On RDNA1&2, the driver needs to support both NGG and legacy for
primitives generated query because we can't know that before starting
queries.
To get the query pool results, we check the availability bit wrote by
the SAMPLE_STREAMOUTSTATS packet but the GDS copy was emitted after,
which means the availability bit might be TRUE before the GDS copy is
actually done.
Fix this by emitting the GDS copy before to ensure the availability is
TRUE for both results.
This fixes recent updates in
dEQP-VK.transform_feedback.primitives_generated_query.* because the
tests no longer wait for the fence.
Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23080>
Since 624e799cc3 ("nir: Drop nir_ssa_def::name and nir_register::name"), SSA
defs don't have names, making the name argument unused. Drop it from the
signature and fix the call sites. This was done with the help of the following
Coccinelle semantic patch:
@@
expression A, B, C, D, E;
@@
-nir_ssa_dest_init(A, B, C, D, E);
+nir_ssa_dest_init(A, B, C, D);
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23078>
There are no more producers of legacy atomics so these calls are inert.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23036>
Now p_wqm always kills its operand, so no movs will be created for it.
Long term we want to remove p_wqm in favor of a Definition flag,
so this is also a step in that direction.
Foz-DB Navi21:
Totals from 45351 (33.63% of 134864) affected shaders:
VGPRs: 2099552 -> 2116192 (+0.79%); split: -0.14%, +0.93%
CodeSize: 179530772 -> 179072104 (-0.26%); split: -0.29%, +0.03%
MaxWaves: 1054740 -> 1052262 (-0.23%); split: +0.10%, -0.33%
Instrs: 33238535 -> 33188347 (-0.15%); split: -0.17%, +0.02%
Latency: 451000471 -> 450869384 (-0.03%); split: -0.11%, +0.08%
InvThroughput: 86026785 -> 86286288 (+0.30%); split: -0.11%, +0.41%
VClause: 633291 -> 623920 (-1.48%); split: -1.91%, +0.43%
SClause: 1436708 -> 1431395 (-0.37%); split: -0.60%, +0.23%
Copies: 2166563 -> 2122592 (-2.03%); split: -2.29%, +0.26%
Branches: 706846 -> 706838 (-0.00%); split: -0.00%, +0.00%
PreSGPRs: 1976162 -> 1976592 (+0.02%)
PreVGPRs: 1797409 -> 1794704 (-0.15%)
MaxWaves regressions in Detroit: Become Human MaxWaves seem to be due
to the scheduler choosing to schedule more aggressively.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22956>
The following flakes were found in the latest stress run:
* dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.2_cmdbuffers_resuming
* dEQP-VK.dynamic_rendering.primary_cmd_buff.basic.contents_secondary_primary_cmdbuffers_resuming
Rather than documenting them directly, let's use a broad regular
expression, to match the already-existing `dEQP-VK.dynamic_rendering.basic.*`.
Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23014>
The VkDescriptorSetVariableDescriptorCountLayoutSupport structure should
be in the pNext chain of VkDescriptorSetLayoutSupport.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23009>
Now both radeonsi and radv call it in driver.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23018>
This is a big delete-the-code win. Tested by diff'ing vulkaninfo output
before/after the patch and confirming no changes (other than the driverInfo git
sha and the pipelineCacheUUID).
Note: removes handling for VkDeviceMemoryOverallocationCreateInfoAMD. This was
surely added as a mistake.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22999>
This is already pretty large.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22989>
When a new fragment shader is bound, the PS epilog needs to be
re-emitted, and this allows us to avoid tracking if the pipeline is
dirty.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22981>
Since we expose them for a few games by default now, it would make sense
to have test coverage for them.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21638>
This is unnecessary because the hardware doesn't execute a FS when it
has no effect and it's possible to execute pre-rasterization stages
without a FS.
This might improve depth-only pass performance very slightly because
the number of packets emitted is reduced a bit.
No fossils-db changes.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22848>
This will be useful for shader objects and also because creating and
emitting a noop FS is useless, the hardware doesn't execute it.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22848>
VK_EXT_sample_locations is only supported on < GFX10 due to some weird
issues on recent GPUs. extendedDynamicState3SampleLocationsEnable is
only enabled on GFX6-GFX9 for the same reason.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22947>
ACO need this to be done in nir. Remove the llvm round code
because both radv and radeonsi do this in nir for both aco
and llvm.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22573>
aco does not implement fpow, need nir to lower it
first. llvm will do by itself in the same way, so
we always lower fpow in nir now.
Remove the llvm fpow implementation that has special
handling for the muliplication. It's not used any
more and does not match GLSL spec as fpow(0,0)=NaN
but here we get 0.
There's some pixel changes for gl-radeonsi-stoney:
ror-default 2 (no tolerance), 0 (1% tol.)
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22573>
ACO only support nir_fsin/cos_amd.
There's some pixel changes for gl-radeonsi-stoney trace.
Different pixels:
furmark 61 (no tolerance), 0 (1% tol.)
gimark 93867 (no tolerance), 888 (1% tol.)
tessmark 39 (no tolerance), 0 (1% tol.)
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22573>
aco does not implement these idiv ops.
nir_lower_idiv is for idiv ops <= 32bit and ported from
llvm amdgpu, so llvm do the same.
nir_lower_divmod64 is for 64bit idiv ops.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22573>
committed has to be a constant so there is no need to have a src and
depend on constant folding to remove the i2b.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22963>
Use ACCESS_* flags in call sites instead of GLC/DLC/SLC.
ACCESS_* flags are extended to describe other aspects of memory instructions
like load/store/atomic/smem.
Then add a function that converts the access flags to GLC, DLC, SLC.
The new functions are also usable by ACO.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22770>
... and expect it to behave like ACCESS_NON_TEMPORAL.
Handling ACCESS_NON_TEMPORAL is sufficient.
This was copied from ac_nir_to_llvm.c.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22770>
