Ian Romanick e301817753 brw: Don't lower phis involved in DPAS instructions to scalar ... On my Arc A380 (DG2), this more than doubles the performance of Jeff Bolz's cooperative matrix benchmark. With llama.cpp modified to use cooperative matrix on DG2, performance is improved by 37%. Closes: #15311 Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> Tested-by: Matt Corallo <git@bluematt.me> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41172>		2026-04-27 18:09:16 +00:00
..
tests	brw: Properly set 'desc as register' for SEND in assembler	2025-12-10 19:46:52 +00:00
brw_analysis.cpp
brw_analysis.h	brw: Don't mark_invalid in update_for_reads for non-VGRF destination	2026-03-05 00:02:50 +00:00
brw_analysis_def.cpp	brw: Also check for ADDRESS file in update_for_reads	2026-03-05 00:02:51 +00:00
brw_analysis_liveness.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_analysis_performance.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_asm.c	brw: Remove global variables from brw_asm parser	2026-01-23 23:13:31 +00:00
brw_asm.h
brw_asm_internal.h	brw: Remove global variables from brw_asm parser	2026-01-23 23:13:31 +00:00
brw_asm_tool.c	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_builder.cpp	intel: Rename wm_prog_data to fs_prog_data	2026-02-06 20:51:59 -08:00
brw_builder.h	brw: Make brw_builder::uniform() ignore previous group	2026-02-20 16:50:41 +00:00
brw_cfg.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_cfg.h	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_compile_bs.cpp	brw: Remove various unused fields	2026-04-24 15:04:25 +00:00
brw_compile_cs.cpp	brw: Have brw_nir_apply_key call brw_nir_lower_simd for all stages	2026-04-09 21:02:16 +00:00
brw_compile_fs.cpp	brw: make the program key available on pass_tracker	2026-04-03 12:17:01 +00:00
brw_compile_gs.cpp	brw: make the program key available on pass_tracker	2026-04-03 12:17:01 +00:00
brw_compile_mesh.cpp	brw: add support for < 32bit io values	2026-04-13 14:01:04 +00:00
brw_compile_tcs.cpp	brw: make the program key available on pass_tracker	2026-04-03 12:17:01 +00:00
brw_compile_tes.cpp	brw: make the program key available on pass_tracker	2026-04-03 12:17:01 +00:00
brw_compile_vs.cpp	brw: make the program key available on pass_tracker	2026-04-03 12:17:01 +00:00
brw_compiler.c	brw: lower ifind_msb for Jay	2026-04-10 18:21:21 +00:00
brw_compiler.h	brw: Remove various unused fields	2026-04-24 15:04:25 +00:00
brw_disasm.c	brw: add missing base offset decoding	2026-01-20 21:25:52 +00:00
brw_disasm.h
brw_disasm_info.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_disasm_info.h	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_disasm_tool.c	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_eu.c	brw: wire up MACL	2026-04-08 16:07:35 +00:00
brw_eu.h	brw: Use brw prefix to LSC helpers tied to brw	2026-04-22 18:25:41 +00:00
brw_eu_compact.c	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_eu_defines.h	brw: wire up MACL	2026-04-08 16:07:35 +00:00
brw_eu_emit.c	brw: Don't set saturate for SYNC instruction	2026-04-22 16:06:42 +00:00
brw_eu_inst.h	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_eu_validate.c	brw/validate: Eliminate duplicate integer multiply validation	2026-03-09 19:21:38 +00:00
brw_from_nir.cpp	brw: Pass write back register for ray query messages	2026-04-21 23:16:09 +00:00
brw_generator.cpp	brw: track push constants shader stats	2026-04-21 16:29:14 +00:00
brw_generator.h	brw: Rework UIP and JIP setting code	2026-01-08 22:01:45 +00:00
brw_gram.y	brw/asm: Don't drop accumulator number in the assembler	2026-03-09 19:21:39 +00:00
brw_inst.cpp	brw: Use brw prefix to LSC helpers tied to brw	2026-04-22 18:25:41 +00:00
brw_inst.h	intel/compiler: expose inferred_exec_pipe from scoreboarding	2026-04-09 07:30:01 +00:00
brw_isa_info.h	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_lex.l	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_load_reg.cpp
brw_lower.cpp	intel: Rename wm_prog_data to fs_prog_data	2026-02-06 20:51:59 -08:00
brw_lower_dpas.cpp
brw_lower_fill_spill.cpp	brw: Use brw prefix to LSC helpers tied to brw	2026-04-22 18:25:41 +00:00
brw_lower_integer_multiplication.cpp
brw_lower_logical_sends.cpp	brw: Use brw prefix to LSC helpers tied to brw	2026-04-22 18:25:41 +00:00
brw_lower_pack.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_lower_regioning.cpp	brw/lower_regioning: Allow integer conversions in SEL	2026-03-24 01:31:26 +00:00
brw_lower_scoreboard.cpp	intel/compiler: expose inferred_exec_pipe from scoreboarding	2026-04-09 07:30:01 +00:00
brw_lower_simd_width.cpp	brw: handle lowering of a couple of opcodes	2026-01-12 12:19:21 +00:00
brw_lower_subgroup_ops.cpp
brw_nir.c	brw: Don't lower phis involved in DPAS instructions to scalar	2026-04-27 18:09:16 +00:00
brw_nir.h	brw: add support for < 32bit io values	2026-04-13 14:01:04 +00:00
brw_nir_fence_shared_stores.c	brw: fence SLM writes between workgroups	2026-03-26 22:38:55 +00:00
brw_nir_lower_alpha_to_coverage.c	intel: Rename intel_msaa_flags to intel_fs_config	2026-02-06 20:51:43 -08:00
brw_nir_lower_cooperative_matrix.c	brw: preserve fp_math_ctrl when lowering cmat alu	2026-03-26 13:15:50 +00:00
brw_nir_lower_cs_intrinsics.c	brw/nir_lower_cs_intrinsics: do some math at 16-bit	2026-04-24 17:13:24 +00:00
brw_nir_lower_fs_barycentrics.c	intel: Rename intel_msaa_flags to intel_fs_config	2026-02-06 20:51:43 -08:00
brw_nir_lower_fs_load_output.c	brw/nir_lower_fs_load_output: optimize pixel coord	2026-04-07 19:32:15 +00:00
brw_nir_lower_fsign.py	brw/nir_lower_fsign: try to fix NaN correctness	2026-02-10 18:42:03 +00:00
brw_nir_lower_immediate_offsets.c	brw/nir: enable constant offsets for global_constant_uniform_block_intel	2026-03-06 06:34:43 +00:00
brw_nir_lower_intersection_shader.c	nir: add and use block predecessor helpers	2026-04-08 15:06:32 +00:00
brw_nir_lower_ray_queries.c	intel/compiler: Handle TerminateOnFirstHit in ray query execution	2026-04-06 10:00:05 -07:00
brw_nir_lower_rt_intrinsics.c	anv/brw: remove push constant load emulation from the backend compiler	2026-03-06 06:34:43 +00:00
brw_nir_lower_rt_intrinsics_pre_trace.c
brw_nir_lower_sample_index_in_coord.c	brw/nir_lower_sample_index_in_coord: use helpers	2025-10-09 09:50:20 -04:00
brw_nir_lower_shader_calls.c	nir: add and use block predecessor helpers	2026-04-08 15:06:32 +00:00
brw_nir_lower_storage_image.c	brw/lower_storage_image: use f2f32 instead of unpack_half	2026-02-06 06:12:36 +00:00
brw_nir_lower_texel_address.c	brw: fixup 64bit atomics emulation on 2D array images	2025-11-14 00:01:50 +00:00
brw_nir_lower_texture.c	brw: Use nir_texop_resinfo_intel for query_levels and txs	2026-03-29 12:53:10 +00:00
brw_nir_opt_divergent_atomics.c	brw: Implement divergent atomics fusion optimization (single message approach)	2026-04-03 12:17:01 +00:00
brw_nir_opt_fsat.c	nir: only set fp_math_ctrl if meaningful	2026-02-18 14:04:22 +00:00
brw_nir_rt.c	brw/nir: handle inline_data_intel more like push_data_intel	2026-02-25 10:44:09 +00:00
brw_nir_rt.h	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_nir_rt_builder.h	intel/compiler: Remove unused brw_nir_memclear_global helper	2026-04-09 05:06:05 +00:00
brw_nir_wa_18019110168.c	intel: Rename intel_msaa_flags to intel_fs_config	2026-02-06 20:51:43 -08:00
brw_nir_workarounds.py	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_opt.cpp	intel/compiler/brw: fix potential unsigned overflow	2025-10-24 07:52:09 -07:00
brw_opt_address_reg_load.cpp
brw_opt_algebraic.cpp	brw/algebraic: Allow mixed types in saturate constant folding	2026-03-24 01:31:25 +00:00
brw_opt_bank_conflicts.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_opt_cmod_propagation.cpp	brw/cmod: Don't propagate from CMP to ADD if there is a write between	2026-02-19 21:28:54 +00:00
brw_opt_cmp_flag_destination.cpp	brw: Change the flags written by some CMP	2026-03-24 01:31:26 +00:00
brw_opt_combine_constants.cpp	brw/const: Don't allow type changes when accumulators are involved	2026-04-07 23:37:26 +00:00
brw_opt_copy_propagation.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_opt_cse.cpp	brw/cse: use copies in operands_match instead of in-place modification	2026-02-11 18:43:02 +00:00
brw_opt_dead_code_eliminate.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_opt_fill_spill.cpp	brw: Eliminate duplicate fills	2025-11-26 17:20:13 +00:00
brw_opt_register_coalesce.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_opt_saturate_propagation.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_opt_txf_combiner.cpp	brw: switch to new sampler payload description scheme	2025-10-16 12:08:15 +00:00
brw_opt_virtual_grfs.cpp	intel/brw: Add BRW_DEPENDENCY_INSTRUCTIONS invalidation when instructions are added or removed in brw_opt_split_virtual_grfs()	2026-02-02 14:46:50 +00:00
brw_packed_float.c	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_print.cpp	brw/print: Don't print extra space at the end	2026-02-06 01:00:31 +00:00
brw_private.h	brw: make the program key available on pass_tracker	2026-04-03 12:17:01 +00:00
brw_reg.cpp	brw/algebraic: Allow mixed types in saturate constant folding	2026-03-24 01:31:25 +00:00
brw_reg.h	brw: brw_reg::nr for an accumulator is not part of the offset	2026-04-08 00:36:39 +00:00
brw_reg_allocate.cpp	brw: Use brw prefix to LSC helpers tied to brw	2026-04-22 18:25:41 +00:00
brw_reg_type.c	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_reg_type.h	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_rt.h	anv: Fix Wa_14021821874, Wa_14018813551, Wa_14026600921	2026-04-21 01:38:34 +00:00
brw_sampler.c	nir: Add nir_texop_sparse_residency[_txf]_intel operations	2026-04-10 18:21:21 +00:00
brw_sampler.h	brw: new Xe2 sampler opcodes	2025-10-16 12:08:16 +00:00
brw_schedule_instructions.cpp	brw: use BITSET_LINEAR_ZALLOC	2026-01-26 20:15:43 +00:00
brw_shader.cpp	brw: add Jay-specific SIMD selection rule	2026-04-10 18:21:21 +00:00
brw_shader.h	intel/compiler: implement dummy mov for Wa_18035690555	2026-04-09 07:30:01 +00:00
brw_simd_selection.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
brw_thread_payload.cpp	brw: Remove various unused fields	2026-04-24 15:04:25 +00:00
brw_thread_payload.h	brw: Purge source_depth_to_render_target	2026-03-12 21:40:37 +00:00
brw_validate.cpp	brw/validate: Perform more 3-src validation in brw_validate instead of brw_eu_emit	2026-03-09 19:21:39 +00:00
brw_vue_map.c	brw: Use memset for initializing varying/slot maps	2026-02-16 15:15:38 -08:00
brw_workaround.cpp	intel/compiler: implement macl part of Wa_18035690555	2026-04-19 11:39:10 +00:00
meson.build	brw: Implement divergent atomics fusion optimization (single message approach)	2026-04-03 12:17:01 +00:00
test_def_analysis.cpp	brw: Also check for ADDRESS file in update_for_reads	2026-03-05 00:02:51 +00:00
test_eu_compact.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
test_eu_validate.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
test_helpers.cpp
test_helpers.h	brw: brw_reg::nr for an accumulator is not part of the offset	2026-04-08 00:36:39 +00:00
test_insert_load_reg.cpp
test_lower_regioning.cpp	brw: brw_reg::nr for an accumulator is not part of the offset	2026-04-08 00:36:39 +00:00
test_lower_scoreboard.cpp	brw/scoreboard: Support local implicit out-of-order dependencies	2026-01-21 22:29:28 +00:00
test_opt_algebraic.cpp	brw/algebraic: Allow mixed types in saturate constant folding	2026-03-24 01:31:25 +00:00
test_opt_cmod_propagation.cpp	brw/cmod: Don't propagate from CMP to ADD if there is a write between	2026-02-19 21:28:54 +00:00
test_opt_combine_constants.cpp	brw/const: Don't allow type changes when accumulators are involved	2026-04-07 23:37:26 +00:00
test_opt_copy_propagation.cpp
test_opt_cse.cpp
test_opt_register_coalesce.cpp
test_opt_saturate_propagation.cpp	brw: switch to new sampler payload description scheme	2025-10-16 12:08:15 +00:00
test_simd_selection.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00
test_vf_float_conversions.cpp	intel/compiler: Use SPDX annotations	2026-01-24 20:37:31 +00:00