With the following test :
dEQP-VK.spirv_assembly.instruction.terminate_invocation.terminate.no_out_of_bounds_load
There is a :
shader_start:
... <- no control flow
g0 = some_alu
g1 = fbl
g2 = broadcast g3, g1
g4 = get_buffer_size g2
... <- no control flow
halt <- on some lanes
g5 = send <surface>, g4
eliminate_find_live_channel will remove the fbl/broadcast because it
assumes lane0 is active at get_buffer_size :
shader_start:
... <- no control flow
g0 = some_alu
g4 = get_buffer_size g0
... <- no control flow
halt <- on some lanes
g5 = send <surface>, g4
But then the instruction scheduler will move the get_buffer_size after
the halt :
shader_start:
... <- no control flow
halt <- on some lanes
g0 = some_alu
g4 = get_buffer_size g0
g5 = send <surface>, g4
get_buffer_size pulls the surface index from lane0 in g0 which could
have been turned off by the halt and we end up accessing an invalid
surface handle.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20765>
Comparisons which produce 32-bit boolean results (0 or 0xFFFFFFFF)
but operate on 16-bit types would first generate a CMP instruction
with W or HF types, before expanding it out. This CMP is a partial
write, which leads us to think the register may contain some prior
contents still. When placed in a loop, this causes its live range
to extend beyond its real life time.
Mark the register with UNDEF first so that we know that no prior
contents exist and need to be preserved.
This affects:
flt32, fge32, feq32, fneu32, ilt32, ult32, ige32, uge32, ieq32, ine32
On one of Cyberpunk 2077's most complex compute shaders, this reduces
the maximum live registers from 696 to 537 (22.8%). Together with the
next patch, Cyberpunk's spills and fills are cut by 10.23% and 9.19%,
respectively.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22835>
Fixes:
```
[829/1646] Compiling C object src/panfrost/vulkan/libpanvk_v6.a.p/panvk_vX_meta_clear.c.o
In function 'panvk_meta_clear_zs_img',
inlined from 'panvk_v6_CmdClearDepthStencilImage' at ../src/panfrost/vulkan/panvk_vX_meta_clear.c:457:7:
../src/panfrost/vulkan/panvk_vX_meta_clear.c:415:26: warning: storing the address of local variable 'view' in '((struct pan_fb_info *)((char *)commandBuffer + 144))[23].zs.view.zs' [-Wdangling-pointer=]
415 | fbinfo->zs.view.zs = &view;
| ~~~~~~~~~~~~~~~~~~~^~~~~~~
../src/panfrost/vulkan/panvk_vX_meta_clear.c: In function 'panvk_v6_CmdClearDepthStencilImage':
../src/panfrost/vulkan/panvk_vX_meta_clear.c:393:26: note: 'view' declared here
393 | struct pan_image_view view = {
| ^~~~
../src/panfrost/vulkan/panvk_vX_meta_clear.c:393:26: note: 'commandBuffer' declared here
[844/1646] Compiling C object src/panfrost/vulkan/libpanvk_v7.a.p/panvk_vX_meta_clear.c.o
In function 'panvk_meta_clear_zs_img',
inlined from 'panvk_v7_CmdClearDepthStencilImage' at ../src/panfrost/vulkan/panvk_vX_meta_clear.c:457:7:
../src/panfrost/vulkan/panvk_vX_meta_clear.c:415:26: warning: storing the address of local variable 'view' in '((struct pan_fb_info *)((char *)commandBuffer + 144))[23].zs.view.zs' [-Wdangling-pointer=]
415 | fbinfo->zs.view.zs = &view;
| ~~~~~~~~~~~~~~~~~~~^~~~~~~
../src/panfrost/vulkan/panvk_vX_meta_clear.c: In function 'panvk_v7_CmdClearDepthStencilImage':
../src/panfrost/vulkan/panvk_vX_meta_clear.c:393:26: note: 'view' declared here
393 | struct pan_image_view view = {
| ^~~~
../src/panfrost/vulkan/panvk_vX_meta_clear.c:393:26: note: 'commandBuffer' declared here
```
Cc: mesa-stable
Suggested-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22829>
pvrsrvkm type sync objects can have a pending state where,
the fence is unsignaled but does not have a valid sync file
due to not yet being submitted to kernel.
The wait function therefore needs to handle these types of syncs
through a spin loop.
This was seen as crashes in dEQP-VK.synchronization.timeline_semaphore.*
Signed-off-by: SoroushIMG <soroush.kashani@imgtec.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22822>
Seems the only thing that really needs this is fpow(0, 0), which should
return NaN, but then gets multiplied with zero. Let's fix that by doing
a bcsel instead of fmul to select the result here. While we're at it,
get rid of the fabs for stop, which isn't needed.
This fixes a piglits failure for most (if not all?) drivers that doesn't
support legacy math rules.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22789>
Prevent GCC warning:
```
[230/1401] Compiling C object src/compiler/nir/libnir.a.p/nir_lower_io_to_vector.c.o
In function 'get_flat_type',
inlined from 'create_new_io_vars' at ../src/compiler/nir/nir_lower_io_to_vector.c:300:10:
../src/compiler/nir/nir_lower_io_to_vector.c:208:14: warning: 'base' may be used uninitialized [-Wmaybe-uninitialized]
208 | return glsl_array_type(glsl_vector_type(base, 4), slots, 0);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../src/compiler/nir/nir_lower_io_to_vector.c: In function 'create_new_io_vars':
../src/compiler/nir/nir_lower_io_to_vector.c:163:24: note: 'base' was declared here
163 | enum glsl_base_type base;
| ^~~~
```
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8957
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22840>
These are unneeded. Events can't be used to indefinitely stall work
like you can with a semaphore or timeline semaphore. The signals
still need to happen so that execution will modify the state that
can be polled from the CPU though.
Fixes dEQP-VK.synchronization.basic.event.single_submit_multi_command_buffer
Fixes: 04fa6c71 ("dzn: Batch command lists together")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22842>
In theory, ResolveSubresourceRegion should be able to resolve just
the depth or just the stencil. In practice, WARP had bugs, which
means that was never tested, so just do it via blits.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22836>
Mike Blumenkrantz