AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
122,070 Commits 1 Branch 0 Tags
9f005f1f850710ea456f9847b1d247aaa8f0d6d2
Commit Graph

81 Commits

Author SHA1 Message Date
Samuel Pitoiset 9f005f1f85 radv: enable lowering of GS intrinsics for the LLVM backend 
This replaces emit_vertex with:
   if (vertex_count < max_vertices) {
      emit_vertex_with_counter vertex_count ...
      vertex_count += 1
   }

Which is exactly what NIR->LLVM was doing but at NIR level. This
pass is already called by ACO.

pipeline-db changes on GFX10:
Totals from affected shaders:
SGPRS: 1952 -> 1912 (-2.05 %)
VGPRS: 2112 -> 2044 (-3.22 %)
Code Size: 189368 -> 185620 (-1.98 %) bytes
Max Waves: 494 -> 491 (-0.61 %)

No pipeline-db changes on other generations.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4182>
 2020-04-08 08:24:05 +02:00
Samuel Pitoiset 3cd5450df5 ac/nir: split 16-bit SSBO stores on GFX6 
Due to possible alignment issues, make sure to split stores of
16-bit vectors.

Doom Eternal requires storageBuffer16BitAccess.

Cc: 20.0 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4339>
 2020-04-03 08:01:28 +00:00
Samuel Pitoiset 55fdcc03de ac/nir: split 16-bit load/store to global memory on GFX6 
Due to possible alignment issues, make sure to split loads/stores
of 16-bit vectors.

Doom Eternal requires storageBuffer16BitAccess.

Cc: 20.0 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4339>
 2020-04-03 08:01:28 +00:00
Samuel Pitoiset c6bf1597d1 ac/nir: split 8-bit SSBO stores on GFX6 
Due to possible alignment issues, make sure to split stores of
8-bit vectors.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4339>
 2020-04-03 08:01:28 +00:00
Samuel Pitoiset 433f3380eb ac/nir: split 8-bit load/store to global memory on GFX6 
Due to possible alignment issues, make sure to split loads/stores
of 8-bit vectors.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4339>
 2020-04-03 08:01:28 +00:00
Eric Engestrom 79af30768d meson: inline inc_common 
Let's make it clear what includes are being added everywhere, so that
they can be cleaned up.

Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4360>
 2020-03-28 21:36:54 +01:00
Samuel Pitoiset ba2ec1f369 ac/nir: use llvm.amdgcn.rcp in ac_build_fdiv() 
Instead of emitting 1.0 / x which includes a slow division that
LLVM doesn't always optimize even if the metadata is correctly set.

No pipeline-db changes with VEGA10/LLVM 9.

pipeline-db (VEGA10/LLVM 10):
Totals from affected shaders:
SGPRS: 6672 -> 6672 (0.00 %)
VGPRS: 6652 -> 6652 (0.00 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Code Size: 561780 -> 561692 (-0.02 %) bytes
Max Waves: 1043 -> 1043 (0.00 %)

pipeline-db (VEGA10/LLVM 11 - 92744f62478):
Totals from affected shaders:
SGPRS: 84608 -> 83768 (-0.99 %)
VGPRS: 106768 -> 106636 (-0.12 %)
Spilled SGPRs: 1625 -> 1713 (5.42 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Code Size: 10850936 -> 10726712 (-1.14 %) bytes
Max Waves: 3152 -> 3180 (0.89 %)

LLVM 11 (master) is more affected than previous versions, but
based on the small impact with LLVM 9/10, I decided to emit it
unconditionally.

Cc: 20.0 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4326>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4326>
 2020-03-27 08:05:43 +01:00
Samuel Pitoiset d548384fc6 ac/nir: use llvm.amdgcn.rsq for nir_op_frsq 
Instead of emitting 1.0 / sqrt(x) which includes a slow division that
LLVM doesn't always optimize even if the metadata is correctly set.

pipeline-db (VEGA10/LLVM 9):
Totals from affected shaders:
SGPRS: 16872 -> 16864 (-0.05 %)
VGPRS: 15320 -> 15464 (0.94 %)
Spilled SGPRs: 2021 -> 2133 (5.54 %)
Code Size: 1915464 -> 1917476 (0.11 %) bytes
Max Waves: 641 -> 639 (-0.31 %)

pipeline-db (VEGA10/LLVM 10):
Totals from affected shaders:
SGPRS: 43936 -> 44120 (0.42 %)
VGPRS: 41776 -> 41972 (0.47 %)
Spilled SGPRs: 875 -> 875 (0.00 %)
Code Size: 4468164 -> 4468120 (-0.00 %) bytes
Max Waves: 2412 -> 2414 (0.08 %)

pipeline-db (VEGA10/LLVM 11 - 92744f62478):
Totals from affected shaders:
SGPRS: 60096 -> 60096 (0.00 %)
VGPRS: 63552 -> 63648 (0.15 %)
Spilled SGPRs: 6135 -> 6117 (-0.29 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Code Size: 6252996 -> 6249772 (-0.05 %) bytes
Max Waves: 2324 -> 2337 (0.56 %)

LLVM 11 (master) is more affected than previous versions, but
based on the small impact with LLVM 9/10, I decided to emit it
unconditionally.

Cc: 20.0 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4326>
 2020-03-27 07:45:47 +01:00
Samuel Pitoiset 66426ce119 ac/nir: use llvm.amdgcn.rcp for nir_op_frcp 
Instead of emitting 1.0 / x which includes a slow division that
LLVM doesn't always optimize even if the metadata is correctly set.

pipeline-db (VEG10/LLVM 9):
Totals from affected shaders:
SGPRS: 50384 -> 50312 (-0.14 %)
VGPRS: 42572 -> 42696 (0.29 %)
Spilled SGPRs: 1372 -> 1372 (0.00 %)
Code Size: 5692040 -> 5691428 (-0.01 %) bytes
Max Waves: 3954 -> 3951 (-0.08 %)

pipeline-db (VEG10/LLVM 10):
Totals from affected shaders:
SGPRS: 78512 -> 78464 (-0.06 %)
VGPRS: 62408 -> 62484 (0.12 %)
Spilled SGPRs: 1502 -> 1502 (0.00 %)
Code Size: 8106188 -> 8103372 (-0.03 %) bytes
Max Waves: 7759 -> 7753 (-0.08 %)

pipeline-db (VEGA10/LLVM 11 - 92744f62478):
Totals from affected shaders:
SGPRS: 112760 -> 113232 (0.42 %)
VGPRS: 111132 -> 110568 (-0.51 %)
Spilled SGPRs: 5870 -> 5940 (1.19 %)
Spilled VGPRs: 650 -> 652 (0.31 %)
Code Size: 11887232 -> 11561744 (-2.74 %) bytes
Max Waves: 8964 -> 9015 (0.57 %)

LLVM 11 (master) is more affected than previous versions, but
based on the small impact with LLVM 9/10, I decided to emit it
unconditionally.

Cc: 20.0 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4326>
 2020-03-27 07:45:43 +01:00
Pierre-Eric Pelloux-Prayer 5533c41541 ac: fix ac_build_is_helper_invocation when postponed_kill is null 
If there was no demote() in the shader, ac_build_is_helper_invocation
behaves exactly the same as ac_build_load_helper_invocation, i.e.
the helper lanes are the same as they were at the beginning of the shader.

Fixes: de57ea2a3d ("amd/llvm: implement nir_intrinsic_demote(_if) and nir_intrinsic_is_helper_invocation")
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4301>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4301>
 2020-03-25 08:19:38 +01:00
Samuel Pitoiset 7ac8bb33cd radv/llvm: fix subgroup shuffle for chips without bpermute 
bpermute only exists on GFX8+ and only with Wave32 on GFX10. Instead
we have to use readlane with a waterfall loop to defeat the LLVM
backend.

This fixes DOOM Eternal which requires subgroup shuffle.

Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4284>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4284>
 2020-03-23 14:19:03 +00:00
Marek Olšák 303842b2db ac: fix fast division 
This stopped working with LLVM 11 and might occasionally have been broken
on older LLVM, because the metadata was set on the mul, not on the rcp.

Cc: 19.3 20.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4268>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4268>
 2020-03-21 22:34:17 +00:00
Bas Nieuwenhuizen 8e4e2cedcf amd/llvm: Fix divergent descriptor regressions with radeonsi. 
piglit/bin/arb_bindless_texture-limit -auto -fbo:
  Needed to deal with non-NULL dynamic_index without deref in tex instructions.

piglit/bin/shader_runner tests/spec/arb_bindless_texture/execution/images/multiple-resident-images-reading.shader_test -auto:
  Need to deal with non-deref images in enter_waterfall_imae.

Fixes: b83c9aca4a "amd/llvm: Fix divergent descriptor indexing. (v3)"
Acked-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4191>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4191>
 2020-03-17 22:53:16 +01:00
Marek Olšák 8dc5e174c7 ac: don't set old denormals flags with LLVM >= 11 
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4196>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4196>
 2020-03-17 20:47:48 +00:00
Marek Olšák 63a5051ea6 ac: set new LLVM denormal flags 
See: https://reviews.llvm.org/D71358

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4196>
 2020-03-17 20:47:48 +00:00
Bas Nieuwenhuizen b83c9aca4a amd/llvm: Fix divergent descriptor indexing. (v3) 
There are multiple LLVM passes that very much move the
intrinsic using the descriptor outside of the loop, defeating
the entire point of creating the loop.

Defeat the optimizer by  splitting the break into a separate
if-statement and putting an optimization barrier on the bool
in between.

v2: Move from a callback based system to begin/end loop.
    This does not make it significantly less intrusive but
    is a bit nicer with all the extra struct and callback
    stubs.
v3: Deal with non-divergent values in divergent path.

Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2160
Fixes: 028ce52739 "radv: Add non-uniform indexing lowering."
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4109>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4109>
 2020-03-12 16:12:02 +00:00
Samuel Pitoiset cc320ef9af ac/llvm: add missing optimization barrier for 64-bit readlanes 
Otherwise, LLVM optimizes it but it's actually incorrect.

Fixes: 0f45d4dc2b ("ac: add ac_build_readlane without optimization barrier")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3585>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3585>
 2020-03-12 08:46:42 +01:00
Marek Olšák fc65df5651 ac: add a bug workaround for the 100% NGG culling case 
Fixes: 8db00a51f8 - radeonsi/gfx10: implement NGG culling for 4x wave32 subgroups
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4079>
 2020-03-09 16:08:11 -04:00
Daniel Schürmann 61fb17e8d7 amd: join emit_kill() from radv and radeonsi in ac_nir_to_llvm 
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4047>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4047>
 2020-03-09 12:29:32 +00:00
Daniel Schürmann de57ea2a3d amd/llvm: implement nir_intrinsic_demote(_if) and nir_intrinsic_is_helper_invocation 
The current implementation uses a temporary helper variable
to ensure correct behavior until LLVM provides an intrinsic.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4047>
 2020-03-09 12:29:32 +00:00
Pierre-Eric Pelloux-Prayer 771f16cf61 radeonsi: remove AMD_DEBUG=sisched option 
sisched is not maintained anymore in LLVM.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4059>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4059>
 2020-03-06 11:35:12 +01:00
Samuel Pitoiset 9e5d2a73c5 ac/llvm: flush denorms for nir_op_fmed3 on GFX8 and older gens 
The hardware doesn't flush denorms, exactly like fmin/fmax, so
we have to do it manually. This doesn't fix anything known.

Fixes: d6a07732c9 ("ac: use llvm.amdgcn.fmed3 intrinsic for nir_op_fmed3")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3962>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3962>
 2020-02-27 08:04:33 +01:00
Samuel Pitoiset 30ac733680 ac/llvm: fix 16-bit fmed3 on GFX8 and older gens 
16-bit med3 is only supported on GFX9+.

Fixes dEQP-VK.spirv_assembly.instruction.amd_trinary_minmax.mid3.f16.*.

Fixes: d6a07732c9 ("ac: use llvm.amdgcn.fmed3 intrinsic for nir_op_fmed3")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3962>
 2020-02-27 08:04:30 +01:00
Samuel Pitoiset 50b8c25274 ac/llvm: fix 64-bit fmed3 
Lower 64-bit fmed3 because LLVM doesn't expose an intrinsic.

Fixes dEQP-VK.spirv_assembly.instruction.amd_trinary_minmax.mid3.f64.*.

Fixes: d6a07732c9 ("ac: use llvm.amdgcn.fmed3 intrinsic for nir_op_fmed3")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3962>
 2020-02-27 08:04:28 +01:00
Samuel Pitoiset 6f4c300919 ac/llvm: implement VK_AMD_shader_explicit_vertex_parameter 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3578>
 2020-01-29 09:49:50 +00:00
Samuel Pitoiset a31bcf2be6 ac/llvm: fix missing casts in ac_build_readlane() 
Because ac_build_optimization_barrier() overwrites the original
src_type, we have to keep track of it before emitting that barrier.
Otherwise, wrong conversions are expected for pointers or small
bitsizes.

By doing this, we no longer need to do the cast dance in
ac_build_readlane_no_opt_barrier(), it was just necessary for
ac_build_optimization_barrier().

This fixes a bunch of crashes with subgroups related tests when
RADV_DEBUG=checkir is enabled, and it also fixes a compiler crash
with The Surge 2.

Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2395
Fixes: 0f45d4dc2b ("ac: add ac_build_readlane without optimization barrier")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3535>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3535>
 2020-01-24 07:40:07 +01:00
Samuel Pitoiset 9e477d79b7 ac/nir: add support for nir_texop_fragment_{mask}_fetch 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3304>
 2020-01-23 10:48:02 +00:00
Marek Olšák 4e4b2d13f0 ac: add helper ac_build_triangle_strip_indices_to_triangle 
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2020-01-20 16:16:11 -05:00
Marek Olšák 0f45d4dc2b ac: add ac_build_readlane without optimization barrier 
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2020-01-20 16:16:11 -05:00
Marek Olšák 77393cf39b ac: add prefix bitcount functions 
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2020-01-20 16:16:11 -05:00
Marek Olšák eeb4a11c11 ac/cull: don't read Position.Z if it's not needed for culling 
It could be NULL.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2020-01-15 15:06:20 -05:00
Jason Ekstrand d3737002ee nir/lower_atomics_to_ssbo: Also lower barriers 
This is more correct for a pass which is supposed to completely lower
away atomic counters.  It also lets us stop supporting atomic counter
barriers in most of the drivers.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3307>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3307>
 2020-01-13 17:23:47 +00:00
Jason Ekstrand e40b11bbcb nir: Rename nir_intrinsic_barrier to control_barrier 
This is a more explicit name now that we don't want it to be doing any
memory barrier stuff for us.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3307>
 2020-01-13 17:23:47 +00:00
Jason Ekstrand 60097cc840 nir: Add a new memory_barrier_tcs_patch intrinsic 
Right now, it's implemented as a no-op for everyone.  For most drivers,
it's a switch case in the NIR -> whatever which just breaks.  For ir3,
they already have code to delete tessellation barriers so we just add a
case to also delete memory_barrier_tcs_patch.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3307>
 2020-01-13 17:23:47 +00:00
Timur Kristóf eccac46cdc ac/llvm: Fix ac_build_reduce in wave32 mode. 
Previously, when cluster_size was set to 0, it always worked as if
the cluster size was 64. This commit fixes it in wave32 mode by
changing to work as if the cluster size was set to 32.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
 2020-01-10 12:30:44 +01:00
Samuel Pitoiset e77ff89914 amd/llvm: handle nir_intrinsic_image_deref_{load,store} with lod 
Use image_load_mip and image_store_mip respectively if the lod
parameter isn't zero.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
 2020-01-09 07:58:33 +01:00
Marek Olšák 9b71041627 ac: add ac_build_s_endpgm 
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2020-01-08 16:03:48 -05:00
Marek Olšák 1c44480538 ac: add 128-bit bitcount 
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2020-01-08 16:00:41 -05:00
Marek Olšák d1c8aeb24f ac: unify primitive export code 
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2020-01-08 16:00:38 -05:00
Marek Olšák 1c77a18cc2 ac: unify build_sendmsg_gs_alloc_req 
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2020-01-08 16:00:36 -05:00
Marek Olšák f90cbd18ff ac: fix the return value in cull_bbox when bbox culling is disabled 
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3095>
 2019-12-16 20:06:07 +00:00
Marek Olšák e5e3ffa6b9 ac: fix ac_get_i1_sgpr_mask for Wave32 
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3095>
 2019-12-16 20:06:07 +00:00
Samuel Pitoiset a0f1a5fa05 ac/nir: fix out-of-bound access when loading constants from global 
Global load/store instructions can't know if the offset is
out-of-bound because they don't use descriptors (no range).

Fix this by clamping the offset for arrays that are indexed
with a non-constant offset that's greater or equal to the array
size.

This fixes VM faults and GPU hangs with Dead Rising 4.

Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2148
Fixes: 71a6794200 ("ac/nir: Enable nir_opt_large_constants")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
 2019-12-12 10:12:56 +00:00
Samuel Pitoiset f63a3132e8 ac/llvm: fix atomic var operations if source isn't a deref 
Fixes some CTS regressions.

Fixes: e61a826f39 ("ac/llvm: fix pointer type for global atomics")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
 2019-12-03 09:41:33 +01:00
Rhys Perry a814f3d8a7 ac/llvm: improve sync scope for global atomics 
Stronger ordering is implemented in SPIRV->NIR with barriers.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
 2019-12-02 10:48:27 +00:00
Rhys Perry e61a826f39 ac/llvm: fix pointer type for global atomics 
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
 2019-12-02 10:48:18 +00:00
Samuel Pitoiset c105e6169c radv,ac/nir: lower deref operations for shared memory 
This shouldn't introduce any functional changes for RadeonSI
when NIR is enabled because these operations are already lowered.

pipeline-db (NAVI10/LLVM):
SGPRS: 9043 -> 9051 (0.09 %)
VGPRS: 7272 -> 7292 (0.28 %)
Code Size: 638892 -> 621628 (-2.70 %) bytes
LDS: 1333 -> 1331 (-0.15 %) blocks
Max Waves: 1614 -> 1608 (-0.37 %)

Found this while glancing at some F12019 shaders.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
 2019-11-29 21:58:18 +01:00
Bas Nieuwenhuizen e09426ad6b amd/llvm: Refactor ac_build_scan. 
Split out the logic for exclusive scans into a separate function
that makes clear what it does instead of having this opaque 60
line if.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
 2019-11-28 11:35:11 +01:00
Samuel Pitoiset 52aadbfd04 ac/llvm: convert src operands to pointers if necessary 
To avoid generating invalid LLVM IR when both operands don't have
the same type. This might happen when performing pointer comparisons
with SPIRV 1.4.

Fixes invalid LLVM IR for:
dEQP-VK.spirv_assembly.instruction.spirv1p4.opptrequal.variable_pointers_ssbo_equal
dEQP-VK.spirv_assembly.instruction.spirv1p4.opptrnotequal.variable_pointers_ssbo_not_equal

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
 2019-11-28 08:26:51 +01:00
Marek Olšák 42318f9197 ac/nir: don't rely on data.patch for tess factors 
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
 2019-11-27 19:28:10 -05:00