Since we split misaligned attributes, we could overwrite one of these
VGPRs in the middle of loading the attribute.
For example:
v_add_u32_e32 v4, vcc, s7, v1
s_waitcnt lgkmcnt(0)
buffer_load_dword v4, v4, s[32:35], 0 idxen
buffer_load_dword v5, v4, s[32:35], 0 idxen offset:4
can overwrite the vertex index in the load of the first component.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27920>
For ESO when TES and GS are compiled separately,
BREAK_WAVE_AT_EOI/IA_SWITCH_ON_EOI must be enabled if the GS uses
PrimID.
This fixes
tests/spec/arb_tessellation_shader/execution/gs-primitiveid-instanced.shader_test
on RDNA2 when shader object is enabled with Zink.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28149>
Setting vindex != NULL (even if it's 0) selects a struct.buffer.load opcode,
which causes LLVM to look for "index * stride + offset" in voffset and
moves "index" to vindex (i.e. not 0 anymore), but the bounds checking
(OOB_SELECT) is set to ignore vindex. Setting vindex = NULL selects
a raw.buffer.load opcode.
Fixes: 6b573c00c9 - ac/nir: use ac_build_buffer_load() for SSBO load operations
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/10794
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28119>
The meta shaders for this feature don't compile with LLVM because of 5/8
component SSBO stores. I'm not sure this was ever expected to work.
This seemed to break vkd3d-proton.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28108>
With RADV, when VS/TES and FS are compiled separately, the PrimitiveId
is exported unconditionally because it's not possible to know if the
FS reads it or not. This happens with fast-link GPL and shader object.
Though, the PrimitiveID should be ignored when it's implicitly exported
because otherwise the stream output LDS offset is incorrect.
This fixes a bunch of failures with transform feedback and Zink/RADV
when shader object is enabled on RDNA3.
Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27981>
When a fragment shader doesn't have any written color outputs it
doesn't need a PS epilog because it's unnecessary. Though, with ESO
the driver still needs to emit the default blend state like graphics
pipelines.
This fixes a bunch of flakes with ESO and Zink.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28008>
Internal operations that set the custom blend mode are monolithic
pipelines only, and the value should always be zero.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28008>
Summary:
- ensure headers used outside runtime are included in dependency source
- drop redundant idep_vulkan_common_entrypoints_h
- drop redundant icd side tricks for the order of header gen
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Yonggang Luo <luoyonggang@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28066>
VK_EXT_shader_object is now supported on all GPUs! It's still not
enabled by default because it needs more testing and more fixing (Renoir
still hangs for some reasons and Zink/RADV on Polaris10 still has
flakes).
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27880>
NGG streamout isn't supported on GFX10/GFX10.3, so NGG GS must be
disabled when pre-rasterization stages are compiled separately because
it's not possible to know that when compiling VS/TES only.
Also GFX10/GFX10.3 needs to disable NGG when extreme geometry are used
with tessellation and we can't know the previous stage if the GS is
compiled separately.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27880>
We can gather the args used by all other stages and use that to optimize
the traversal shader.
Totals from 4 (1.06% of 379) affected shaders:
Instrs: 2912 -> 2864 (-1.65%)
CodeSize: 15424 -> 15232 (-1.24%)
Latency: 69342 -> 69074 (-0.39%)
InvThroughput: 11558 -> 11512 (-0.40%)
Copies: 712 -> 676 (-5.06%)
Branches: 152 -> 148 (-2.63%)
PreVGPRs: 324 -> 316 (-2.47%)
Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27137>
DebugMarkerSetObjectNameEXT is just a less powerful version of
SetDebugUtilsObjectNameEXT. Fixes the objectType cast warning as well.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27975>
This prevents potential app bugs causing GPU hangs and other issues.
Note that the draw ID is undefined in mesh shaders when also
using a task shader.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27879>
Fixes usage of uninitialised value in dead_space/065515347dca4851 and
other dead_space shaders.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Fixes: def0c275c4 ("aco: Eliminate SCC copies when possible.")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28011>
