clear value to a dword. */
|
|
|
|
switch (clear_value_size) {
|
|
|
|
case 1:
|
|
|
|
dword_value = *(uint8_t*)clear_value_ptr;
|
|
|
|
dword_value |= (dword_value << 8) |
|
|
|
|
(dword_value << 16) |
|
|
|
|
(dword_value << 24);
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
dword_value = *(uint16_t*)clear_value_ptr;
|
|
|
|
dword_value |= dword_value << 16;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
dword_value = *(uint32_t*)clear_value_ptr;
|
|
|
|
}
|
|
|
|
|
|
|
|
sctx->b.clear_buffer(ctx, dst, offset, size, dword_value, false);
|
|
|
|
}
|
|
|
|
|
2014-01-11 15:00:50 +00:00
|
|
|
void si_init_blit_functions(struct si_context *sctx)
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
{
|
2014-01-11 15:00:50 +00:00
|
|
|
sctx->b.b.clear = si_clear;
|
2015-11-03 18:35:46 +00:00
|
|
|
sctx->b.b.clear_buffer = si_pipe_clear_buffer;
|
2014-01-11 15:00:50 +00:00
|
|
|
sctx->b.b.clear_render_target = si_clear_render_target;
|
|
|
|
sctx->b.b.clear_depth_stencil = si_clear_depth_stencil;
|
|
|
|
sctx->b.b.resource_copy_region = si_resource_copy_region;
|
|
|
|
sctx->b.b.blit = si_blit;
|
|
|
|
sctx->b.b.flush_resource = si_flush_resource;
|
|
|
|
sctx->b.blit_decompress_depth = si_blit_decompress_depth;
|
2016-02-24 21:04:47 +00:00
|
|
|
sctx->b.decompress_dcc = si_decompress_dcc;
|
radeonsi: initial WIP SI code
This commit adds initial support for acceleration
on SI chips. egltri is starting to work.
The SI/R600 llvm backend is currently included in mesa
but that may change in the future.
The plan is to write a single gallium driver and
use gallium to support X acceleration.
This commit contains patches from:
Tom Stellard <thomas.stellard@amd.com>
Michel Dänzer <michel.daenzer@amd.com>
Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The following commits were squashed in:
======================================================================
radeonsi: Remove unused winsys pointer
This was removed from r600g in commit:
commit 96d882939d612fcc8332f107befec470ed4359de
Author: Marek Olšák <maraeo@gmail.com>
Date: Fri Feb 17 01:49:49 2012 +0100
gallium: remove unused winsys pointers in pipe_screen and pipe_context
A winsys is already a private object of a driver.
======================================================================
radeonsi: Copy color clamping CAPs from r600
Not sure if the values of these CAPS are correct for radeonsi, but the
same changed were made to r600g in commit:
commit bc1c8369384b5e16547c5bf9728aa78f8dfd66cc
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Jan 23 03:11:17 2012 +0100
st/mesa: do vertex and fragment color clamping in shaders
For ARB_color_buffer_float. Most hardware can't do it and st/mesa is
the perfect place for a fallback.
The exceptions are:
- r500 (vertex clamp only)
- nv50 (both)
- nvc0 (both)
- softpipe (both)
We also have to take into account that r300 can do CLAMPED vertex colors only,
while r600 can do UNCLAMPED vertex colors only. The difference can be expressed
with the two new CAPs.
======================================================================
radeonsi: Remove PIPE_CAP_OUTPUT_READ
This CAP was dropped in commit:
commit 04e324008759282728a95a1394bac2c4c2a1a3f9
Author: Marek Olšák <maraeo@gmail.com>
Date: Thu Feb 23 23:44:36 2012 +0100
gallium: remove PIPE_SHADER_CAP_OUTPUT_READ
r600g is the only driver which has made use of it. The reason the CAP was
added was to fix some piglit tests when the GLSL pass lower_output_reads
didn't exist.
However, not removing output reads breaks the fallback for glClampColorARB,
which assumes outputs are not readable. The fix would be non-trivial
and my personal preference is to remove the CAP, considering that reading
outputs is uncommon and that we can now use lower_output_reads to fix
the issue that the CAP was supposed to workaround in the first place.
======================================================================
radeonsi: Add missing parameters to rws->buffer_get_tiling() call
This was changed in commit:
commit c0c979eebc076b95cc8d18a013ce2968fe6311ad
Author: Jerome Glisse <jglisse@redhat.com>
Date: Mon Jan 30 17:22:13 2012 -0500
r600g: add support for common surface allocator for tiling v13
Tiled surface have all kind of alignment constraint that needs to
be met. Instead of having all this code duplicated btw ddx and
mesa use common code in libdrm_radeon this also ensure that both
ddx and mesa compute those alignment in the same way.
v2 fix evergreen
v3 fix compressed texture and workaround cube texture issue by
disabling 2D array mode for cubemap (need to check if r7xx and
newer are also affected by the issue)
v4 fix texture array
v5 fix evergreen and newer, split surface values computation from
mipmap tree generation so that we can get them directly from the
ddx
v6 final fix to evergreen tile split value
v7 fix mipmap offset to avoid to use random value, use color view
depth view to address different layer as hardware is doing some
magic rotation depending on the layer
v8 fix COLOR_VIEW on r6xx for linear array mode, use COLOR_VIEW on
evergreen, align bytes per pixel to a multiple of a dword
v9 fix handling of stencil on evergreen, half fix for compressed
texture
v10 fix evergreen compressed texture proper support for stencil
tile split. Fix stencil issue when array mode was clear by
the kernel, always program stencil bo. On evergreen depth
buffer bo need to be big enough to hold depth buffer + stencil
buffer as even with stencil disabled things get written there.
v11 rebase on top of mesa, fix pitch issue with 1d surface on evergreen,
old ddx overestimate those. Fix linear case when pitch*height < 64.
Fix r300g.
v12 Fix linear case when pitch*height < 64 for old path, adapt to
libdrm API change
v13 add libdrm check
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
======================================================================
radeonsi: Remove PIPE_TRANSFER_MAP_PERMANENTLY
This was removed in commit:
commit 62f44f670bb0162e89fd4786af877f8da9ff607c
Author: Marek Olšák <maraeo@gmail.com>
Date: Mon Mar 5 13:45:00 2012 +0100
Revert "gallium: add flag PIPE_TRANSFER_MAP_PERMANENTLY"
This reverts commit 0950086376b1c8b7fb89eda81ed7f2f06dee58bc.
It was decided to refactor the transfer API instead of adding workarounds
to address the performance issues.
======================================================================
radeonsi: Handle PIPE_VIDEO_CAP_PREFERED_FORMAT.
Reintroduced in commit 9d9afcb5bac2931d4b8e6d1aa571e941c5110c90.
======================================================================
radeonsi: nuke the fallback for vertex and fragment color clamping
Ported from r600g commit c2b800cf38b299c1ab1c53dc0e4ea00c7acef853.
======================================================================
radeonsi: don't expose transform_feedback2 without kernel support
Ported from r600g commit 15146fd1bcbb08e44a1cbb984440ee1a5de63d48.
======================================================================
radeonsi: Handle PIPE_CAP_GLSL_FEATURE_LEVEL.
Ported from r600g part of commit 171be755223d99f8cc5cc1bdaf8bd7b4caa04b4f.
======================================================================
radeonsi: set minimum point size to 1.0 for non-sprite non-aa points.
Ported from r600g commit f183cc9ce3ad1d043bdf8b38fd519e8f437714fc.
======================================================================
radeonsi: rework and consolidate stencilref state setting.
Ported from r600g commit a2361946e782b57f0c63587841ca41c0ea707070.
======================================================================
radeonsi: cleanup setting DB_SHADER_CONTROL.
Ported from r600g commit 3d061caaed13b646ff40754f8ebe73f3d4983c5b.
======================================================================
radeonsi: Get rid of register masks.
Ported from r600g commits
3d061caaed13b646ff40754f8ebe73f3d4983c5b..9344ab382a1765c1a7c2560e771485edf4954fe2.
======================================================================
radeonsi: get rid of r600_context_reg.
Ported from r600g commits
9344ab382a1765c1a7c2560e771485edf4954fe2..bed20f02a771f43e1c5092254705701c228cfa7f.
======================================================================
radeonsi: Fix regression from 'Get rid of register masks'.
======================================================================
radeonsi: optimize r600_resource_va.
Ported from r600g commit 669d8766ff3403938794eb80d7769347b6e52174.
======================================================================
radeonsi: remove u8,u16,u32,u64 types.
Ported from r600g commit 78293b99b23268e6698f1267aaf40647c17d95a5.
======================================================================
radeonsi: merge r600_context with r600_pipe_context.
Ported from r600g commit e4340c1908a6a3b09e1a15d5195f6da7d00494d0.
======================================================================
radeonsi: Miscellaneous context cleanups.
Ported from r600g commits
e4340c1908a6a3b09e1a15d5195f6da7d00494d0..621e0db71c5ddcb379171064a4f720c9cf01e888.
======================================================================
radeonsi: add a new simple API for state emission.
Ported from r600g commits
621e0db71c5ddcb379171064a4f720c9cf01e888..f661405637bba32c2cfbeecf6e2e56e414e9521e.
======================================================================
radeonsi: Also remove sbu_flags member of struct r600_reg.
Requires using sid.h instead of r600d.h for the new CP_COHER_CNTL definitions,
so some code needs to be disabled for now.
======================================================================
radeonsi: Miscellaneous simplifications.
Ported from r600g commits 38bf2763482b4f1b6d95cd51aecec75601d8b90f and
b0337b679ad4c2feae59215104cfa60b58a619d5.
======================================================================
radeonsi: Handle PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION.
Ported from commit 8b4f7b0672d663273310fffa9490ad996f5b914a.
======================================================================
radeonsi: Use a fake reloc to sleep for fences.
Ported from r600g commit 8cd03b933cf868ff867e2db4a0937005a02fd0e4.
======================================================================
radeonsi: adapt to get_query_result interface change.
Ported from r600g commit 4445e170bee23a3607ece0e010adef7058ac6a11.
2012-01-06 22:38:37 +00:00
|
|
|
}
|