Draft: nir: implement loop invariant code motion (LICM) pass
Draft for now since checking for nir_intrinsic_can_reorder is not sufficient and we first need to make use of ACCESS_CAN_SPECULATE.
nir: implement loop invariant code motion (LICM) pass
This simple LICM pass hoists all loop-invariant instructions
from the loops' top-level control flow, skipping any nested CF.
The hoisted instructions are placed right before the loop.Some GFX11 stats for the LICM pass. To me it looks quite beneficial.
Totals:
MaxWaves: 2415652 -> 2415504 (-0.01%); split: +0.00%, -0.01%
Instrs: 45309794 -> 45296989 (-0.03%); split: -0.06%, +0.04%
CodeSize: 235735256 -> 235634772 (-0.04%); split: -0.08%, +0.04%
VGPRs: 3082480 -> 3083368 (+0.03%); split: -0.01%, +0.04%
SpillSGPRs: 6935 -> 6982 (+0.68%); split: -13.68%, +14.36%
SpillVGPRs: 653 -> 745 (+14.09%)
Scratch: 6959872 -> 6963968 (+0.06%)
Latency: 300361437 -> 298365091 (-0.66%); split: -0.69%, +0.03%
InvThroughput: 48908954 -> 48591557 (-0.65%); split: -0.72%, +0.07%
VClause: 884873 -> 884762 (-0.01%); split: -0.07%, +0.05%
SClause: 1388421 -> 1377818 (-0.76%); split: -0.82%, +0.06%
Copies: 2937998 -> 2944504 (+0.22%); split: -0.25%, +0.48%
Branches: 899234 -> 899368 (+0.01%); split: -0.01%, +0.02%
PreSGPRs: 2445827 -> 2462883 (+0.70%); split: -0.02%, +0.71%
PreVGPRs: 2249572 -> 2250266 (+0.03%); split: -0.00%, +0.03%
VALU: 25207745 -> 25211343 (+0.01%); split: -0.01%, +0.02%
SALU: 4398056 -> 4399755 (+0.04%); split: -0.20%, +0.24%
VMEM: 1598134 -> 1598254 (+0.01%); split: -0.00%, +0.01%
SMEM: 2469245 -> 2460311 (-0.36%)
VOPD: 8801 -> 8791 (-0.11%); split: +0.31%, -0.42%
Totals from 5575 (7.02% of 79395) affected shaders:
MaxWaves: 128709 -> 128561 (-0.11%); split: +0.02%, -0.13%
Instrs: 15018601 -> 15005796 (-0.09%); split: -0.19%, +0.11%
CodeSize: 78159916 -> 78059432 (-0.13%); split: -0.25%, +0.12%
VGPRs: 393568 -> 394456 (+0.23%); split: -0.06%, +0.29%
SpillSGPRs: 4216 -> 4263 (+1.11%); split: -22.51%, +23.62%
SpillVGPRs: 518 -> 610 (+17.76%)
Scratch: 6108160 -> 6112256 (+0.07%)
Latency: 177467906 -> 175471560 (-1.12%); split: -1.17%, +0.05%
InvThroughput: 31979180 -> 31661783 (-0.99%); split: -1.10%, +0.11%
VClause: 306050 -> 305939 (-0.04%); split: -0.19%, +0.15%
SClause: 425042 -> 414439 (-2.49%); split: -2.69%, +0.20%
Copies: 1059298 -> 1065804 (+0.61%); split: -0.71%, +1.32%
Branches: 407064 -> 407198 (+0.03%); split: -0.02%, +0.05%
PreSGPRs: 302651 -> 319707 (+5.64%); split: -0.14%, +5.78%
PreVGPRs: 346505 -> 347199 (+0.20%); split: -0.02%, +0.22%
VALU: 8439279 -> 8442877 (+0.04%); split: -0.03%, +0.07%
SALU: 1523515 -> 1525214 (+0.11%); split: -0.58%, +0.69%
VMEM: 505281 -> 505401 (+0.02%); split: -0.00%, +0.03%
SMEM: 636590 -> 627656 (-1.40%)
VOPD: 3201 -> 3191 (-0.31%); split: +0.84%, -1.16%
PERCENTAGE DELTAS Shaders MaxWaves Instrs CodeSize VGPRs SpillSGPRs SpillVGPRs Scratch Latency InvThroughput VClause SClause Copies Branches PreSGPRs PreVGPRs VALU SALU VMEM SMEM VOPD
assassins_creed_valhalla 830 . -0.15% -0.20% -0.07% -53.73% . . -2.78% -3.36% +0.11% -0.75% +0.15% -0.04% +1.21% +0.12% -0.20% +0.27% . -0.07% -0.89%
control 4242 -0.01% -0.02% -0.02% +0.05% +77.27% . . -0.77% -0.74% +0.01% +0.04% -0.23% . +0.86% +0.03% +0.02% -0.15% . -0.04% .
cyberpunk_2077 2221 -0.00% -0.20% -0.27% +0.01% +74.09% . . -0.45% -0.59% +0.04% -1.13% -0.64% +0.01% +1.15% +0.05% -0.06% -0.57% . -0.88% .
dead_space 400 . -0.95% -0.91% . +4.05% . . -1.67% -1.72% -0.07% -6.01% -5.84% -0.03% +2.64% +0.13% +0.01% -6.24% . -6.16% .
deathloop 1130 +0.03% -0.00% -0.02% -0.05% +3.17% . . -0.30% -0.39% +0.12% +0.22% +0.21% +0.00% +3.82% +0.00% +0.02% +0.13% . . .
detroit_become_human 995 -0.01% -0.01% -0.01% +0.03% +86.05% . . -1.10% -0.65% +0.01% -1.48% +0.08% . +2.65% . +0.02% +0.05% . -0.23% .
deus_ex_md 8063 . -0.03% -0.04% . . . . -0.42% -0.10% . -0.28% -0.02% . +0.10% +0.00% -0.00% -0.02% . -0.17% .
doom_eternal 461 . -0.17% -0.27% +0.06% . . . -0.67% -1.17% -0.44% -7.60% +0.97% -0.00% +2.76% +0.17% +0.08% +0.29% . -4.51% -40.62%
dota2 3840 . . . . . . . . . . . . . . . . . . . .
dow3 280 . +0.00% +0.01% . . . . -0.77% -0.44% . -0.29% +0.06% . +0.35% +0.04% +0.01% +0.01% . . .
dredge 2517 . +0.06% +0.07% . . . . +0.04% +0.03% . +0.20% +0.01% . +0.82% . +0.00% . . -0.01% .
elden_ring 3733 -0.02% +0.02% +0.03% +0.20% +94.12% . . -0.21% -0.15% . +0.01% +0.34% +0.02% +0.24% +0.02% +0.01% +0.22% . . -0.14%
f1_23 4006 . -0.04% -0.03% . -11.11% . . -0.42% -0.27% +0.01% +0.21% -0.07% +0.02% +1.43% +0.01% +0.00% -0.16% . . +0.56%
farcry5 805 -0.04% -0.05% -0.07% +0.11% . . . -0.88% -2.32% . -1.36% -0.00% . +0.89% +0.14% +0.00% -0.03% . -0.26% .
god_of_war 1029 -0.01% -0.02% -0.03% +0.09% +3.29% . . -1.23% -0.44% . -0.53% +0.11% . +0.89% +0.01% +0.01% +0.06% . -0.34% .
gtav 2955 . +0.00% +0.00% . . . . -0.02% -0.01% . -0.01% . . +0.00% . . . . . .
heaven_d3d9_strict_float_ultra_4xaa 286 . +0.00% -0.00% . . . . -0.18% -0.11% . -0.19% -0.09% . +0.05% . . -0.07% . -0.02% .
hogwarts_legacy 887 -0.04% +0.01% +0.00% +0.09% +50.00% . . -0.54% -0.60% +0.02% -0.13% +0.21% . +0.54% +0.13% +0.01% +0.10% . -0.07% .
horizon_zero_dawn 1800 . +0.00% +0.00% . . . . -0.00% -0.00% . . +0.00% . . . +0.00% . . . .
madmax 922 . +0.00% -0.03% . . . . -0.41% -0.08% . -0.03% . . +0.10% . . . . . .
metro_exodus 2627 . -0.01% +0.00% . . . . -0.23% -0.22% -0.01% -0.05% -0.07% . +0.37% . -0.00% -0.06% . . +1.29%
PERCENTAGE DELTAS Shaders MaxWaves Instrs CodeSize VGPRs SpillSGPRs SpillVGPRs Scratch Latency InvThroughput VClause SClause Copies Branches PreSGPRs PreVGPRs VALU SALU VMEM SMEM VOPD
monster_hunter_world 1103 . -0.01% -0.02% . . . . -0.18% -0.19% +0.04% -0.49% +0.15% +0.02% +0.55% +0.01% +0.00% +0.05% . -0.08% .
nier 2422 -0.02% -0.04% -0.04% +0.03% . . . -0.33% -0.06% -0.02% -0.47% -0.01% . +0.04% +0.05% -0.00% -0.01% . -0.24% .
no_mans_sky 1978 . -0.00% -0.00% . . . . -0.01% -0.04% +0.02% . +0.00% . +0.00% . . +0.00% . . .
parallel_rdp 539 . +0.64% +0.72% +1.16% +inf% +20.63% +13.11% -0.56% +0.12% +1.14% -0.08% +3.73% +0.21% +1.91% +0.11% +0.92% +0.09% +1.34% . .
persona_5 1566 . . . . . . . . . . . . . . . . . . . .
q2rtx 48 . -0.07% -0.01% . . . . -0.19% -0.64% . . -0.71% . +1.94% +0.97% -0.00% -0.77% . -0.04% .
rdr2 3649 -0.04% -0.05% -0.09% +0.09% +4.69% . . -1.79% -0.99% +0.46% -1.00% +0.27% +0.00% +1.76% +0.15% +0.00% +0.13% . -0.33% +0.92%
redout 2485 . -0.01% -0.00% . +inf% . . +0.11% -0.00% +0.01% -0.11% -0.00% . +0.17% +0.00% +0.00% -0.02% . -0.05% .
resident_evil_village 1157 . -0.27% -0.25% -0.05% +2.90% . . -1.60% -0.44% -0.09% -2.40% -1.98% . +0.82% +0.02% +0.01% -1.39% . -1.16% .
second_extinction 437 . +0.10% +0.07% . . . . -0.20% -0.25% . -0.40% +1.76% . +0.48% +0.01% . +1.16% . . .
skyrim_dx9 3305 . -0.03% -0.02% +0.03% . . . -0.27% -0.38% . -0.01% -0.41% . +0.99% +0.03% . -0.20% . -0.00% .
sniper_elite_5 4278 . +0.13% +0.09% . . . . +0.01% +0.01% -0.00% . +1.29% +0.01% +0.03% +0.02% +0.00% +0.90% . . .
sottr 542 -0.01% +0.00% -0.01% +0.04% . . . +0.10% +0.25% . -0.45% -0.17% . +0.07% +0.02% -0.00% -0.08% . -0.00% .
spiderman_remastered 648 . -0.24% -0.06% . . . . -2.13% -2.19% . -1.90% -0.54% +0.01% +5.47% +0.01% -0.03% -0.82% . -1.06% .
talos_principle 1999 . +0.01% -0.03% . . . . -0.55% -0.03% . +0.05% +0.06% . +0.08% +0.00% +0.00% +0.02% . . .
the_last_of_us_part1 1198 -0.09% +0.22% +0.27% +0.24% +62.16% . . -0.68% +0.12% -0.15% -0.04% +1.12% +0.41% +1.67% +0.30% +0.19% +0.47% -0.05% -0.06% -1.53%
thewitness 3150 . +0.00% -0.00% . . . . -0.00% -0.01% . . . . . . . . . . .
total_warhammer_3 632 . -0.00% -0.01% . . . . -0.08% -0.01% -0.01% -0.33% +0.01% -0.04% +0.20% -0.00% -0.00% . . -0.01% +9.09%
war_thunder 2100 . -0.02% +0.03% . . . . -1.69% -0.82% +0.02% -0.30% -0.05% . +0.22% +0.03% -0.00% -0.02% . -0.02% -0.32%
worldofwarships 2130 . -0.04% -0.04% . . . . -1.04% -0.14% . -0.46% -0.04% . +0.90% +0.02% +0.00% -0.05% . -0.01% .
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
All affected 5575 -0.11% -0.09% -0.13% +0.23% +1.11% +17.76% +0.07% -1.12% -0.99% -0.04% -2.49% +0.61% +0.03% +5.64% +0.20% +0.04% +0.11% +0.02% -1.40% -0.31%
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Total 79395 -0.01% -0.03% -0.04% +0.03% +0.68% +14.09% +0.06% -0.66% -0.65% -0.01% -0.76% +0.22% +0.01% +0.70% +0.03% +0.01% +0.04% +0.01% -0.36% -0.11%