Commit graph

15393 commits

Author SHA1 Message Date
Daniel Schürmann
043ec096c1 aco/validate: use live-in variables for RA validation
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30120>
2024-07-16 14:00:49 +00:00
Daniel Schürmann
976dd71942 aco/cssa: use live-in variables instead of live-out variables
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30120>
2024-07-16 14:00:49 +00:00
Daniel Schürmann
c146d4b6b6 aco/spill: use live-in variables directly rather than computing them
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30120>
2024-07-16 14:00:49 +00:00
Daniel Schürmann
162876c875 aco/ra: use live-in variables directly rather than computing them
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30120>
2024-07-16 14:00:49 +00:00
Daniel Schürmann
29262f8cf3 aco: compute live-in variables in addition to live-out variables
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30120>
2024-07-16 14:00:49 +00:00
Daniel Schürmann
79875737cc radv: use NIR loop invariant code motion pass
Totals from 3469 (4.37% of 79395) affected shaders: (GFX11)
MaxWaves: 78690 -> 78622 (-0.09%); split: +0.03%, -0.11%
Instrs: 11093592 -> 11092346 (-0.01%); split: -0.09%, +0.07%
CodeSize: 57979444 -> 58077232 (+0.17%); split: -0.12%, +0.29%
VGPRs: 257892 -> 258336 (+0.17%); split: -0.08%, +0.25%
SpillSGPRs: 2958 -> 2521 (-14.77%); split: -32.83%, +18.05%
Latency: 135247583 -> 134446992 (-0.59%); split: -0.61%, +0.02%
InvThroughput: 25654328 -> 25478620 (-0.68%); split: -0.73%, +0.05%
VClause: 244799 -> 244499 (-0.12%); split: -0.17%, +0.05%
SClause: 313323 -> 315081 (+0.56%); split: -0.40%, +0.96%
Copies: 835953 -> 842457 (+0.78%); split: -0.38%, +1.15%
Branches: 330136 -> 330210 (+0.02%); split: -0.03%, +0.05%
PreSGPRs: 193374 -> 200277 (+3.57%); split: -0.38%, +3.95%
PreVGPRs: 223947 -> 224227 (+0.13%); split: -0.02%, +0.15%
VALU: 6312413 -> 6314841 (+0.04%); split: -0.02%, +0.06%
SALU: 1222275 -> 1227329 (+0.41%); split: -0.26%, +0.67%
VMEM: 408421 -> 408412 (-0.00%)
SMEM: 430966 -> 430399 (-0.13%)
VOPD: 2482 -> 2440 (-1.69%); split: +0.44%, -2.14%

Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28783>
2024-07-16 12:29:08 +00:00
Dave Airlie
814a2da2f4 radv/video: advertise mutable/extended for dst video images.
This allows zink video to create planar image views if needed.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30203>
2024-07-16 07:04:15 +00:00
Samuel Pitoiset
8863704c6b radv/meta: add a helper to create descriptor set layout
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30187>
2024-07-16 06:17:07 +00:00
Samuel Pitoiset
3d322b787e radv/meta: add a helper to create pipeline layout
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30187>
2024-07-16 06:17:07 +00:00
Samuel Pitoiset
c6a626e000 radv/meta: add a helper to create compute pipeline
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30187>
2024-07-16 06:17:07 +00:00
Samuel Pitoiset
bf3b2d2912 radv/meta: remove useless checks for NULL handles before destroying
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30187>
2024-07-16 06:17:07 +00:00
Samuel Pitoiset
4deb138e7d radv/meta: remove unused number of rectangles for internal operations
It was always 1.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30187>
2024-07-16 06:17:07 +00:00
Samuel Pitoiset
ecd3bbf826 radv/meta: remove redundant check for hw resolve pipelines
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30187>
2024-07-16 06:17:07 +00:00
Samuel Pitoiset
76e4edefbf radv/meta: remove unnecessary blit2d_dst_temps struct
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30187>
2024-07-16 06:17:07 +00:00
Samuel Pitoiset
e739d0e5bb radv/meta: remove non-valuable comments
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30187>
2024-07-16 06:17:07 +00:00
Josh Simmons
1ced840632 radv: Add RADV_PROFILE_PSTATE envvar
Enable selecting the specific pstate to enter when using thread tracing
and when acquiring the profiling lock for performance queries.

Signed-off-by: Josh Simmons <josh@nega.tv>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30139>
2024-07-15 20:32:01 +00:00
Marek Olšák
a5b4ae67ae ac: add radeon_info::has_scratch_base_registers
Fixes: 3b0bfd254f - radeonsi/gfx11: make flat_scratch changes for compute

Reviewed-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30071>
2024-07-15 13:52:25 -04:00
Georg Lehmann
b0ad3c2160 aco: fix s_delay_alu with salu and trans dependency
These events were silently truncated in get_counters_for_event.
The integer types in this pass are a bit all over the place, maybe we should
consider using typedefs for clarity or a different solution with type safety.

Fixes: 9e9cabd2fa ("aco/waitcnt: support GFX12 in waitcnt pass")
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30163>
2024-07-15 12:02:35 +00:00
Samuel Pitoiset
15a3aff0f1 radv: use zero allocation for the device queues
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
5fa22f9fec radv: regroup all tools initialization in one helper
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
9d2751bbde radv: add radv_device_init_rmv()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
27a90f1f29 radv: add helpers for init/deinit device fault detection
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
351fba7ee3 radv: add radv_device_init_trap_handler()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
4ab6357c42 radv: simplify keeping shader info for GPU hangs debugging
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
fe09a6d72b radv: add helpers for init/deinit RGP
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
033084c912 radv: add helpers for init/deinit device memory cache
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
01339c6c93 radv: add radv_device_init_perf_counter()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Samuel Pitoiset
5657f21fcf radv: destroy the perf counter BO in radv_device_finish_perf_counter()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30153>
2024-07-15 09:34:42 +02:00
Josh Simmons
c68408d195 radv: Fix crash when using SQTT and NO_COMPUTE
Signed-off-by: Josh Simmons <josh@nega.tv>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30177>
2024-07-15 07:05:40 +00:00
Marek Olšák
0bb83a4060 ac/surface: finish display DCC for gfx12
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30114>
2024-07-13 02:17:37 +00:00
Marek Olšák
46071c90c7 ac/surface: finish display DCC for gfx11.5
Fixes: 6835257246 - amd/common: update DCC for gfx11.5

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30114>
2024-07-13 02:17:37 +00:00
Marek Olšák
678d520162 as/llvm: add s_nops before the ordered add loop and s_wait_alu workaround
The s_nops improve performance.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30063>
2024-07-13 01:32:48 +00:00
Marek Olšák
bd8d20543d ac/llvm: fix inline assembly register constraints for ordered_add_loop_gfx12_amd
This is only known to fix the assembly code when num_atomics > 6, which is
not currently used.

The VGPRs are reordered to simplify the clobber constraint.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30063>
2024-07-13 01:32:48 +00:00
Marek Olšák
b617c3b06e ac/llvm: remove s_nop from ordered_add_loop_gfx12_amd
This is faster.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30063>
2024-07-13 01:32:48 +00:00
Marek Olšák
11272a8d82 ac/nir: remove sleeps from gfx12 streamout code
This is faster.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30063>
2024-07-13 01:32:48 +00:00
Marek Olšák
1b2cd628b8 nir: rename ordered_xfb_counter_add_gfx12_amd -> ordered_add_loop_gfx12_amd
because it can also be used by compute.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30063>
2024-07-13 01:32:48 +00:00
Marek Olšák
1fd43bca2c radeonsi: don't use CP DMA on GFX940
It's been defeatured.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30115>
2024-07-13 00:58:30 +00:00
Samuel Pitoiset
0a6852907d radv: fix marking RADV_DYNAMIC_COLOR_ATTACHMENT_MAP as dirty
Due to the cmdbuf dirty split.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30119>
2024-07-12 06:37:52 +00:00
David Rosca
6cc32c609b radeonsi/vcn: Add low latency encode support
This feature should be enabled for use cases when the lowest encoding
latency is desired, such as real-time game streaming.
Disabled by default due to increased power usage.

There is no libva interface currently that could be used for this, so
for now it can only be enabled by setting AMD_DEBUG=lowlatencyenc
environment variable.

See: https://gitlab.freedesktop.org/drm/amd/-/issues/3336

Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30039>
2024-07-11 18:33:41 +00:00
Rhys Perry
4b36668575 radv: remove unecessary nir_remove_unused_varyings cleanup passes
I think the comment meant to refer to nir_remove_unused_varyings.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25590>
2024-07-10 19:11:38 +00:00
Georg Lehmann
cd9187a1e1 aco/ra: fix affinity for s_addk
The first, non SCC def matters.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Fixes: 39380d475a ("aco: add affinities for possible sopk optimizations")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29943>
2024-07-10 13:36:00 +00:00
Daniel Schürmann
fab95c78f9 aco/ra: remove special-casing of p_logical_end
There is always enough registers available and this code
was broken anyway and had no effect.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00
Daniel Schürmann
ad01e473f5 aco/live_var_analysis: use separate allocator for temporary live sets
This drastically reduces the memory footprint of the live sets.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00
Daniel Schürmann
7c466157d0 aco/live_var_analysis: remove unused includes
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00
Daniel Schürmann
6729e81d15 aco/live_var_analysis: inline block->register_demand updates
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00
Daniel Schürmann
2f4fb9eecf aco/live_var_analysis: ignore phi definition and operand demand at predecessors
The linear_phi changes are already reflected in the live-in demand
of the successor. The logical_phi_sgpr_ops can only reduce the
register demand at the predecessor. Although this might slightly
overestimate the register-demand, no differences in code quality
were found.

Totals from 1610 (2.03% of 79395) affected shaders: (GFX11)

PreSGPRs: 54002 -> 56954 (+5.47%)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00
Daniel Schürmann
68c1e7237c aco/live_var_analysis: refactor using ctx struct
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00
Daniel Schürmann
daac18f2ce aco/util: skip empty blocks in IDSet::insert(IDSet)
Since we don't remove empty blocks on erase(), this avoids
duplicating them unnecessarily.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00
Daniel Schürmann
6c6f382d68 aco: add RegisterDemand member to Instruction
Since we never need both at the same time, we can use
a union with pass_flags.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00
Daniel Schürmann
dc851c0aa6 aco/ra: use live_in_demand in should_compact_linear_vgprs()
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29962>
2024-07-10 12:31:02 +00:00