We have to disable the GLSL unit tests because with asan it runs way too
much code under qemu and times out. Those unit tests have coverage on
x86, anyway.
I also included a vulkan run, which is disabled by default due to timeouts
that I need to sort out still. It should be a useful tool for turnip
devs, though.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9070>
This atom tries to unbind cb0 when it's not used any more (the params &&
params->NumParameters check), but if you transitioned to not having a
gs/tess enabled at all, you'd skip unbinding it. This was mostly
harmless, since if you don't have a GS, why are you looking at GS
constants? However, if a new program came along that didn't use cb0 at
all, we wouldn't end up in this atom to get the disable, and now you have
a GS enabled but a GS constbuf pointing at potentially freed data.
Dereferencing the freed cb0 data ended up happening in freedreno's
fallback UBO upload path with this combination of tests (which execute in
that order):
dEQP-GLES31.functional.shaders.opaque_type_indexing.sampler.uniform.geometry.sampler2darray
dEQP-GLES31.functional.shaders.opaque_type_indexing.ubo.const_literal_fragment
dEQP-GLES31.functional.shaders.opaque_type_indexing.ubo.dynamically_uniform_geometry
and it seems also affected softpipe as well.
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9070>
Use clangs thread-safety annotations to implement a virtual lock
protecting context fields that should only be accessed from driver-
thread. This should let the compiler help us detect problems where
ctx is used unsafely from things that could be called by the fe/st
thread.
This does end up sprinkled far and wide, it would be nice if the
compiler could be a bit smarter about understanding call-graphs
(at least with static fxns), but at least it makes it clear where
things are called from which thread.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9061>
They've all supported it since either forever or Iron Lake which is
equivalent to forever for Vulkan.
From Kenneth Graunke's GitLab review:
"Linear blending of depth buffer data is usually fairly nonsense
(something's 2 meters away? another thing's 6 meters away? let's
just report 4 meters?)...but it's definitely a thing we can do, so
we may as well let apps do it, and trust them not when it doesn't
make sense."
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9110>
Test the sampler->conversion for NULL pointer before dereferencing it.
Fixes: Regressions in VulkanCTS.
Fixes: 226316116c "intel/anv: Fix condition to set MipModeFilter for YUV surface"
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
load_helper_invocation is an Input Builtin, for which the
value should not change during the execution of a shader.
This new pass inserts an is_helper intrinsic before any
demote() instruction and re-uses its value.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9058>
On Intel platforms before Gen6, there is no min or max instruction.
Instead, a comparison instruction (*more on this below) and a SEL
instruction are used. Per other IEEE rules, the regular comparison
instruction, CMP, will always return false if either source is NaN. A
sequence like
cmp.l.f0.0(16) null<1>F g30<8,8,1>F g22<8,8,1>F
(+f0.0) sel(16) g8<1>F g30<8,8,1>F g22<8,8,1>F
will generate the wrong result for min if g22 is NaN. The CMP will
return false, and the SEL will pick g22.
To account for this, the hardware has a special comparison instruction
CMPN. This instruction behaves just like CMP, except if the second
source is NaN, it will return true. The intention is to use it for min
and max. This sequence will always generate the correct result:
cmpn.l.f0.0(16) null<1>F g30<8,8,1>F g22<8,8,1>F
(+f0.0) sel(16) g8<1>F g30<8,8,1>F g22<8,8,1>F
The problem is... for whatever reason, we don't emit CMPN. There was
even a comment in lower_minmax that calls out this very issue! The bug
is actually older than the "Fixes" below even implies. That's just when
the comment was added. That we know of, we never observed a failure
until #4254.
If src1 is known to be a number, either because it's not float or it's
an immediate number, use CMP. This allows cmod propagation to still do
its thing. Without this slight optimization, about 8,300 shaders from
shader-db are hurt on Iron Lake.
Fixes the following piglit tests (from piglit!475):
tests/spec/glsl-1.20/execution/fs-nan-builtin-max.shader_test
tests/spec/glsl-1.20/execution/fs-nan-builtin-min.shader_test
tests/spec/glsl-1.20/execution/vs-nan-builtin-max.shader_test
tests/spec/glsl-1.20/execution/vs-nan-builtin-min.shader_test
Closes: #4254
Fixes: 2f2c00c727 ("i965: Lower min/max after optimization on Gen4/5.")
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Iron Lake and GM45 had similar results. (Iron Lake shown)
total instructions in shared programs: 8115134 -> 8115135 (<.01%)
instructions in affected programs: 229 -> 230 (0.44%)
helped: 0
HURT: 1
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9027>
Mip Mode Filter must be set to MIPFILTER_NONE for Planar YUV surfaces.
Add the missing condition to check for planar format.
Fixes: b24b93d584 "anv: enable VK_KHR_sampler_ycbcr_conversion"
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
After the previous commit, when running the following
deqp-gles31 caselist:
dEQP-GLES31.functional.image_load_store.2d.format_reinterpret.rgba32f_rgba32ui
dEQP-GLES31.functional.image_load_store.2d.format_reinterpret.rgba32f_rgba32i
The second test always fails on gfx10. I don't know why,
but forcing the dcc clear from si_decompress_dcc to use
compute fixes the problem.
The test caselist wasn't failing before because the dcc
disable step was done in si_resource_copy_region, before
calling si_compute_copy_image.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8958>
This was implemented in 1d3bffaf9c,
but missing the WRITE_COMPRESS_ENABLE bit, then disabled by
4dc6ed2a59040f04648eadbffeb1522587d00f3.
This commits reimplements it to:
- avoid disabling dcc when uploading FP16 textures
(see si_use_compute_copy_for_float_formats)
- being able to use compute to upload textures in more cases, rather
than using the blit path
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8958>
The new enum has 3 values:
- SI_CP_DMA_CLEAR_METHOD: equivalent to force_cp_dma = true
- SI_COMPUTE_CLEAR_METHOD: to force the clear to use compute
- SI_AUTO_SELECT_CLEAR_METHOD: equivalent to force_cp_dma = false
No functional change yet, but this will be used later.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8958>
We want glthread to ignore variable-sized parameters if the only thing
we want is to pass the pointer parameter as-is, e.g. when a PBO is bound.
Making it conditional on marshal_sync is kinda hacky, but it's the easiest
path towards handling PBOs, which will use marshal_sync to check whether
a PBO is bound.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9029>
Eric Anholt