fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-22 00:30:13 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1
mesa/src/intel/compiler
History
Lionel Landwerlin e5dfff0946 intel/fs: reduce liveness of variables in lowering passes 
When lowering a single instruction with a destination VGRF to 2 or
more, the VGRF is now considered partially written by each generated
instruction and that increases its liveness especially in loops. Thus
potentially increasing the number of spills/fills due to register
allocation.

Putting an UNDEF instruction in front of the lowered instructions
allows the IR to limit the liveness of the VGRF, reducing register
pressure.

This has a pretty dramatic effect on spills/fills for RT shaders. Here
the stats on Q2RTX shaders on DG2 (wipping out any spills/fills due to
register allocation) :

Instructions in all programs: 26150 -> 24955 (-4.6%)
SENDs in all programs: 1148 -> 1148 (+0.0%)
Loops in all programs: 4 -> 4 (+0.0%)
Cycles in all programs: 392179 -> 332787 (-15.1%)
Spills in all programs: 132 -> 116 (-12.1%)
Fills in all programs: 262 -> 154 (-41.2%)

Shader-db results on TGL :

total instructions in shared programs: 21158140 -> 21158377 (<.01%)
instructions in affected programs: 76629 -> 76866 (0.31%)
helped: 18
HURT: 20
helped stats (abs) min: 1 max: 60 x̄: 18.89 x̃: 12
helped stats (rel) min: 0.21% max: 3.61% x̄: 1.02% x̃: 0.77%
HURT stats (abs)   min: 1 max: 79 x̄: 28.85 x̃: 18
HURT stats (rel)   min: 0.04% max: 2.81% x̄: 1.13% x̃: 0.79%
95% mean confidence interval for instructions value: -4.82 17.30
95% mean confidence interval for instructions %-change: -0.34% 0.57%
Inconclusive result (value mean confidence interval includes 0).

total loops in shared programs: 5753 -> 5753 (0.00%)
loops in affected programs: 0 -> 0
helped: 0
HURT: 0

total cycles in shared programs: 798856834 -> 798870688 (<.01%)
cycles in affected programs: 6208395 -> 6222249 (0.22%)
helped: 22
HURT: 17
helped stats (abs) min: 2 max: 8794 x̄: 1438.18 x̃: 782
helped stats (rel) min: 0.05% max: 2.28% x̄: 0.63% x̃: 0.44%
HURT stats (abs)   min: 2 max: 19178 x̄: 2676.12 x̃: 1358
HURT stats (rel)   min: 0.04% max: 23.49% x̄: 2.25% x̃: 0.71%
95% mean confidence interval for cycles value: -952.19 1662.65
95% mean confidence interval for cycles %-change: -0.64% 1.90%
Inconclusive result (value mean confidence interval includes 0).

total spills in shared programs: 4078 -> 4066 (-0.29%)
spills in affected programs: 40 -> 28 (-30.00%)
helped: 2
HURT: 0

total fills in shared programs: 2856 -> 2832 (-0.84%)
fills in affected programs: 127 -> 103 (-18.90%)
helped: 2
HURT: 0

total sends in shared programs: 998554 -> 998554 (0.00%)
sends in affected programs: 0 -> 0
helped: 0
HURT: 0

LOST:   0
GAINED: 0

Total CPU time (seconds): 2346.06 -> 2304.80 (-1.76%)

Signed-off-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/18657>
2022-10-27 21:05:00 +00:00
..
brw_cfg.cpp intel/fs: Add physical fall-through CFG edge for unconditional BREAK instruction. 2021-12-21 00:43:29 +00:00
brw_cfg.h
brw_clip.h
brw_clip_line.c intel/compiler: Split 3DPRIM_* defines out to a separate header. 2022-06-30 23:46:35 +00:00
brw_clip_point.c
brw_clip_tri.c intel/compiler: Split 3DPRIM_* defines out to a separate header. 2022-06-30 23:46:35 +00:00
brw_clip_unfilled.c intel/compiler: Split 3DPRIM_* defines out to a separate header. 2022-06-30 23:46:35 +00:00
brw_clip_util.c intel: move away from booleans to identify platforms 2021-11-08 16:48:06 +00:00
brw_compile_clip.c intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_compile_ff_gs.c intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_compile_sf.c intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_compiler.c intel/compiler: Vectorize gl_TessLevelInner/Outer[] writes [v2] 2022-10-13 11:38:21 -07:00
brw_compiler.h anv: support VK_PIPELINE_CREATE_RAY_TRACING_SKIP_* 2022-10-20 00:03:55 +00:00
brw_dead_control_flow.cpp
brw_dead_control_flow.h
brw_debug_recompile.c intel/compiler: Stop including src/mesa/main/config.h 2022-06-30 23:46:35 +00:00
brw_disasm.c intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_disasm_info.c intel/eu: Handle compaction when inserting validation errors 2022-07-28 21:31:45 +00:00
brw_disasm_info.h intel/eu: Handle compaction when inserting validation errors 2022-07-28 21:31:45 +00:00
brw_eu.c intel/compiler: Convert brw_eu.cpp back to brw_eu.c 2022-06-30 23:46:35 +00:00
brw_eu.h intel/fs: switch register allocation spilling to use LSC on Gfx12.5+ 2022-08-24 17:51:40 +00:00
brw_eu_compact.c intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_eu_defines.h intel/fs: remove unused opcode 2022-08-24 17:51:40 +00:00
brw_eu_emit.c intel/compiler: Broadcast lower code should check 64-bit int support 2022-10-27 09:22:09 +00:00
brw_eu_util.c
brw_eu_validate.c intel/fs: fixup SEND validation check on overlapping src0/src1 2022-08-24 17:51:40 +00:00
brw_fs.cpp intel/fs: reduce liveness of variables in lowering passes 2022-10-27 21:05:00 +00:00
brw_fs.h intel/fs: take a builder arg for resolve_source_modifiers() 2022-09-28 05:38:36 +00:00
brw_fs_bank_conflicts.cpp intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_fs_builder.h intel/fs: reduce liveness of variables in lowering passes 2022-10-27 21:05:00 +00:00
brw_fs_cmod_propagation.cpp intel: fix typos found by codespell 2022-06-27 10:20:55 +00:00
brw_fs_combine_constants.cpp
brw_fs_copy_propagation.cpp intel/compiler: Avoid copy propagating large registers into EOT messages 2022-07-07 20:20:01 +00:00
brw_fs_cse.cpp intel/compiler: Implement nir_intrinsic_last_invocation 2022-03-26 00:28:19 +00:00
brw_fs_dead_code_eliminate.cpp
brw_fs_generator.cpp intel/fs: Always use integer types for indirect MOVs 2022-09-28 05:38:36 +00:00
brw_fs_live_variables.cpp
brw_fs_live_variables.h
brw_fs_lower_pack.cpp intel/fs: reduce liveness of variables in lowering passes 2022-10-27 21:05:00 +00:00
brw_fs_lower_regioning.cpp intel/compiler/fs: Fix compilation of shaders with SHADER_OPCODE_SHUFFLE of float64 type 2022-09-14 19:32:43 +00:00
brw_fs_nir.cpp intel/compiler: Support 16 bit float ops 2022-10-17 15:56:28 +02:00
brw_fs_reg_allocate.cpp intel/fs: switch register allocation spilling to use LSC on Gfx12.5+ 2022-08-24 17:51:40 +00:00
brw_fs_register_coalesce.cpp
brw_fs_saturate_propagation.cpp
brw_fs_scoreboard.cpp intel/fs: SEL_EXEC uses the integer pipe for 64-bit stuff 2022-09-28 05:38:36 +00:00
brw_fs_sel_peephole.cpp
brw_fs_thread_payload.cpp intel/compiler: Use brw_ud* helpers in thread payload code 2022-09-13 01:44:24 +00:00
brw_fs_validate.cpp intel/compiler: Print more details when fs_visitor::validate() fails 2022-08-22 18:58:55 +00:00
brw_fs_visitor.cpp intel/compiler: Make thread_payload struct abstract 2022-09-13 01:44:24 +00:00
brw_gfx_ver_enum.h intel/compiler: Fix brw_gfx_ver_enum.h to be a proper header file 2022-06-30 23:46:35 +00:00
brw_inst.h intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_interpolation_map.c
brw_ir.h intel/fs: switch register allocation spilling to use LSC on Gfx12.5+ 2022-08-24 17:51:40 +00:00
brw_ir_allocator.h
brw_ir_analysis.h
brw_ir_fs.h intel/compiler: Make component() work for FIXED_GRF/ARF 2022-08-23 19:52:38 +00:00
brw_ir_performance.cpp intel/fs: Remove non-_LOGICAL URB messages 2022-07-08 19:45:34 +00:00
brw_ir_performance.h
brw_ir_vec4.h intel: fix typos found by codespell 2022-06-27 10:20:55 +00:00
brw_isa_info.h intel/compiler: Remove use of thread_local for opcode tables 2022-06-30 23:46:35 +00:00
brw_kernel.c intel/clc: assert when libclc shader is not found 2022-10-27 08:53:55 +00:00
brw_kernel.h intel/compiler: fix singleton pointer coverity warning 2022-04-19 12:36:10 +03:00
brw_lower_logical_sends.cpp intel/compiler: Support 16 bit float ops 2022-10-17 15:56:28 +02:00
brw_mesh.cpp intel/compiler: fix loading of draw_id from task & mesh payload 2022-09-29 17:17:25 +00:00
brw_nir.c nir/lower_idiv: Remove imprecise_32bit_lowering 2022-10-27 19:37:14 +00:00
brw_nir.h driconf: Add a limit_trig_input_range option 2022-05-13 06:47:53 +00:00
brw_nir_analyze_boolean_resolves.c
brw_nir_analyze_ubo_ranges.c
brw_nir_attribute_workarounds.c intel/compiler: Use named NIR intrinsic const index accessors 2022-08-16 05:44:30 +00:00
brw_nir_clamp_image_1d_2d_array_sizes.c intel/compiler: use nir_shader_instructions_pass in brw_nir_clamp_image_1d_2d_array_sizes 2021-10-05 10:02:54 +00:00
brw_nir_lower_alpha_to_coverage.c
brw_nir_lower_conversions.c intel/compiler: use nir_shader_instructions_pass in brw_nir_lower_conversions 2021-10-05 10:02:54 +00:00
brw_nir_lower_cs_intrinsics.c intel/compiler: Lower Task/Mesh local_invocation_{id,index} 2021-12-04 00:41:46 +00:00
brw_nir_lower_intersection_shader.c intel/rt: Handle halts in any-hit shaders properly 2022-08-05 11:51:31 +00:00
brw_nir_lower_mem_access_bit_sizes.c intel/compiler: add support for 8/16 bits task payload loads 2022-09-21 09:16:20 +00:00
brw_nir_lower_ray_queries.c intel/nir: fixup preserved metadata in rayquery lowering 2022-09-28 05:38:36 +00:00
brw_nir_lower_rt_intrinsics.c anv: support VK_PIPELINE_CREATE_RAY_TRACING_SKIP_* 2022-10-20 00:03:55 +00:00
brw_nir_lower_scoped_barriers.c intel/compiler: use nir_shader_instructions_pass in brw_nir_lower_scoped_barriers 2021-10-05 10:02:54 +00:00
brw_nir_lower_shader_calls.c anv: support VK_PIPELINE_CREATE_RAY_TRACING_SKIP_* 2022-10-20 00:03:55 +00:00
brw_nir_lower_shading_rate_output.c intel: fix typos found by codespell 2022-06-27 10:20:55 +00:00
brw_nir_lower_storage_image.c nir/builder: Add a nir_trim_vector helper 2022-05-11 14:47:33 +00:00
brw_nir_opt_peephole_ffma.c Revert "nir: Drop the unused instr arg for src/dest copy functions." 2022-08-30 18:21:44 +00:00
brw_nir_rt.c intel/nir: fix potential invalid function impl ptr usage 2022-09-28 05:38:36 +00:00
brw_nir_rt.h anv: support VK_PIPELINE_CREATE_RAY_TRACING_SKIP_* 2022-10-20 00:03:55 +00:00
brw_nir_rt_builder.h intel/nir: disable assert on async stack id 2022-09-28 05:38:36 +00:00
brw_nir_tcs_workarounds.c intel/compiler: use nir_metadata_none instead of its value 2021-10-05 10:02:54 +00:00
brw_nir_trig_workarounds.py driconf: Add a limit_trig_input_range option 2022-05-13 06:47:53 +00:00
brw_packed_float.c
brw_predicated_break.cpp intel/compiler: Don't predicate a WHILE if there is a CONT 2021-12-08 14:56:32 -08:00
brw_prim.h intel/compiler: Split 3DPRIM_* defines out to a separate header. 2022-06-30 23:46:35 +00:00
brw_private.h intel,anv,iris,crocus: Drop subgroup size from the shader key 2022-07-08 22:47:22 +00:00
brw_reg.h intel/compiler: Add a few more brw_ud* helpers 2022-09-13 01:44:24 +00:00
brw_reg_type.c
brw_reg_type.h intel/compiler: Move type_is_unsigned_int to brw_reg_type.h 2021-08-30 14:00:14 -07:00
brw_rt.h intel/devinfo: Rename & implement num_dual_subslices 2022-09-28 05:38:36 +00:00
brw_schedule_instructions.cpp intel/fs: Lower URB messages to SEND 2022-07-08 19:45:34 +00:00
brw_shader.cpp intel/compiler: Use FS thread payload only for FS 2022-09-13 01:44:24 +00:00
brw_shader.h intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_simd_selection.c intel/fs: fixup simd selection with shader calls 2022-08-05 11:51:31 +00:00
brw_vec4.cpp intel/compiler: Use FS thread payload only for FS 2022-09-13 01:44:24 +00:00
brw_vec4.h intel/compiler: remove gfx6 gather wa from backend. 2021-12-22 21:37:55 +00:00
brw_vec4_builder.h
brw_vec4_cmod_propagation.cpp
brw_vec4_copy_propagation.cpp intel/compiler: Introduce a new brw_isa_info structure 2022-06-30 23:46:35 +00:00
brw_vec4_cse.cpp intel/compiler: Rename vec4 state URB opcodes to have VEC4_ prefix 2022-07-08 19:45:34 +00:00
brw_vec4_dead_code_eliminate.cpp
brw_vec4_generator.cpp intel/compiler: Delete unused Gfx8+ code in brw_find_live_channel() 2022-08-02 08:41:43 +00:00
brw_vec4_gs_nir.cpp intel/compiler: Use named NIR intrinsic const index accessors 2022-08-16 05:44:30 +00:00
brw_vec4_gs_visitor.cpp intel/compiler: Use FS thread payload only for FS 2022-09-13 01:44:24 +00:00
brw_vec4_gs_visitor.h intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
brw_vec4_live_variables.cpp
brw_vec4_live_variables.h
brw_vec4_nir.cpp gallium,glsl: Delete PIPE_CAP_VERTEXID_NOBASE and lower_vertex_id. 2022-08-31 22:57:03 +00:00
brw_vec4_reg_allocate.cpp intel/compiler: Don't create vec4 reg-set for gen8+ 2022-07-14 17:49:01 +00:00
brw_vec4_surface_builder.cpp intel: move away from booleans to identify platforms 2021-11-08 16:48:06 +00:00
brw_vec4_surface_builder.h
brw_vec4_tcs.cpp intel/compiler: Use FS thread payload only for FS 2022-09-13 01:44:24 +00:00
brw_vec4_tcs.h intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
brw_vec4_tes.cpp intel/compiler: Rename vec4 state URB opcodes to have VEC4_ prefix 2022-07-08 19:45:34 +00:00
brw_vec4_tes.h intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
brw_vec4_visitor.cpp intel/vec4: Inline emit_texture and move helpers to brw_vec4_nir.cpp 2021-12-16 00:09:45 -08:00
brw_vec4_vs.h intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
brw_vec4_vs_visitor.cpp intel/compiler: Rename vec4 state URB opcodes to have VEC4_ prefix 2022-07-08 19:45:34 +00:00
brw_vue_map.c intel/compiler: Store the number of position slots in the VUE map 2022-08-31 02:00:18 +00:00
gfx6_gs_visitor.cpp intel/compiler: Rename vec4 state URB opcodes to have VEC4_ prefix 2022-07-08 19:45:34 +00:00
gfx6_gs_visitor.h intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
intel_clc.c anv: added proper handling for input argument in intel_clc 2022-10-20 02:24:39 +00:00
meson.build intel/compiler: Explicitly include build-id when linking intel_clc 2022-10-01 16:42:07 +00:00
test_eu_compact.cpp intel/compiler: Fixes [-Wdeprecated-declarations] in test_eu_compact.cpp 2022-08-23 15:19:16 +00:00
test_eu_validate.cpp intel/compiler: Fixes [-Wdeprecated-declarations] in test_eu_validate.cpp 2022-08-23 15:19:16 +00:00
test_fs_cmod_propagation.cpp intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
test_fs_copy_propagation.cpp intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
test_fs_saturate_propagation.cpp intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
test_fs_scoreboard.cpp intel/fs/xehp: Add unit test for handling of RaR deps across multiple pipelines. 2022-01-25 22:40:44 +00:00
test_simd_selection.cpp intel/compiler: Initialize SIMDSelectionTest member error. 2021-11-03 04:22:35 +00:00
test_vec4_cmod_propagation.cpp intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
test_vec4_copy_propagation.cpp intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
test_vec4_dead_code_eliminate.cpp intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
test_vec4_register_coalesce.cpp intel/fs,vec4: Drop support for shader time 2021-12-10 21:20:47 +00:00
test_vf_float_conversions.cpp