Skip to content

Draft: nir/opt_load_store_vectorize: add ir3 intrinsics

Mike Blumenkrantz requested to merge zmike/mesa:ir3-vectorize into main 
MaxWaves: 1404678 -> 1481506 (+5.47%); split: +5.47%, -0.00%
Instrs: 40290756 -> 38081820 (-5.48%); split: -5.51%, +0.03%
CodeSize: 81578630 -> 75783980 (-7.10%); split: -7.10%, +0.00%
NOPs: 8259900 -> 8309987 (+0.61%); split: -0.79%, +1.40%
MOVs: 2211475 -> 1502588 (-32.05%); split: -32.58%, +0.53%
Full: 1519053 -> 1386889 (-8.70%); split: -8.71%, +0.01%
(ss): 1105397 -> 1067728 (-3.41%); split: -4.47%, +1.06%
(sy): 465086 -> 490393 (+5.44%); split: -1.10%, +6.54%
(ss)-stall: 3414712 -> 3469508 (+1.60%); split: -0.56%, +2.17%
(sy)-stall: 12832261 -> 18202020 (+41.85%); split: -0.22%, +42.07%
STPs: 21854 -> 6240 (-71.45%)
LDPs: 23966 -> 6809 (-71.59%)
Subgroup size: 10745792 -> 10765568 (+0.18%)
Cat0: 8843708 -> 8894713 (+0.58%); split: -0.75%, +1.32%
Cat1: 2830843 -> 2030572 (-28.27%); split: -28.71%, +0.44%
Cat2: 14839178 -> 13886163 (-6.42%)
Cat3: 10725232 -> 10724858 (-0.00%)
Cat5: 1418824 -> 769671 (-45.75%)
Cat6: 609135 -> 752007 (+23.45%); split: -2.13%, +25.59%

Totals from 24535 (19.32% of 126999) affected shaders:
MaxWaves: 195394 -> 272222 (+39.32%); split: +39.35%, -0.03%
Instrs: 16691578 -> 14482642 (-13.23%); split: -13.31%, +0.07%
CodeSize: 33544774 -> 27750124 (-17.27%); split: -17.28%, +0.00%
NOPs: 3673019 -> 3723106 (+1.36%); split: -1.79%, +3.15%
MOVs: 1252091 -> 543204 (-56.62%); split: -57.55%, +0.93%
Full: 476719 -> 344555 (-27.72%); split: -27.74%, +0.02%
(ss): 417321 -> 379652 (-9.03%); split: -11.83%, +2.80%
(sy): 165170 -> 190477 (+15.32%); split: -3.09%, +18.41%
(ss)-stall: 1305079 -> 1359875 (+4.20%); split: -1.47%, +5.67%
(sy)-stall: 4797685 -> 10167444 (+111.92%); split: -0.60%, +112.52%
STPs: 17260 -> 1646 (-90.46%)
LDPs: 19906 -> 2749 (-86.19%)
Subgroup size: 1931136 -> 1950912 (+1.02%)
Cat0: 3923463 -> 3974468 (+1.30%); split: -1.68%, +2.98%
Cat1: 1544790 -> 744519 (-51.80%); split: -52.60%, +0.80%
Cat2: 6112542 -> 5159527 (-15.59%)
Cat3: 3722567 -> 3722193 (-0.01%)
Cat5: 851500 -> 202347 (-76.24%)
Cat6: 152361 -> 295233 (+93.77%); split: -8.52%, +102.29%

Merge request reports