Sampler TS is an hardware optimization that can be used when rendering
to textures. After rendering to a resource with TS enabled, the
texture unit can use this to bypass lookups to empty tiles. This also
means a resolve-in-place can be avoided to flush the TS.
This commit is also an optimization when not using sampler TS, as
resolve-in-place will now be skipped if a resource has no (valid) TS.
Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
This is to make sure that the TS is properly flushed to memory before
rendering to a new surface starts.
Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Sampler TS introduces yet another format enumeration for
renderable+textureable formats. Introduce it into the etnaviv_format
table as another column.
Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Resources only need a resolve-to-itself if their TS is valid for any
level, not just if it happens to be allocated.
Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
This reverts two of the vk_error changes:
reporting unsupported format is common,
and testing non-amdgpu drivers and ignoring them is also common.
Fixes: cd64a4f70 (radv: use vk_error() everywhere an error is returned)
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
BATCH_RESERVED was deleted in commit 2c46a67b41 (i965: Delete
BATCH_RESERVED handling.) The reserved_space field is dead code, and
the comments aren't useful these days.
Having this separate could potentially make programs that rebind atomics
but no other surfaces ever so slightly faster. But it's a tiny amount
of code to add to the existing UBO/SSBO atom, and very related.
The extra atoms have a cost on every draw call, and so dropping some of
them would be nice. This also reclaims a dirty bit.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
We use the same hardware mechanism for both atomic counters and SSBO
atomics, so there's really no benefit to maintaining separate code to
handle each case. Instead, we can just use Rob's shiny new NIR pass to
convert atomic_uints to SSBOs, and delete piles of code.
The ssbo_start section of the binding table becomes a combined ABO and
SSBO section, with ABOs first, then SSBOs.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This fixes the missing AutomaticSize handling in the ABO code, removes
a bunch of duplicated code, and drops an extra layer of wrapping around
brw_emit_buffer_surface_state().
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This should reduce the overhead of adding a BO to the current
list, especially when the list is huge. Also, when a new pipeline
is bound, we only need to update the descriptor, the buffer objects
should already be in the list.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
I don't think we will need a 64-bit unsigned integer for the
dirty flags in the future, and there is still 20 bits left.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
VkDeviceQueueCreateInfo::queueFamilyIndex is an unsigned 32-bit
integer.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
radv_fill_buffer() ensures that the image BO is added to the list.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
radv_fill_buffer() ensures that the image BO is added to the list.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Float rts were always set as unorm instead of float.
Not sure of the consequences, but at least it looks like the blend clamp
would have been enabled, which is against the rules (only eg really bothered
to even attempt to specify this correctly, r600 always used clamp anyway).
Albeit r600 (not r700) setup still looks bugged to me due to never setting
BLEND_FLOAT32 which must be set according to docs...
Not sure if the hw really cares, no piglit change (on eg/juniper).
Reviewed-by: Dave Airlie <airlied@redhat.com>
Both r600 and evergreen used the clamped version, whereas cayman used the
ieee one. I don't think there's a valid reason for this discrepancy, so let's
switch to the ieee version for r600 and evergreen too, since we generally
want to stick to ieee arithmetic.
With this, behavior for both rcp and rsq should now be the same for all of
r600, eg, cm, all using ieee versions (albeit note rsq retains the abs
behavior for everybody, which may not be a good idea ultimately).
Reviewed-by: Dave Airlie <airlied@redhat.com>
r600 used the clamped version for rcp, whereas both evergreen and cayman
used the ieee version. I don't know why that discrepancy exists (it does so
since day 1) but there does not seem to be a valid reason for this, so make
it consistent. This seems now safer than before the previous commit (using
the dx10 clamp bit).
Note that rsq still uses clamped version (as before even though the table
may have suggested otherwise for evergreen) for r600/eg, but not for cayman.
Will be changed separately for better regression tracking...
Reviewed-by: Dave Airlie <airlied@redhat.com>
The docs are not very concise in what this really does, however both
Alex Deucher and Nicolai Hähnle suggested this only really affects instructions
using the CLAMP output modifier, and I've confirmed that with the newly
changed piglit isinf_and_isnan test.
So, with this bit set, if an instruction has the CLAMP modifier bit (which
clamps to [0,1]) set, then NaNs will be converted to zero, otherwise the result
will be NaN.
D3D10 would require this, glsl doesn't have modifiers (with mesa
clamp(x,0,1) would get converted to such a modifier) coupled with a
whatever-floats-your-boat specified NaN behavior, but the clamp behavior
should probably always be used (this also matches what a decomposition into
min(1.0, max(x, 0.0)) would do, if min/max also adhere to the ieee spec of
picking the non-nan result).
Some apps may in fact rely on this, as this prevents misrenderings in
This War of Mine since using ieee muls
(ce7a045fee), without having to use clamped
rcp opcode, which would also fix this bug there.
radeonsi also seems to set this bit nowadays if I see that righ (albeit the
llvm amdgpu code comment now says "Make clamp modifier on NaN input returns 0"
instead of "Do not clamp NAN to 0" since it was changed, which also looks
a bit misleading).
v2: set it in all shader stages.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103544
Reviewed-by: Dave Airlie <airlied@redhat.com>
I believe this is the safe thing to do, especially ever since the driver
actually generates NaNs for muls too.
The ISA docs are not very helpful here, however the dx10 versions will pick
a non-nan result over a NaN one (this is also the ieee754 behavior), whereas
the non-dx10 ones will pick the NaN (verified by newly changed piglit
isinf-and-isnan test).
Other "modern" drivers will most likely do the same.
This was shown to make some difference for bug 103544, albeit it is not
required to fix it.
Reviewed-by: Dave Airlie <airlied@redhat.com>
A lot of cubemap array piglits fail, port the texture type
picking code from radeonsi which seems to fix most of them.
For images I will port the rest of the code.
Fixes:
getteximage-depth gl_texture_cube_map_array-*
fbo-generatemipmap-cubemap array
getteximage-targets cube_array
amongst others.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
A couple failures in piglit tests w/ TF or gl_VertexID + indirect draws.
OTOH all the deqp tests (although they don't test those combinations).
I suspect this could be fixed by a firmware update, but I don't think
there is much we can do in mesa for that.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Number of dwords in MI_FLUSH_DW changed from 4 to 5 in gen8+.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: <mesa-stable@lists.freedesktop.org>
We already have this workaround in OpenGL driver.
See Mesa commit 3cf4fe2219.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: Nanley Chery <nanley.g.chery@intel.com>
Cc: Rafael Antognolli <rafael.antognolli@intel.com>
The MOV instruction can extract bytes to words/double words, and
words/double words to quadwords, but not byte to quadwords.
For unsigned byte to quadword, we can read them as words and AND off the
high byte and extract to quadword in one instruction. For signed bytes,
we need to first sign extend to word and the sign extend that word to a
quadword.
Fixes the following test on CHV, BXT, and GLK:
KHR-GL46.shader_ballot_tests.ShaderBallotBitmasks
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103628
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Fixes the following tests on CHV, BXT, and GLK:
KHR-GL46.shader_ballot_tests.ShaderBallotFunctionBallot
dEQP-VK.spirv_assembly.instruction.compute.uconvert.uint32_to_int64
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103115
Speed up simd16 frontend (default) on avx/avx2 platforms;
fixes performance regression caused by switch to simdlib.
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
Cc: mesa-stable@lists.freedesktop.org
Speed up avx512 platforms; fixes performance regression caused
by swithc to simdlib.
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
Cc: mesa-stable@lists.freedesktop.org
Brian Paul
Wladimir J. van der Laan
Dave Airlie
Kenneth Graunke
Samuel Pitoiset
Roland Scheidegger
Rob Clark
Timothy Arceri
Anuj Phogat
Andres Rodriguez
Matt Turner
Tim Rowley