Georg Lehmann
edfba93f4c
aco/optimizer: apply f2f32 without label_usedef
...
This means we apply it to instructions where an omod label removed label_usedef.
Foz-DB Navi21:
Totals from 52 (0.07% of 79789) affected shaders:
Instrs: 144945 -> 144304 (-0.44%); split: -0.45%, +0.01%
CodeSize: 835956 -> 833992 (-0.23%); split: -0.30%, +0.06%
Latency: 2850900 -> 2850018 (-0.03%); split: -0.05%, +0.02%
InvThroughput: 1299486 -> 1299116 (-0.03%); split: -0.05%, +0.02%
Copies: 29638 -> 29803 (+0.56%); split: -0.16%, +0.71%
VALU: 94971 -> 94329 (-0.68%); split: -0.69%, +0.01%
SALU: 29272 -> 29274 (+0.01%)
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:25 +00:00
Georg Lehmann
c0e88c376a
aco/optimizer: validate context data
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:25 +00:00
Georg Lehmann
906b7dbcec
aco: replace novalidateir with novalidate debug option
...
The next commits will add more validation that's enabled by default.
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:25 +00:00
Georg Lehmann
1540db244b
aco/optimizer: store parent_instr for all temps
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:25 +00:00
Georg Lehmann
918359b41e
aco/optimizer: add semantic aliases for info.instr
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:25 +00:00
Georg Lehmann
c62d7e680c
aco/optimizer: remove label_mul
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:24 +00:00
Georg Lehmann
f773860a23
aco/optimizer: remove label_bitwise
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:24 +00:00
Georg Lehmann
cf3ec4a28f
aco/optimizer: remove label_split
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:24 +00:00
Georg Lehmann
907e86e8fb
aco/optimizer: remove label_vec
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:24 +00:00
Georg Lehmann
2c0a924521
aco/optimizer: remove label_minmax
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:24 +00:00
Georg Lehmann
dca8a7981d
aco/optimizer: remove label_f2f32
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:24 +00:00
Georg Lehmann
17a973c6fa
aco/optimizer: remove label_dpp8 and label_dpp16
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:24 +00:00
Georg Lehmann
dfa7e56f23
aco/optimizer: remove label_add_sub
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:23 +00:00
Georg Lehmann
345bf8a2f2
aco/optimizer: remove label_vop3p
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:23 +00:00
Georg Lehmann
6667ee66d5
aco/optimizer: remove label_vopc
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34858 >
2025-05-09 14:23:23 +00:00
Georg Lehmann
f5a5905e37
aco: support nir_op_bfdot2_bfadd
...
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34768 >
2025-05-09 11:20:26 +00:00
Georg Lehmann
5ca98bf99e
aco: support bf16 wmma
...
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34768 >
2025-05-09 11:20:25 +00:00
Georg Lehmann
e8f5c335ff
radv,aco,nir: keep the A and B base type for cmat_muladd_amd
...
With bfloat16, and the two fp8 formats in the future, using just the bit size
to identify the types is no longer possible.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34768 >
2025-05-09 11:20:25 +00:00
Samuel Pitoiset
dae45adc9d
aco: adjust an assertion in select_trap_handler_shader()
...
macOS-CI / macOS-CI (dri) (push) Waiting to run
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34840 >
2025-05-09 09:04:57 +00:00
Samuel Pitoiset
ae6d3df139
radv,aco: dump more SQ_WAVE registers from the trap handler on GFX12
...
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34840 >
2025-05-09 09:04:57 +00:00
Samuel Pitoiset
50a01a6559
radv: fix save/restore SCC in the trap handler on GFX12
...
SQ_WAVE_STATE_PRIV now contains SCC instead of SQ_WAVE_STATUS.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34840 >
2025-05-09 09:04:57 +00:00
Rhys Perry
18a53230eb
aco: don't check dst_bitsize in apply_load_extract
...
I don't think this is necessary.
fossil-db (gfx1201):
Totals from 12 (0.02% of 79377) affected shaders:
Instrs: 73041 -> 72669 (-0.51%); split: -0.51%, +0.00%
CodeSize: 417376 -> 413852 (-0.84%); split: -0.85%, +0.00%
Latency: 1301862 -> 1301533 (-0.03%); split: -0.03%, +0.00%
InvThroughput: 599874 -> 599723 (-0.03%)
VClause: 1344 -> 1346 (+0.15%)
Copies: 15855 -> 15832 (-0.15%); split: -0.37%, +0.23%
VALU: 42138 -> 41883 (-0.61%); split: -0.61%, +0.00%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:50 +00:00
Rhys Perry
eb95f7cc0e
aco: support sign extension in apply_load_extract
...
fossil-db (gfx1201):
Totals from 10 (0.01% of 79377) affected shaders:
Instrs: 28954 -> 28938 (-0.06%)
CodeSize: 164552 -> 164472 (-0.05%)
Latency: 1249341 -> 1247037 (-0.18%)
InvThroughput: 297077 -> 296618 (-0.15%)
VALU: 15951 -> 15941 (-0.06%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:50 +00:00
Rhys Perry
0de0fd38b4
aco: support more opcodes in apply_ds_extract
...
fossil-db (gfx1201):
Totals from 320 (0.40% of 79377) affected shaders:
Instrs: 3439754 -> 3432384 (-0.21%)
CodeSize: 18008696 -> 17973180 (-0.20%); split: -0.20%, +0.00%
VGPRs: 16016 -> 15404 (-3.82%)
Latency: 20246168 -> 20295740 (+0.24%); split: -0.08%, +0.33%
InvThroughput: 4462916 -> 4478546 (+0.35%); split: -0.08%, +0.43%
VClause: 87123 -> 87099 (-0.03%)
Copies: 261779 -> 261948 (+0.06%); split: -0.05%, +0.12%
Branches: 94611 -> 94601 (-0.01%); split: -0.01%, +0.00%
VALU: 1870695 -> 1865738 (-0.26%)
SALU: 488351 -> 487557 (-0.16%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:50 +00:00
Rhys Perry
4fa1c92862
aco/gfx12: allow 8/16-bit smem loads
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:50 +00:00
Rhys Perry
75efc218f5
aco: support 8/16-bit loads in smem_combine()
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:50 +00:00
Rhys Perry
13b0131edc
aco/gfx12: select dwordx3 smem loads
...
fossil-db (gfx1201):
Totals from 47814 (60.24% of 79377) affected shaders:
MaxWaves: 1436871 -> 1436855 (-0.00%); split: +0.00%, -0.00%
Instrs: 36653621 -> 36649461 (-0.01%); split: -0.07%, +0.06%
CodeSize: 194102884 -> 194076060 (-0.01%); split: -0.06%, +0.04%
VGPRs: 2267944 -> 2269648 (+0.08%); split: -0.01%, +0.08%
SpillSGPRs: 8301 -> 8295 (-0.07%); split: -0.08%, +0.01%
Latency: 249627561 -> 249631829 (+0.00%); split: -0.04%, +0.04%
InvThroughput: 40004042 -> 40003575 (-0.00%); split: -0.02%, +0.02%
VClause: 680488 -> 680429 (-0.01%); split: -0.08%, +0.07%
SClause: 1062835 -> 1066206 (+0.32%); split: -0.20%, +0.52%
Copies: 2393981 -> 2393607 (-0.02%); split: -0.23%, +0.22%
Branches: 728117 -> 728113 (-0.00%); split: -0.00%, +0.00%
VALU: 20358269 -> 20358585 (+0.00%); split: -0.01%, +0.01%
SALU: 4737317 -> 4736411 (-0.02%); split: -0.07%, +0.06%
SMEM: 1712349 -> 1710075 (-0.13%); split: -0.13%, +0.00%
VOPD: 5808 -> 5813 (+0.09%); split: +0.12%, -0.03%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:50 +00:00
Rhys Perry
90a5c93ea5
aco: prepare for dwordx3 smem loads
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:50 +00:00
Rhys Perry
208d62430f
aco/gfx12: use s_load_dwordx3 to load ray launch sizes
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:49 +00:00
Rhys Perry
cbd718506b
aco: add smem opcode helper
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34162 >
2025-05-08 13:30:49 +00:00
Marek Olšák
b960137ebf
aco: remove unused aco_shader_info::tcs_offchip_layout
...
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34863 >
2025-05-08 02:54:13 +00:00
Rhys Perry
9ca71b52aa
aco: swap the correct v_mov_b32 if there are two of them
...
Previously, this function tried to swap the instruction which is not
v_mov_b32, so that it doesn't introduce any new OPY-only instructions. If
both were v_mov_b32, it swapped Y. Since this makes Y opy-only, this can't
be done if X is also opy-only.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Fixes: 408fa33c09 ("aco/gfx12: don't use second VALU for VOPD's OPX if there is a WaR")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/13101
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34841 >
2025-05-06 21:04:28 +00:00
Rhys Perry
752f5f317e
aco: replace max_const_offset_plus_one with max_const_offset
...
macOS-CI / macOS-CI (dri) (push) Waiting to run
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34766 >
2025-05-01 09:19:02 +00:00
Rhys Perry
a85ebe16b3
aco: fix max_const_offset_plus_one overflow
...
smem_offset_max is UINT32_MAX on GFX7 and setting
max_const_offset_plus_one to 0 causes divisions by zero later.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Fixes: c26851b80b ("aco: increase max_const_offset_plus_one for SMEM load_global")
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Tested-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34766 >
2025-05-01 09:19:02 +00:00
Rhys Perry
6338ed44c5
aco/gfx12: increase maximum vbuffer offset
...
fossil-db (gfx1201):
Totals from 301 (0.38% of 79377) affected shaders:
Instrs: 2734478 -> 2728816 (-0.21%); split: -0.21%, +0.00%
CodeSize: 14347476 -> 14306568 (-0.29%)
Latency: 15508055 -> 15502202 (-0.04%); split: -0.04%, +0.00%
InvThroughput: 2846419 -> 2842387 (-0.14%); split: -0.14%, +0.00%
VClause: 68286 -> 68101 (-0.27%); split: -0.30%, +0.03%
SClause: 49487 -> 49500 (+0.03%)
Copies: 207179 -> 206093 (-0.52%); split: -0.57%, +0.04%
Branches: 72941 -> 72942 (+0.00%); split: -0.00%, +0.00%
VALU: 1549156 -> 1544727 (-0.29%); split: -0.29%, +0.00%
SALU: 339620 -> 338989 (-0.19%); split: -0.19%, +0.00%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34730 >
2025-04-29 17:44:41 +00:00
Rhys Perry
d987d5e341
aco/gfx12: increase maximum global/scratch offset
...
fossil-db (gfx1201):
Totals from 29 (0.04% of 79377) affected shaders:
Instrs: 1439082 -> 1439070 (-0.00%)
CodeSize: 7641688 -> 7641608 (-0.00%)
Latency: 9836296 -> 9836280 (-0.00%)
VALU: 799504 -> 799496 (-0.00%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34730 >
2025-04-29 17:44:41 +00:00
Rhys Perry
02d193f058
aco/gfx12: increase maximum smem offset
...
fossil-db (gfx1201):
Totals from 55 (0.07% of 79377) affected shaders:
Instrs: 3203775 -> 3200809 (-0.09%)
CodeSize: 16817140 -> 16813440 (-0.02%); split: -0.04%, +0.02%
Latency: 17838315 -> 17829658 (-0.05%)
InvThroughput: 3352905 -> 3351689 (-0.04%)
SClause: 57377 -> 57273 (-0.18%)
Copies: 231006 -> 230941 (-0.03%)
SALU: 436900 -> 435234 (-0.38%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34730 >
2025-04-29 17:44:41 +00:00
Rhys Perry
c26851b80b
aco: increase max_const_offset_plus_one for SMEM load_global
...
fossil-db (gfx1201):
Totals from 1115 (1.40% of 79377) affected shaders:
Instrs: 1473805 -> 1467571 (-0.42%); split: -0.43%, +0.01%
CodeSize: 7852972 -> 7819656 (-0.42%); split: -0.44%, +0.02%
SpillSGPRs: 1632 -> 1460 (-10.54%); split: -11.27%, +0.74%
Latency: 11975762 -> 11971915 (-0.03%); split: -0.05%, +0.02%
InvThroughput: 2496961 -> 2496448 (-0.02%); split: -0.03%, +0.01%
VClause: 25213 -> 25218 (+0.02%); split: -0.00%, +0.02%
SClause: 28822 -> 28565 (-0.89%); split: -1.41%, +0.52%
Copies: 106377 -> 105715 (-0.62%); split: -1.23%, +0.61%
Branches: 27497 -> 27473 (-0.09%)
PreSGPRs: 52071 -> 51310 (-1.46%)
VALU: 871051 -> 870694 (-0.04%); split: -0.04%, +0.00%
SALU: 186090 -> 181811 (-2.30%); split: -2.32%, +0.02%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34730 >
2025-04-29 17:44:41 +00:00
Rhys Perry
f390893a64
aco/gfx12: use s_sub_u64
...
fossil-db (gfx1201):
Totals from 2 (0.00% of 79377) affected shaders:
Instrs: 243999 -> 243993 (-0.00%)
CodeSize: 1288176 -> 1288152 (-0.00%)
Latency: 1894093 -> 1894091 (-0.00%)
InvThroughput: 378819 -> 378818 (-0.00%)
SALU: 33048 -> 33044 (-0.01%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34730 >
2025-04-29 17:44:41 +00:00
Rhys Perry
5b4813c4f0
aco/gfx12: use s_add_u64
...
fossil-db (gfx1201):
Totals from 122 (0.15% of 79377) affected shaders:
Instrs: 3640138 -> 3637577 (-0.07%); split: -0.07%, +0.00%
CodeSize: 19133796 -> 19120080 (-0.07%); split: -0.08%, +0.01%
SpillSGPRs: 666 -> 650 (-2.40%)
SpillVGPRs: 2147 -> 2159 (+0.56%)
Scratch: 254208 -> 255232 (+0.40%)
Latency: 21529337 -> 21522317 (-0.03%); split: -0.04%, +0.00%
InvThroughput: 4048519 -> 4047233 (-0.03%); split: -0.03%, +0.00%
VClause: 90453 -> 90455 (+0.00%)
SClause: 67846 -> 67674 (-0.25%); split: -0.28%, +0.03%
Copies: 287449 -> 287476 (+0.01%); split: -0.04%, +0.05%
Branches: 104526 -> 104530 (+0.00%); split: -0.00%, +0.01%
PreSGPRs: 9795 -> 9723 (-0.74%); split: -0.78%, +0.04%
VALU: 2004219 -> 2003031 (-0.06%); split: -0.06%, +0.00%
SALU: 492651 -> 491737 (-0.19%); split: -0.19%, +0.00%
VMEM: 161317 -> 161341 (+0.01%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34730 >
2025-04-29 17:44:41 +00:00
Rhys Perry
20279c28c8
aco/tests: add pseudo-scalar transcendental and fallback path RA tests
...
macOS-CI / macOS-CI (dri) (push) Waiting to run
macOS-CI / macOS-CI (xlib) (push) Waiting to run
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34343 >
2025-04-29 15:15:11 +00:00
Rhys Perry
96e49b7904
aco/ra: add ra_test_policy::use_compact_relocate
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34343 >
2025-04-29 15:15:10 +00:00
Rhys Perry
3c1dbc1d9b
aco/ra: cleanup compact_relocate_vars fallback path
...
We don't need to duplicate these loops.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34343 >
2025-04-29 15:15:10 +00:00
Rhys Perry
a780345e01
aco: fix compact_relocate_vars fallback with scc/exec/m0 precolored regs
...
This probably doesn't fix anything in practice. I don't think this path is
ever taken for SGPRs except for pseudo-scalar transcendental instructions,
and those don't have any precolored operands/definitions.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34343 >
2025-04-29 15:15:10 +00:00
Rhys Perry
f6581b41c4
aco/ra: don't require alignment for NPOT SGPR temporaries
...
Aligning these can create situations where register allocation is
impossible. Only pseudo-instructions can use these, and they don't require
any alignment.
I'm not sure if these temporaries actually happen in practice. This
probably only affects the get_reg()'s compact_relocate_vars fallback path,
which doesn't usually happen for SGPRs.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34343 >
2025-04-29 15:15:10 +00:00
Rhys Perry
623230a6ef
aco/ra: change sorting in compact_relocate_vars
...
D16 MIMG or pseudo-scalar transcendental instructions might give lower
limits to the maximum register available for their definitions, so just
try to place them earlier.
This is also part of fixing compact_relocate_vars with aligned NPOT
def/killed-op space (the second part is the later commit which changes
get_stride()).
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34343 >
2025-04-29 15:15:10 +00:00
Rhys Perry
de896234d8
aco: improve spilling of clobbered operands
...
macOS-CI / macOS-CI (dri) (push) Waiting to run
macOS-CI / macOS-CI (xlib) (push) Waiting to run
We can ignore live_changes for these.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34699 >
2025-04-28 10:04:43 +00:00
Rhys Perry
7fe84024cb
aco: fix get_temp_reg_changes with clobbered operands
...
The spiller might have tried to spill a live-through first or second
s_fmac_f32 operand, but this wouldn't have reduced the SGPRs if the third
operand wasn't killed
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/13038
Fixes: d6cb45dbb0 ("aco/spill: Allow spilling live-through operands")
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34699 >
2025-04-28 10:04:43 +00:00
Rhys Perry
3c021b79b4
aco/ra: use a correct stride for subdword get_reg_impl
...
macOS-CI / macOS-CI (dri) (push) Waiting to run
macOS-CI / macOS-CI (xlib) (push) Waiting to run
fossil-db (gfx1201):
Totals from 3 (0.00% of 79377) affected shaders:
Instrs: 1312 -> 1308 (-0.30%)
CodeSize: 7112 -> 7096 (-0.22%)
Latency: 5381 -> 5382 (+0.02%)
InvThroughput: 753 -> 752 (-0.13%)
Copies: 69 -> 68 (-1.45%)
VALU: 836 -> 835 (-0.12%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34679 >
2025-04-25 14:43:41 +00:00
Rhys Perry
ae6d4f1195
aco/ra: update_renames() before add_subdword_definition()
...
The register file tests here should be done after update_renames().
Normally, get_reg() wouldn't have to move anything to make space for a 1-3
byte definition. This was encountered with skip_optimistic_path=true and a
get_reg_impl() bug (fixed in a later commit) which caused suboptimal
register assignment.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34679 >
2025-04-25 14:43:41 +00:00