This code uses a vector shift which has to be emulated on x86 unless
there's AVX2. Luckily in some cases we can actually avoid the shift
altogether, so do that.
Also make sure we hit the fast lp_build_conv() path when applicable,
albeit that's quite the hack...
That said, this path is taken for AoS sampling for small unorm (smaller
than rgba8) formats, and it is completely hopeless even with those
changes, with or without AVX.
(Probably should have some code similar to the one in the llvmpipe fs
backend code, using bit replication to extend to rgba8888 - rounding
is not quite 100% accurate but if it's good enough there it should be
here as well.)
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
If we only feed one source vector at a time, we cannot use pack intrinsics
(as we only have a 64bit destination dst vector). lp_bld_conv_auto is
specifically designed to alter the length and number of destination vectors,
so this works just fine (if we use single source vectors at a time, afterwards
we immediately reassemble the vectors).
For AVX though this isn't really possible, since we expect 128bit output
already for a single 256bit input. (One day we should handle AVX2 which again
would need multiple inputs, however there's the problem that we get different
ordered output there and we don't want to reorder, so would need to be able
to tell build_conv to handle upper and lower halfs independently.)
A similar strategy would probably work for 32->8bit too (if it doesn't hit
the special case) but I'm going to try something different for that...
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
simd instruction sets usually have comparisons for equal, not unequal.
So use a different comparison against the mask itself - which also means
we don't need a all-zero as well as a all-one (for the pxor) reg.
Also add code to avoid scalar expansion of i1 values which we definitely
shouldn't do. There's problems with this though with llvm select
interaction, so it's disabled (basically using llvm select instead of
intrinsics may still produce atrocious code, even in cases where we
figured it should not, albeit I think this could probably be fixed
with some better selection of optimization passes, but I have zero
idea there really).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
The specification section 9.4 says :
When an application attempts to create many pipelines in a single
command, it is possible that some subset may fail creation. In that
case, the corresponding entries in the pPipelines output array will
be filled with VK_NULL_HANDLE values. If any pipeline fails
creation (for example, due to out of memory errors), the
vkCreate*Pipelines commands will return an error code. The
implementation will attempt to create all pipelines, and only
return VK_NULL_HANDLE values for those that actually failed.
Fixes :
dEQP-VK.api.object_management.alloc_callback_fail_multiple.graphics_pipeline
dEQP-VK.api.object_management.alloc_callback_fail_multiple.compute_pipeline
v2: C is hard let's go shopping (Lionel)
v3: Remove unnecessary condition in for loops (Lionel)
v4: Document why we return on first failure (Eduardo)
Move i declaration inside for() (Eduardo)
v5: Move array cleanup out of loop (Jason)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
In OpenGL 3.0 and later it is legal to make a context current without
a default framebuffer.
This has been broken since DRI3 support was introduced.
Cc: "13.0 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Draw calls no longer flush SDMA IBs. r600_need_dma_space is
responsible for synchronizing execution between both IBs.
Initial buffer clears and fast clears will stay unflushed in the SDMA IB
(up to 64 MB) as long as the GFX IB isn't flushed either.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Call r600_dma_emit_wait_idle only when there is a possibility of
a read-after-write hazard. Buffers not yet used by the SDMA IB don't
have to wait.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
r600_dma_emit_wait_idle is going away in its current form.
The only difference is that the moved code is executed before DMA calls
instead of after them.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
This fixes the mistake introduced in commit
b6737a8bcd
Signed-off-by: Nayan Deshmukh <nayan26deshmukh@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Old test caused breakage with llvm-svn (4.0.0svn), and not needed as
the minimum required llvm version for swr is 3.6.
Reviewed-by: George Kyriazis <george.kyriazis@intel.com>
wrap lp_bld_type.h around extern "C".
Windows decorates global variables, so when used from .cpp files, need
to use an undecorated version.
Also, removed related and unneeded code from swr_screen.cpp
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
This seems to fix the GPU hangs caused by:
commit ed3190b3f3
Author: Marek Olšák <marek.olsak@amd.com>
Date: Sun Nov 13 18:41:43 2016 +0100
radeonsi: don't export ClipVertex and ClipDistance[] if clipping is disabled
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99219
Tested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
GLSL compilation now takes 24% less time with the Gallium noop driver.
I used my shader-db for the measurement. The difference for the whole
radeonsi driver can be ~10%.
The generated TGSI is mostly the same. For example, the compilation success
rate with a TGSI->GCN bytecode converter without any optimizations is
the same. Note that glsl_to_tgsi does its own copy propagation and simple
register allocation.
shader-db GCN report:
- Talos spills fewer SGPRs.
- DOTA 2 spills more SGPRs.
- The average shader-db score is better, but it's just due to randomness.
29045 shaders in 17564 tests
Totals:
SGPRS: 1325929 -> 1325017 (-0.07 %)
VGPRS: 1010808 -> 1010172 (-0.06 %)
Spilled SGPRs: 1432 -> 1399 (-2.30 %)
Spilled VGPRs: 93 -> 92 (-1.08 %)
Private memory VGPRs: 688 -> 688 (0.00 %)
Scratch size: 2540 -> 2484 (-2.20 %) dwords per thread
Code Size: 39336732 -> 39342936 (0.02 %) bytes
Max Waves: 217937 -> 217969 (0.01 %)
Reviewed-by: Eric Anholt <eric@anholt.net>
We need to move this to the shared layer.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
If the VUE map has slots at the end which the shader does not write,
then we'd "flush" (constructing an URB write) on the last output it
actually wrote. Then, we'd construct another SEND with EOT, but with
no actual payload data. That's not legal.
For example, SSO programs have clip distance slots allocated no matter
what, but the shader may not write them. If it doesn't write any user
defined varyings, then the clip distance slots will be the last ones.
Found while debugging
dEQP-VK.tessellation.shader_input_output.gl_position_vs_to_tcs_to_tes
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
We can use this to track various features that may or may not be supported
by the hw / kernel. Currently, we usually do this by checking the generation
and supported command parser versions in various places thoughtout the driver
code. With this patch, we centralize all these checks in just once place at
screen creation time, then we just query the bitfield wherever we need to
check if a particular feature is supported.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Moving the test to the screen places it alongside the other global HW
feature tests that want to be shared between contexts.
Also, we need to know if we support pipelined register writes at
screen creation time so that we can tell if we can expose OpenGL 4.0
in gen7.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>