Skip to content

Conversation

@avik-pal
Copy link
Collaborator

@avik-pal avik-pal commented Dec 29, 2025

Mostly experimenting.

  • Add reduce to fusible with reshape (delete dims case)
  • create helper functions for the fusions so that we can process those in one go, else phase ordering creates issues
  • instead of using the worklist to process in isfusible, we keep that only for immediate cases
  • for reshape_elementwise, we process a full chain of elementwise and immediately eliminate all of the reshapes

fixes #1862 #1865

@avik-pal avik-pal force-pushed the ap/more_reshape_simplify branch 2 times, most recently from 680e48d to b2739f2 Compare December 29, 2025 05:43
Copy link
Contributor

@github-actions github-actions bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

EnzymeJAX Benchmarks

Details
Benchmark suite Current: 19db57f Previous: 426a717 Ratio
actmtch / JaXPipe / cpu / Primal 0.000006742300001860713 s 0.0000070261399650917154 s 0.96
actmtch / Jax / cpu / Primal 0.000006291740005508473 s 0.000007681720026084804 s 0.82
actmtch / HLOOpt / cpu / Primal 0.000007618559995989927 s 0.000008456099985778564 s 0.90
actmtch / PartOpt / cpu / Primal 0.00000646157999881325 s 0.000008354660039913142 s 0.77
actmtch / IPartOpt / cpu / Primal 0.000006491059998552373 s 0.000008498499983033981 s 0.76
actmtch / DefOpt / cpu / Primal 0.000007149299999582582 s 0.000008734280045246124 s 0.82
actmtch / IDefOpt / cpu / Primal 0.000006879740008116641 s 0.000008587439951952547 s 0.80
actmtch / JaXPipe / cpu / Forward 0.000010864259993468294 s 0.000012429179996615858 s 0.87
actmtch / Jax / cpu / Forward 0.000009452840008634667 s 0.000010656700023901068 s 0.89
actmtch / HLOOpt / cpu / Forward 0.000010726260004503274 s 0.000012589040015882348 s 0.85
actmtch / PartOpt / cpu / Forward 0.000010353419997954916 s 0.000011920439947061825 s 0.87
actmtch / IPartOpt / cpu / Forward 0.000010768940007892523 s 0.000012381419974190069 s 0.87
actmtch / DefOpt / cpu / Forward 0.00001013312000395672 s 0.000012271340056031477 s 0.83
actmtch / IDefOpt / cpu / Forward 0.00001064875999418291 s 0.000012316779984757888 s 0.86
actmtch / JaXPipe / cpu / PreRev 0.000010897239983478355 s 0.00001274931999432738 s 0.85
actmtch / JaXPipe / cpu / PostRev 0.000009988120011712451 s 0.000011261919989919989 s 0.89
actmtch / JaXPipe / cpu / BothRev 0.000010810720007157214 s 0.000012745980011459324 s 0.85
actmtch / Jax / cpu / BothRev 0.00000941728000270814 s 0.00001086691999262257 s 0.87
actmtch / HLOOpt / cpu / PreRev 0.000010659940001005452 s 0.000012499179983933571 s 0.85
actmtch / HLOOpt / cpu / PostRev 0.000013300379998781864 s 0.000014753979976376283 s 0.90
actmtch / HLOOpt / cpu / BothRev 0.000010641840003700051 s 0.00001177669998469355 s 0.90
actmtch / PartOpt / cpu / PreRev 0.000010860399997909552 s 0.000012723199997708434 s 0.85
actmtch / PartOpt / cpu / PostRev 0.000009844079995673384 s 0.000011315820029267342 s 0.87
actmtch / PartOpt / cpu / BothRev 0.0000115475599977799 s 0.000012766059990099166 s 0.90
actmtch / IPartOpt / cpu / PreRev 0.000010369440003614729 s 0.000012880159993073904 s 0.81
actmtch / IPartOpt / cpu / PostRev 0.000009463919993777382 s 0.000010472339963598645 s 0.90
actmtch / IPartOpt / cpu / BothRev 0.000010794539998641994 s 0.000012367339950287716 s 0.87
actmtch / DefOpt / cpu / PreRev 0.000010389639987806733 s 0.000012020940012007486 s 0.86
actmtch / DefOpt / cpu / PostRev 0.000011129979998258933 s 0.000012939700009155783 s 0.86
actmtch / DefOpt / cpu / BothRev 0.000010952219993214385 s 0.000012443320019883686 s 0.88
actmtch / IDefOpt / cpu / PreRev 0.000010425560001294798 s 0.000012142540026616188 s 0.86
actmtch / IDefOpt / cpu / PostRev 0.000011277080000127171 s 0.00001286107997657382 s 0.88
actmtch / IDefOpt / cpu / BothRev 0.00001113764000365336 s 0.000012399619981806609 s 0.90
actmtch / JaXPipe / cuda / Primal 0.0000024 s 0.000002016 s 1.19
actmtch / Jax / cuda / Primal 0.0000024 s 0.000002015 s 1.19
actmtch / HLOOpt / cuda / Primal 0.0000024 s 0.000002016 s 1.19
actmtch / PartOpt / cuda / Primal 0.0000024 s 0.000002015 s 1.19
actmtch / IPartOpt / cuda / Primal 0.0000024 s 0.000002015 s 1.19
actmtch / DefOpt / cuda / Primal 0.000002399 s 0.000002015 s 1.19
actmtch / IDefOpt / cuda / Primal 0.0000024 s 0.000002016 s 1.19
actmtch / JaXPipe / cuda / Forward 0.000010336 s 0.000009887 s 1.05
actmtch / Jax / cuda / Forward 0.000010336 s 0.000009888 s 1.05
actmtch / HLOOpt / cuda / Forward 0.000010336 s 0.000010048 s 1.03
actmtch / PartOpt / cuda / Forward 0.000010367 s 0.000010112 s 1.03
actmtch / IPartOpt / cuda / Forward 0.000010464 s 0.00000992 s 1.05
actmtch / DefOpt / cuda / Forward 0.000010272 s 0.000009696 s 1.06
actmtch / IDefOpt / cuda / Forward 0.000010432 s 0.000010272 s 1.02
actmtch / JaXPipe / cuda / PreRev 0.00001088 s 0.000009215 s 1.18
actmtch / JaXPipe / cuda / PostRev 0.000010464 s 0.000010624 s 0.98
actmtch / JaXPipe / cuda / BothRev 0.00001056 s 0.00000992 s 1.06
actmtch / Jax / cuda / BothRev 0.000010879 s 0.000010271 s 1.06
actmtch / HLOOpt / cuda / PreRev 0.000010272 s 0.000010144 s 1.01
actmtch / HLOOpt / cuda / PostRev 0.000010848 s 0.00001024 s 1.06
actmtch / HLOOpt / cuda / BothRev 0.000010593 s 0.000010304 s 1.03
actmtch / PartOpt / cuda / PreRev 0.000010624 s 0.00000992 s 1.07
actmtch / PartOpt / cuda / PostRev 0.0000104 s 0.00001008 s 1.03
actmtch / PartOpt / cuda / BothRev 0.00001024 s 0.000010015 s 1.02
actmtch / IPartOpt / cuda / PreRev 0.000013056 s 0.000010336 s 1.26
actmtch / IPartOpt / cuda / PostRev 0.000010592 s 0.000010272 s 1.03
actmtch / IPartOpt / cuda / BothRev 0.000012 s 0.0000104 s 1.15
actmtch / DefOpt / cuda / PreRev 0.000013088 s 0.000010112 s 1.29
actmtch / DefOpt / cuda / PostRev 0.000010944 s 0.000011712 s 0.93
actmtch / DefOpt / cuda / BothRev 0.000010624 s 0.000010081 s 1.05
actmtch / IDefOpt / cuda / PreRev 0.000010816 s 0.000010177 s 1.06
actmtch / IDefOpt / cuda / PostRev 0.000010304 s 0.000009696 s 1.06
actmtch / IDefOpt / cuda / BothRev 0.000010976 s 0.000010208 s 1.08
actmtch / JaXPipe / tpu / Primal 5.6325e-7 s 5.63175e-7 s 1.00
actmtch / Jax / tpu / Primal 5.969000000000001e-7 s 5.96825e-7 s 1.00
actmtch / HLOOpt / tpu / Primal 0.0000021011 s 0.000002097775 s 1.00
actmtch / PartOpt / tpu / Primal 5.96725e-7 s 5.9665e-7 s 1.00
actmtch / IPartOpt / tpu / Primal 5.521000000000001e-7 s 5.52675e-7 s 1.00
actmtch / DefOpt / tpu / Primal 0.0000021594 s 0.000002164025 s 1.00
actmtch / IDefOpt / tpu / Primal 0.00000210305 s 0.000002095375 s 1.00
actmtch / JaXPipe / tpu / Forward 0.000003819525 s 0.0000038255 s 1.00
actmtch / Jax / tpu / Forward 0.0000012143 s 0.0000012064999999999998 s 1.01
actmtch / HLOOpt / tpu / Forward 0.000003936225 s 0.0000039346750000000006 s 1.00
actmtch / PartOpt / tpu / Forward 0.000003917849999999999 s 0.00000392895 s 1.00
actmtch / IPartOpt / tpu / Forward 0.000003934375000000001 s 0.00000393315 s 1.00
actmtch / DefOpt / tpu / Forward 0.0000039066 s 0.00000391195 s 1.00
actmtch / IDefOpt / tpu / Forward 0.000003941225 s 0.0000039288250000000005 s 1.00
actmtch / JaXPipe / tpu / PreRev 0.0000034789750000000004 s 0.00000348385 s 1.00
actmtch / JaXPipe / tpu / PostRev 0.000001632725 s 0.000001634375 s 1.00
actmtch / JaXPipe / tpu / BothRev 0.0000034667500000000003 s 0.000003499325 s 0.99
actmtch / Jax / tpu / BothRev 0.0000016350999999999998 s 0.00000163605 s 1.00
actmtch / HLOOpt / tpu / PreRev 0.00000347495 s 0.00000348265 s 1.00
actmtch / HLOOpt / tpu / PostRev 0.000003410075 s 0.000003403275 s 1.00
actmtch / HLOOpt / tpu / BothRev 0.000003473925 s 0.00000347005 s 1.00
actmtch / PartOpt / tpu / PreRev 0.0000034061000000000003 s 0.00000340765 s 1.00
actmtch / PartOpt / tpu / PostRev 0.0000016018 s 0.0000015938749999999998 s 1.00
actmtch / PartOpt / tpu / BothRev 0.0000034093750000000003 s 0.000003406875 s 1.00
actmtch / IPartOpt / tpu / PreRev 0.000003480125 s 0.00000347205 s 1.00
actmtch / IPartOpt / tpu / PostRev 0.0000016339 s 0.00000163595 s 1.00
actmtch / IPartOpt / tpu / BothRev 0.0000034837249999999994 s 0.000003471625 s 1.00
actmtch / DefOpt / tpu / PreRev 0.000003405775 s 0.000003417425 s 1.00
actmtch / DefOpt / tpu / PostRev 0.000003415775 s 0.00000340655 s 1.00
actmtch / DefOpt / tpu / BothRev 0.000003416125 s 0.000003411675 s 1.00
actmtch / IDefOpt / tpu / PreRev 0.000003487 s 0.00000347155 s 1.00
actmtch / IDefOpt / tpu / PostRev 0.000003415475 s 0.000003411 s 1.00
actmtch / IDefOpt / tpu / BothRev 0.000003469975 s 0.000003475375 s 1.00
actmtch / JaXPipe / cpu / Primal 0.000013009 s 0.0000070261399650917154 s 1.85
actmtch / Jax / cpu / Primal 0.000013267 s 0.000007681720026084804 s 1.73
actmtch / HLOOpt / cpu / Primal 0.000013973 s 0.000008456099985778564 s 1.65
actmtch / PartOpt / cpu / Primal 0.000013176 s 0.000008354660039913142 s 1.58
actmtch / IPartOpt / cpu / Primal 0.00001348 s 0.000008498499983033981 s 1.59
actmtch / DefOpt / cpu / Primal 0.000014176 s 0.000008734280045246124 s 1.62
actmtch / IDefOpt / cpu / Primal 0.000013974 s 0.000008587439951952547 s 1.63
actmtch / JaXPipe / cpu / Forward 0.000019276000000000003 s 0.000012429179996615858 s 1.55
actmtch / Jax / cpu / Forward 0.000017835 s 0.000010656700023901068 s 1.67
actmtch / HLOOpt / cpu / Forward 0.00001858 s 0.000012589040015882348 s 1.48
actmtch / PartOpt / cpu / Forward 0.000018141 s 0.000011920439947061825 s 1.52
actmtch / IPartOpt / cpu / Forward 0.000018977 s 0.000012381419974190069 s 1.53
actmtch / DefOpt / cpu / Forward 0.000018433 s 0.000012271340056031477 s 1.50
actmtch / IDefOpt / cpu / Forward 0.000018822 s 0.000012316779984757888 s 1.53
actmtch / JaXPipe / cpu / PreRev 0.000019474 s 0.00001274931999432738 s 1.53
actmtch / JaXPipe / cpu / PostRev 0.000017204 s 0.000011261919989919989 s 1.53
actmtch / JaXPipe / cpu / BothRev 0.000019094 s 0.000012745980011459324 s 1.50
actmtch / Jax / cpu / BothRev 0.000017536 s 0.00001086691999262257 s 1.61
actmtch / HLOOpt / cpu / PreRev 0.000019077 s 0.000012499179983933571 s 1.53
actmtch / HLOOpt / cpu / PostRev 0.000019135 s 0.000014753979976376283 s 1.30
actmtch / HLOOpt / cpu / BothRev 0.000019217000000000003 s 0.00001177669998469355 s 1.63
actmtch / PartOpt / cpu / PreRev 0.000018395000000000003 s 0.000012723199997708434 s 1.45
actmtch / PartOpt / cpu / PostRev 0.000017412000000000002 s 0.000011315820029267342 s 1.54
actmtch / PartOpt / cpu / BothRev 0.000019476 s 0.000012766059990099166 s 1.53
actmtch / IPartOpt / cpu / PreRev 0.000019292 s 0.000012880159993073904 s 1.50
actmtch / IPartOpt / cpu / PostRev 0.000017224 s 0.000010472339963598645 s 1.64
actmtch / IPartOpt / cpu / BothRev 0.000020125 s 0.000012367339950287716 s 1.63
actmtch / DefOpt / cpu / PreRev 0.000018937 s 0.000012020940012007486 s 1.58
actmtch / DefOpt / cpu / PostRev 0.00001883 s 0.000012939700009155783 s 1.46
actmtch / DefOpt / cpu / BothRev 0.000019222 s 0.000012443320019883686 s 1.54
actmtch / IDefOpt / cpu / PreRev 0.000019002 s 0.000012142540026616188 s 1.56
actmtch / IDefOpt / cpu / PostRev 0.000019448 s 0.00001286107997657382 s 1.51
actmtch / IDefOpt / cpu / BothRev 0.000020138 s 0.000012399619981806609 s 1.62
actmtch / JaXPipe / cpu / Primal 0.00001 s 0.0000070261399650917154 s 1.42
actmtch / Jax / cpu / Primal 0.000008999999999999999 s 0.000007681720026084804 s 1.17
actmtch / HLOOpt / cpu / Primal 0.00001 s 0.000008456099985778564 s 1.18
actmtch / PartOpt / cpu / Primal 0.00001 s 0.000008354660039913142 s 1.20
actmtch / IPartOpt / cpu / Primal 0.000008999999999999999 s 0.000008498499983033981 s 1.06
actmtch / DefOpt / cpu / Primal 0.00001 s 0.000008734280045246124 s 1.14
actmtch / IDefOpt / cpu / Primal 0.00001 s 0.000008587439951952547 s 1.16
actmtch / JaXPipe / cpu / Forward 0.000015 s 0.000012429179996615858 s 1.21
actmtch / Jax / cpu / Forward 0.000013 s 0.000010656700023901068 s 1.22
actmtch / HLOOpt / cpu / Forward 0.000015 s 0.000012589040015882348 s 1.19
actmtch / PartOpt / cpu / Forward 0.000015 s 0.000011920439947061825 s 1.26
actmtch / IPartOpt / cpu / Forward 0.000014 s 0.000012381419974190069 s 1.13
actmtch / DefOpt / cpu / Forward 0.000014 s 0.000012271340056031477 s 1.14
actmtch / IDefOpt / cpu / Forward 0.000014 s 0.000012316779984757888 s 1.14
actmtch / JaXPipe / cpu / PreRev 0.000013 s 0.00001274931999432738 s 1.02
actmtch / JaXPipe / cpu / PostRev 0.000012 s 0.000011261919989919989 s 1.07
actmtch / JaXPipe / cpu / BothRev 0.000014 s 0.000012745980011459324 s 1.10
actmtch / Jax / cpu / BothRev 0.000013 s 0.00001086691999262257 s 1.20
actmtch / HLOOpt / cpu / PreRev 0.000014 s 0.000012499179983933571 s 1.12
actmtch / HLOOpt / cpu / PostRev 0.000014 s 0.000014753979976376283 s 0.95
actmtch / HLOOpt / cpu / BothRev 0.000014 s 0.00001177669998469355 s 1.19
actmtch / PartOpt / cpu / PreRev 0.000015 s 0.000012723199997708434 s 1.18
actmtch / PartOpt / cpu / PostRev 0.000013 s 0.000011315820029267342 s 1.15
actmtch / PartOpt / cpu / BothRev 0.000015 s 0.000012766059990099166 s 1.17
actmtch / IPartOpt / cpu / PreRev 0.000014 s 0.000012880159993073904 s 1.09
actmtch / IPartOpt / cpu / PostRev 0.000013 s 0.000010472339963598645 s 1.24
actmtch / IPartOpt / cpu / BothRev 0.000015 s 0.000012367339950287716 s 1.21
actmtch / DefOpt / cpu / PreRev 0.000014 s 0.000012020940012007486 s 1.16
actmtch / DefOpt / cpu / PostRev 0.000014 s 0.000012939700009155783 s 1.08
actmtch / DefOpt / cpu / BothRev 0.000014 s 0.000012443320019883686 s 1.13
actmtch / IDefOpt / cpu / PreRev 0.000014 s 0.000012142540026616188 s 1.15
actmtch / IDefOpt / cpu / PostRev 0.000015 s 0.00001286107997657382 s 1.17
actmtch / IDefOpt / cpu / BothRev 0.000014 s 0.000012399619981806609 s 1.13
add_one / JaXPipe / cpu / Primal 0.000006527319999349856 s 0.00000767638000070292 s 0.85
add_one / Jax / cpu / Primal 0.000006504780008071975 s 0.000007881059946157621 s 0.83
add_one / HLOOpt / cpu / Primal 0.000006477259994426276 s 0.00000769876004596881 s 0.84
add_one / PartOpt / cpu / Primal 0.000006698579993553722 s 0.00000769573998695705 s 0.87
add_one / IPartOpt / cpu / Primal 0.000006570080004166812 s 0.000007311640019906917 s 0.90
add_one / DefOpt / cpu / Primal 0.000006708960004289111 s 0.000006927979984538979 s 0.97
add_one / IDefOpt / cpu / Primal 0.000006464819994107529 s 0.000007248140027513728 s 0.89
add_one / JaXPipe / cpu / Forward 0.00001014962000454034 s 0.000011312520000501535 s 0.90
add_one / Jax / cpu / Forward 0.000009547139995902398 s 0.000010935220007013414 s 0.87
add_one / HLOOpt / cpu / Forward 0.000010000679994845995 s 0.00001113865996558161 s 0.90
add_one / PartOpt / cpu / Forward 0.00000981631999138699 s 0.000011150479995194472 s 0.88
add_one / IPartOpt / cpu / Forward 0.000010321519998797158 s 0.00001149292000263813 s 0.90
add_one / DefOpt / cpu / Forward 0.000009844679993875616 s 0.000011125779992653409 s 0.88
add_one / IDefOpt / cpu / Forward 0.000009882099989226844 s 0.000010953519959002734 s 0.90
add_one / JaXPipe / cpu / PreRev 0.000011326540000027308 s 0.00001329155999883369 s 0.85
add_one / JaXPipe / cpu / PostRev 0.00001145702001394966 s 0.00001282565998735663 s 0.89
add_one / JaXPipe / cpu / BothRev 0.000011704820005888903 s 0.000013342939955691691 s 0.88
add_one / Jax / cpu / BothRev 0.000011226359993088408 s 0.000013148820007700124 s 0.85
add_one / HLOOpt / cpu / PreRev 0.000011924759999146772 s 0.000013094879996060629 s 0.91
add_one / HLOOpt / cpu / PostRev 0.00001298861999885048 s 0.000014685159976579598 s 0.88
add_one / HLOOpt / cpu / BothRev 0.00001104727999518218 s 0.000013228520047050553 s 0.84
add_one / PartOpt / cpu / PreRev 0.00001146232000337477 s 0.000012965020023329998 s 0.88
add_one / PartOpt / cpu / PostRev 0.000010983319991737516 s 0.000012663999968935967 s 0.87
add_one / PartOpt / cpu / BothRev 0.000011836359999506384 s 0.000014077939986236744 s 0.84
add_one / IPartOpt / cpu / PreRev 0.000011301260001346235 s 0.00001328809998994984 s 0.85
add_one / IPartOpt / cpu / PostRev 0.000010790559999804829 s 0.00001288855998609506 s 0.84
add_one / IPartOpt / cpu / BothRev 0.000011510620004173688 s 0.000013109620003888268 s 0.88
add_one / DefOpt / cpu / PreRev 0.00001107387999581988 s 0.000013311499997143982 s 0.83
add_one / DefOpt / cpu / PostRev 0.000011495440014641643 s 0.000012807700004486833 s 0.90
add_one / DefOpt / cpu / BothRev 0.000011469480002688214 s 0.00001317183998253313 s 0.87
add_one / IDefOpt / cpu / PreRev 0.000011350160002621124 s 0.00001243973999407899 s 0.91
add_one / IDefOpt / cpu / PostRev 0.000011141860009047375 s 0.00001230946002579003 s 0.91
add_one / IDefOpt / cpu / BothRev 0.00001152155999761817 s 0.00001255659997696057 s 0.92
add_one / JaXPipe / cuda / Primal 0.000002303 s 0.0000019200000000000003 s 1.20
add_one / Jax / cuda / Primal 0.000002303 s 0.0000019200000000000003 s 1.20
add_one / HLOOpt / cuda / Primal 0.000002304 s 0.0000019200000000000003 s 1.20
add_one / PartOpt / cuda / Primal 0.000002303 s 0.0000019200000000000003 s 1.20
add_one / IPartOpt / cuda / Primal 0.000002303 s 0.0000019200000000000003 s 1.20
add_one / DefOpt / cuda / Primal 0.000002303 s 0.0000019200000000000003 s 1.20
add_one / IDefOpt / cuda / Primal 0.000002272 s 0.0000019200000000000003 s 1.18
add_one / JaXPipe / cuda / Forward 0.00001024 s 0.000010016 s 1.02
add_one / Jax / cuda / Forward 0.0000104 s 0.000009792 s 1.06
add_one / HLOOpt / cuda / Forward 0.000010464 s 0.000010145 s 1.03
add_one / PartOpt / cuda / Forward 0.00001088 s 0.00001008 s 1.08
add_one / IPartOpt / cuda / Forward 0.000010433 s 0.00000992 s 1.05
add_one / DefOpt / cuda / Forward 0.000009919 s 0.000010144 s 0.98
add_one / IDefOpt / cuda / Forward 0.00001072 s 0.000009761 s 1.10
add_one / JaXPipe / cuda / PreRev 0.000026176 s 0.000024704 s 1.06
add_one / JaXPipe / cuda / PostRev 0.0000256 s 0.000024256 s 1.06
add_one / JaXPipe / cuda / BothRev 0.000024928 s 0.000024577 s 1.01
add_one / Jax / cuda / BothRev 0.00002544 s 0.000025088 s 1.01
add_one / HLOOpt / cuda / PreRev 0.000025632 s 0.000025088 s 1.02
add_one / HLOOpt / cuda / PostRev 0.000025344 s 0.000024256 s 1.04
add_one / HLOOpt / cuda / BothRev 0.0000256 s 0.0000312 s 0.82
add_one / PartOpt / cuda / PreRev 0.00002512 s 0.000024896 s 1.01
add_one / PartOpt / cuda / PostRev 0.000025824 s 0.000024609 s 1.05
add_one / PartOpt / cuda / BothRev 0.000025728 s 0.000024768 s 1.04
add_one / IPartOpt / cuda / PreRev 0.000025856 s 0.000024768 s 1.04
add_one / IPartOpt / cuda / PostRev 0.000025984 s 0.000024704 s 1.05
add_one / IPartOpt / cuda / BothRev 0.000025696 s 0.000023808 s 1.08
add_one / DefOpt / cuda / PreRev 0.000025664 s 0.000025024 s 1.03
add_one / DefOpt / cuda / PostRev 0.000025696 s 0.000025217 s 1.02
add_one / DefOpt / cuda / BothRev 0.00002576 s 0.00002528 s 1.02
add_one / IDefOpt / cuda / PreRev 0.000026336 s 0.00002496 s 1.06
add_one / IDefOpt / cuda / PostRev 0.000024992 s 0.00002496 s 1.00
add_one / IDefOpt / cuda / BothRev 0.000025791 s 0.00002432 s 1.06
add_one / JaXPipe / tpu / Primal 0.000001422725 s 0.0000014234749999999995 s 1.00
add_one / Jax / tpu / Primal 0.000001399125 s 0.00000140335 s 1.00
add_one / HLOOpt / tpu / Primal 0.0000014349999999999998 s 0.000001423325 s 1.01
add_one / PartOpt / tpu / Primal 0.00000141155 s 0.0000014036 s 1.01
add_one / IPartOpt / tpu / Primal 0.0000014268500000000002 s 0.000001422475 s 1.00
add_one / DefOpt / tpu / Primal 0.0000014048749999999998 s 0.000001397825 s 1.01
add_one / IDefOpt / tpu / Primal 0.0000014229 s 0.000001422125 s 1.00
add_one / JaXPipe / tpu / Forward 0.00000185045 s 0.00000184455 s 1.00
add_one / Jax / tpu / Forward 0.000001838 s 0.000001838475 s 1.00
add_one / HLOOpt / tpu / Forward 0.00000184775 s 0.000001845525 s 1.00
add_one / PartOpt / tpu / Forward 0.00000183505 s 0.00000183625 s 1.00
add_one / IPartOpt / tpu / Forward 0.00000185085 s 0.0000018463 s 1.00
add_one / DefOpt / tpu / Forward 0.000001838075 s 0.00000183405 s 1.00
add_one / IDefOpt / tpu / Forward 0.000001843775 s 0.00000185615 s 0.99
add_one / JaXPipe / tpu / PreRev 0.000002246525 s 0.0000022316 s 1.01
add_one / JaXPipe / tpu / PostRev 0.000002234025 s 0.000002234475 s 1.00
add_one / JaXPipe / tpu / BothRev 0.00000223895 s 0.000002245975 s 1.00
add_one / Jax / tpu / BothRev 0.0000022354250000000003 s 0.00000224375 s 1.00
add_one / HLOOpt / tpu / PreRev 0.0000022336750000000003 s 0.000002241625 s 1.00
add_one / HLOOpt / tpu / PostRev 0.00000224425 s 0.0000022375999999999995 s 1.00
add_one / HLOOpt / tpu / BothRev 0.000002241425 s 0.0000022452 s 1.00
add_one / PartOpt / tpu / PreRev 0.00000224165 s 0.00000223565 s 1.00
add_one / PartOpt / tpu / PostRev 0.000002253175 s 0.00000224485 s 1.00
add_one / PartOpt / tpu / BothRev 0.00000224235 s 0.0000022485 s 1.00
add_one / IPartOpt / tpu / PreRev 0.000002242075 s 0.0000022385 s 1.00
add_one / IPartOpt / tpu / PostRev 0.000002237775 s 0.000002240525 s 1.00
add_one / IPartOpt / tpu / BothRev 0.000002242175 s 0.0000022301 s 1.01
add_one / DefOpt / tpu / PreRev 0.00000224545 s 0.0000022470250000000003 s 1.00
add_one / DefOpt / tpu / PostRev 0.000002236075 s 0.00000223715 s 1.00
add_one / DefOpt / tpu / BothRev 0.000002252375 s 0.000002235075 s 1.01
add_one / IDefOpt / tpu / PreRev 0.000002242 s 0.000002230725 s 1.01
add_one / IDefOpt / tpu / PostRev 0.000002245225 s 0.000002232775 s 1.01
add_one / IDefOpt / tpu / BothRev 0.0000022384 s 0.000002229975 s 1.00
add_one / JaXPipe / cpu / Primal 0.000013332 s 0.00000767638000070292 s 1.74
add_one / Jax / cpu / Primal 0.00001297 s 0.000007881059946157621 s 1.65
add_one / HLOOpt / cpu / Primal 0.000012952 s 0.00000769876004596881 s 1.68
add_one / PartOpt / cpu / Primal 0.000012952 s 0.00000769573998695705 s 1.68
add_one / IPartOpt / cpu / Primal 0.000012478 s 0.000007311640019906917 s 1.71
add_one / DefOpt / cpu / Primal 0.000013041 s 0.000006927979984538979 s 1.88
add_one / IDefOpt / cpu / Primal 0.000013255 s 0.000007248140027513728 s 1.83
add_one / JaXPipe / cpu / Forward 0.000017367 s 0.000011312520000501535 s 1.54
add_one / Jax / cpu / Forward 0.000017168 s 0.000010935220007013414 s 1.57
add_one / HLOOpt / cpu / Forward 0.000017364000000000002 s 0.00001113865996558161 s 1.56
add_one / PartOpt / cpu / Forward 0.000017371 s 0.000011150479995194472 s 1.56
add_one / IPartOpt / cpu / Forward 0.000016952 s 0.00001149292000263813 s 1.47
add_one / DefOpt / cpu / Forward 0.000017641 s 0.000011125779992653409 s 1.59
add_one / IDefOpt / cpu / Forward 0.00001714 s 0.000010953519959002734 s 1.56
add_one / JaXPipe / cpu / PreRev 0.00001976 s 0.00001329155999883369 s 1.49
add_one / JaXPipe / cpu / PostRev 0.000019379 s 0.00001282565998735663 s 1.51
add_one / JaXPipe / cpu / BothRev 0.000019312 s 0.000013342939955691691 s 1.45
add_one / Jax / cpu / BothRev 0.000019316 s 0.000013148820007700124 s 1.47
add_one / HLOOpt / cpu / PreRev 0.000019938 s 0.000013094879996060629 s 1.52
add_one / HLOOpt / cpu / PostRev 0.000019886 s 0.000014685159976579598 s 1.35
add_one / HLOOpt / cpu / BothRev 0.0000196 s 0.000013228520047050553 s 1.48
add_one / PartOpt / cpu / PreRev 0.000019641 s 0.000012965020023329998 s 1.51
add_one / PartOpt / cpu / PostRev 0.000019701 s 0.000012663999968935967 s 1.56
add_one / PartOpt / cpu / BothRev 0.000019832 s 0.000014077939986236744 s 1.41
add_one / IPartOpt / cpu / PreRev 0.000020519 s 0.00001328809998994984 s 1.54
add_one / IPartOpt / cpu / PostRev 0.000019593 s 0.00001288855998609506 s 1.52
add_one / IPartOpt / cpu / BothRev 0.000020029 s 0.000013109620003888268 s 1.53
add_one / DefOpt / cpu / PreRev 0.0000198 s 0.000013311499997143982 s 1.49
add_one / DefOpt / cpu / PostRev 0.000019896 s 0.000012807700004486833 s 1.55
add_one / DefOpt / cpu / BothRev 0.000020492 s 0.00001317183998253313 s 1.56
add_one / IDefOpt / cpu / PreRev 0.000019734 s 0.00001243973999407899 s 1.59
add_one / IDefOpt / cpu / PostRev 0.000020003 s 0.00001230946002579003 s 1.63
add_one / IDefOpt / cpu / BothRev 0.000020178 s 0.00001255659997696057 s 1.61
add_one / JaXPipe / cpu / Primal 0.000008999999999999999 s 0.00000767638000070292 s 1.17
add_one / Jax / cpu / Primal 0.000008999999999999999 s 0.000007881059946157621 s 1.14
add_one / HLOOpt / cpu / Primal 0.000008999999999999999 s 0.00000769876004596881 s 1.17
add_one / PartOpt / cpu / Primal 0.000008999999999999999 s 0.00000769573998695705 s 1.17
add_one / IPartOpt / cpu / Primal 0.000008999999999999999 s 0.000007311640019906917 s 1.23
add_one / DefOpt / cpu / Primal 0.000008999999999999999 s 0.000006927979984538979 s 1.30
add_one / IDefOpt / cpu / Primal 0.000008999999999999999 s 0.000007248140027513728 s 1.24
add_one / JaXPipe / cpu / Forward 0.000012 s 0.000011312520000501535 s 1.06
add_one / Jax / cpu / Forward 0.000012 s 0.000010935220007013414 s 1.10
add_one / HLOOpt / cpu / Forward 0.000012 s 0.00001113865996558161 s 1.08
add_one / PartOpt / cpu / Forward 0.000012 s 0.000011150479995194472 s 1.08
add_one / IPartOpt / cpu / Forward 0.000012 s 0.00001149292000263813 s 1.04
add_one / DefOpt / cpu / Forward 0.000012 s 0.000011125779992653409 s 1.08
add_one / IDefOpt / cpu / Forward 0.000012 s 0.000010953519959002734 s 1.10
add_one / JaXPipe / cpu / PreRev 0.000014 s 0.00001329155999883369 s 1.05
add_one / JaXPipe / cpu / PostRev 0.000014 s 0.00001282565998735663 s 1.09
add_one / JaXPipe / cpu / BothRev 0.000014 s 0.000013342939955691691 s 1.05
add_one / Jax / cpu / BothRev 0.000014 s 0.000013148820007700124 s 1.06
add_one / HLOOpt / cpu / PreRev 0.000014 s 0.000013094879996060629 s 1.07
add_one / HLOOpt / cpu / PostRev 0.000014 s 0.000014685159976579598 s 0.95
add_one / HLOOpt / cpu / BothRev 0.000014 s 0.000013228520047050553 s 1.06
add_one / PartOpt / cpu / PreRev 0.000014 s 0.000012965020023329998 s 1.08
add_one / PartOpt / cpu / PostRev 0.000014 s 0.000012663999968935967 s 1.11
add_one / PartOpt / cpu / BothRev 0.000014 s 0.000014077939986236744 s 0.99
add_one / IPartOpt / cpu / PreRev 0.000014 s 0.00001328809998994984 s 1.05
add_one / IPartOpt / cpu / PostRev 0.000014 s 0.00001288855998609506 s 1.09
add_one / IPartOpt / cpu / BothRev 0.000015 s 0.000013109620003888268 s 1.14
add_one / DefOpt / cpu / PreRev 0.000014 s 0.000013311499997143982 s 1.05
add_one / DefOpt / cpu / PostRev 0.000015 s 0.000012807700004486833 s 1.17
add_one / DefOpt / cpu / BothRev 0.000014 s 0.00001317183998253313 s 1.06
add_one / IDefOpt / cpu / PreRev 0.000014 s 0.00001243973999407899 s 1.13
add_one / IDefOpt / cpu / PostRev 0.000014 s 0.00001230946002579003 s 1.14
add_one / IDefOpt / cpu / BothRev 0.000014 s 0.00001255659997696057 s 1.11
add_two / JaXPipe / cpu / Primal 0.000007181719995514868 s 0.00000840646003780421 s 0.85
add_two / Jax / cpu / Primal 0.000007412120000935829 s 0.000007408079991364502 s 1.00
add_two / HLOOpt / cpu / Primal 0.000006910080007855868 s 0.0000074600599691621025 s 0.93
add_two / PartOpt / cpu / Primal 0.000007012280004801141 s 0.000007752780065857223 s 0.90
add_two / IPartOpt / cpu / Primal 0.000006975419996706477 s 0.000008030020026126295 s 0.87
add_two / DefOpt / cpu / Primal 0.000007349439990775863 s 0.00000800805999460863 s 0.92
add_two / IDefOpt / cpu / Primal 0.000006735419992764946 s 0.000007063699977152282 s 0.95
add_two / JaXPipe / cpu / Forward 0.00001016190000655115 s 0.000011770780010920136 s 0.86
add_two / Jax / cpu / Forward 0.000010153839998565672 s 0.000011452759990788764 s 0.89
add_two / HLOOpt / cpu / Forward 0.000010582480003904493 s 0.00001166930002909794 s 0.91
add_two / PartOpt / cpu / Forward 0.000010770500002763584 s 0.000011410620018068583 s 0.94
add_two / IPartOpt / cpu / Forward 0.000009896839992507012 s 0.000011299139960101456 s 0.88
add_two / DefOpt / cpu / Forward 0.0000102615999912814 s 0.000011692319976646103 s 0.88
add_two / IDefOpt / cpu / Forward 0.000010313920001863152 s 0.000011732559996744386 s 0.88
add_two / JaXPipe / cpu / PreRev 0.000013803239999106156 s 0.000015161400006036274 s 0.91
add_two / JaXPipe / cpu / PostRev 0.00001332511999862618 s 0.00001595720002114831 s 0.84
add_two / JaXPipe / cpu / BothRev 0.00001358063998623038 s 0.00001547035995827173 s 0.88
add_two / Jax / cpu / BothRev 0.000013045000009697106 s 0.00001497119998020935 s 0.87
add_two / HLOOpt / cpu / PreRev 0.000013749100003224157 s 0.000015616819973729433 s 0.88
add_two / HLOOpt / cpu / PostRev 0.000014910239988239482 s 0.000016793039994809077 s 0.89
add_two / HLOOpt / cpu / BothRev 0.000013744620005127218 s 0.000015394640031445305 s 0.89
add_two / PartOpt / cpu / PreRev 0.000013719960006710608 s 0.00001552674003505672 s 0.88
add_two / PartOpt / cpu / PostRev 0.000012994620001336442 s 0.00001590923999174265 s 0.82
add_two / PartOpt / cpu / BothRev 0.000013382699999056058 s 0.0000157065999792394 s 0.85
add_two / IPartOpt / cpu / PreRev 0.000013732899994920444 s 0.00001480517998970754 s 0.93
add_two / IPartOpt / cpu / PostRev 0.000013543900015520194 s 0.000015164839996941735 s 0.89
add_two / IPartOpt / cpu / BothRev 0.000013044039990290912 s 0.000015474340025320998 s 0.84
add_two / DefOpt / cpu / PreRev 0.00001376413999878423 s 0.00001505929999439104 s 0.91
add_two / DefOpt / cpu / PostRev 0.000013500779998594223 s 0.00001657381996665208 s 0.81
add_two / DefOpt / cpu / BothRev 0.000013454059997002331 s 0.000015267580029103555 s 0.88
add_two / IDefOpt / cpu / PreRev 0.00001369823999084474 s 0.000015865159994064016 s 0.86
add_two / IDefOpt / cpu / PostRev 0.00001319982000040909 s 0.00001576899999236048 s 0.84
add_two / IDefOpt / cpu / BothRev 0.000013646219995280264 s 0.000015866280000409462 s 0.86
add_two / JaXPipe / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
add_two / Jax / cuda / Primal 0.0000024 s 0.000001888 s 1.27
add_two / HLOOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
add_two / PartOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
add_two / IPartOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
add_two / DefOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
add_two / IDefOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
add_two / JaXPipe / cuda / Forward 0.00001024 s 0.000009504 s 1.08
add_two / Jax / cuda / Forward 0.000010144 s 0.00000992 s 1.02
add_two / HLOOpt / cuda / Forward 0.000010368 s 0.000009888 s 1.05
add_two / PartOpt / cuda / Forward 0.00000992 s 0.000009153 s 1.08
add_two / IPartOpt / cuda / Forward 0.000010367 s 0.00000992 s 1.05
add_two / DefOpt / cuda / Forward 0.000011328 s 0.000009536 s 1.19
add_two / IDefOpt / cuda / Forward 0.000010368 s 0.000009568 s 1.08
add_two / JaXPipe / cuda / PreRev 0.000034111 s 0.000031585 s 1.08
add_two / JaXPipe / cuda / PostRev 0.000033119999999999995 s 0.000031264 s 1.06
add_two / JaXPipe / cuda / BothRev 0.00003808 s 0.000031744 s 1.20
add_two / Jax / cuda / BothRev 0.000034528000000000006 s 0.00003152 s 1.10
add_two / HLOOpt / cuda / PreRev 0.000033664 s 0.000032 s 1.05
add_two / HLOOpt / cuda / PostRev 0.000034144000000000004 s 0.000031424 s 1.09
add_two / HLOOpt / cuda / BothRev 0.000032864 s 0.000041184 s 0.80
add_two / PartOpt / cuda / PreRev 0.00003296 s 0.000031584 s 1.04
add_two / PartOpt / cuda / PostRev 0.00003824 s 0.000031392 s 1.22
add_two / PartOpt / cuda / BothRev 0.000038304 s 0.00003136 s 1.22
add_two / IPartOpt / cuda / PreRev 0.000033728 s 0.000032896000000000005 s 1.03
add_two / IPartOpt / cuda / PostRev 0.000033536000000000006 s 0.000031616 s 1.06
add_two / IPartOpt / cuda / BothRev 0.000033567 s 0.000032192 s 1.04
add_two / DefOpt / cuda / PreRev 0.0000416 s 0.000031776 s 1.31
add_two / DefOpt / cuda / PostRev 0.000032064 s 0.00003184 s 1.01
add_two / DefOpt / cuda / BothRev 0.000033311 s 0.000032608 s 1.02
add_two / IDefOpt / cuda / PreRev 0.000033312 s 0.000036225 s 0.92
add_two / IDefOpt / cuda / PostRev 0.000032416 s 0.000035904 s 0.90
add_two / IDefOpt / cuda / BothRev 0.000033855 s 0.000035327 s 0.96
add_two / JaXPipe / tpu / Primal 0.000001434825 s 0.0000014340249999999998 s 1.00
add_two / Jax / tpu / Primal 0.000001472925 s 0.000001485725 s 0.99
add_two / HLOOpt / tpu / Primal 0.000001430725 s 0.000001440225 s 0.99
add_two / PartOpt / tpu / Primal 0.000001479875 s 0.00000147715 s 1.00
add_two / IPartOpt / tpu / Primal 0.00000143225 s 0.000001432175 s 1.00
add_two / DefOpt / tpu / Primal 0.00000147735 s 0.00000147175 s 1.00
add_two / IDefOpt / tpu / Primal 0.000001445225 s 0.00000144305 s 1.00
add_two / JaXPipe / tpu / Forward 0.000001835 s 0.000001820975 s 1.01
add_two / Jax / tpu / Forward 0.000001836925 s 0.000001826575 s 1.01
add_two / HLOOpt / tpu / Forward 0.0000018329 s 0.0000018237 s 1.01
add_two / PartOpt / tpu / Forward 0.000001821525 s 0.000001831525 s 0.99
add_two / IPartOpt / tpu / Forward 0.000001830775 s 0.0000018267 s 1.00
add_two / DefOpt / tpu / Forward 0.0000018227 s 0.000001830275 s 1.00
add_two / IDefOpt / tpu / Forward 0.0000018356 s 0.000001832075 s 1.00
add_two / JaXPipe / tpu / PreRev 0.0000028284 s 0.000002834025 s 1.00
add_two / JaXPipe / tpu / PostRev 0.00000274375 s 0.000002750375 s 1.00
add_two / JaXPipe / tpu / BothRev 0.000002837675 s 0.0000028336250000000003 s 1.00
add_two / Jax / tpu / BothRev 0.0000027488499999999995 s 0.0000027459250000000003 s 1.00
add_two / HLOOpt / tpu / PreRev 0.000002827175 s 0.0000028386 s 1.00
add_two / HLOOpt / tpu / PostRev 0.0000027509 s 0.000002751625 s 1.00
add_two / HLOOpt / tpu / BothRev 0.000002825475 s 0.0000028408 s 0.99
add_two / PartOpt / tpu / PreRev 0.000002755475 s 0.0000027445750000000003 s 1.00
add_two / PartOpt / tpu / PostRev 0.0000028429500000000005 s 0.000002841075 s 1.00
add_two / PartOpt / tpu / BothRev 0.000002756675 s 0.00000274485 s 1.00
add_two / IPartOpt / tpu / PreRev 0.000002831275 s 0.00000282995 s 1.00
add_two / IPartOpt / tpu / PostRev 0.000002745675 s 0.0000027511750000000004 s 1.00
add_two / IPartOpt / tpu / BothRev 0.0000028472 s 0.0000028333000000000005 s 1.00
add_two / DefOpt / tpu / PreRev 0.000002760225 s 0.000002745525 s 1.01
add_two / DefOpt / tpu / PostRev 0.0000028344000000000003 s 0.00000283105 s 1.00
add_two / DefOpt / tpu / BothRev 0.000002744925 s 0.000002757425 s 1.00
add_two / IDefOpt / tpu / PreRev 0.0000028378500000000003 s 0.000002837325 s 1.00
add_two / IDefOpt / tpu / PostRev 0.000002747075 s 0.000002747225 s 1.00
add_two / IDefOpt / tpu / BothRev 0.0000028333000000000005 s 0.00000283425 s 1.00
add_two / JaXPipe / cpu / Primal 0.000013482 s 0.00000840646003780421 s 1.60
add_two / Jax / cpu / Primal 0.000013202 s 0.000007408079991364502 s 1.78
add_two / HLOOpt / cpu / Primal 0.000012844 s 0.0000074600599691621025 s 1.72
add_two / PartOpt / cpu / Primal 0.00001331 s 0.000007752780065857223 s 1.72
add_two / IPartOpt / cpu / Primal 0.000013283 s 0.000008030020026126295 s 1.65
add_two / DefOpt / cpu / Primal 0.000013259 s 0.00000800805999460863 s 1.66
add_two / IDefOpt / cpu / Primal 0.000013055 s 0.000007063699977152282 s 1.85
add_two / JaXPipe / cpu / Forward 0.00001848 s 0.000011770780010920136 s 1.57
add_two / Jax / cpu / Forward 0.000018288 s 0.000011452759990788764 s 1.60
add_two / HLOOpt / cpu / Forward 0.000018378 s 0.00001166930002909794 s 1.57
add_two / PartOpt / cpu / Forward 0.000018022 s 0.000011410620018068583 s 1.58
add_two / IPartOpt / cpu / Forward 0.000017988 s 0.000011299139960101456 s 1.59
add_two / DefOpt / cpu / Forward 0.000017684 s 0.000011692319976646103 s 1.51
add_two / IDefOpt / cpu / Forward 0.000018017 s 0.000011732559996744386 s 1.54
add_two / JaXPipe / cpu / PreRev 0.000022969 s 0.000015161400006036274 s 1.51
add_two / JaXPipe / cpu / PostRev 0.000023239 s 0.00001595720002114831 s 1.46
add_two / JaXPipe / cpu / BothRev 0.000023364 s 0.00001547035995827173 s 1.51
add_two / Jax / cpu / BothRev 0.000022512 s 0.00001497119998020935 s 1.50
add_two / HLOOpt / cpu / PreRev 0.000022923 s 0.000015616819973729433 s 1.47
add_two / HLOOpt / cpu / PostRev 0.000023641 s 0.000016793039994809077 s 1.41
add_two / HLOOpt / cpu / BothRev 0.000023294 s 0.000015394640031445305 s 1.51
add_two / PartOpt / cpu / PreRev 0.000023112 s 0.00001552674003505672 s 1.49
add_two / PartOpt / cpu / PostRev 0.000023105 s 0.00001590923999174265 s 1.45
add_two / PartOpt / cpu / BothRev 0.000023044 s 0.0000157065999792394 s 1.47
add_two / IPartOpt / cpu / PreRev 0.000023136 s 0.00001480517998970754 s 1.56
add_two / IPartOpt / cpu / PostRev 0.000022811 s 0.000015164839996941735 s 1.50
add_two / IPartOpt / cpu / BothRev 0.000023166 s 0.000015474340025320998 s 1.50
add_two / DefOpt / cpu / PreRev 0.000023319 s 0.00001505929999439104 s 1.55
add_two / DefOpt / cpu / PostRev 0.000022747 s 0.00001657381996665208 s 1.37
add_two / DefOpt / cpu / BothRev 0.000023235 s 0.000015267580029103555 s 1.52
add_two / IDefOpt / cpu / PreRev 0.000023282 s 0.000015865159994064016 s 1.47
add_two / IDefOpt / cpu / PostRev 0.000022865 s 0.00001576899999236048 s 1.45
add_two / IDefOpt / cpu / BothRev 0.000023368 s 0.000015866280000409462 s 1.47
add_two / JaXPipe / cpu / Primal 0.000008999999999999999 s 0.00000840646003780421 s 1.07
add_two / Jax / cpu / Primal 0.000008999999999999999 s 0.000007408079991364502 s 1.21
add_two / HLOOpt / cpu / Primal 0.000008999999999999999 s 0.0000074600599691621025 s 1.21
add_two / PartOpt / cpu / Primal 0.00001 s 0.000007752780065857223 s 1.29
add_two / IPartOpt / cpu / Primal 0.000008999999999999999 s 0.000008030020026126295 s 1.12
add_two / DefOpt / cpu / Primal 0.000008999999999999999 s 0.00000800805999460863 s 1.12
add_two / IDefOpt / cpu / Primal 0.000008999999999999999 s 0.000007063699977152282 s 1.27
add_two / JaXPipe / cpu / Forward 0.000013 s 0.000011770780010920136 s 1.10
add_two / Jax / cpu / Forward 0.000013 s 0.000011452759990788764 s 1.14
add_two / HLOOpt / cpu / Forward 0.000013 s 0.00001166930002909794 s 1.11
add_two / PartOpt / cpu / Forward 0.000013 s 0.000011410620018068583 s 1.14
add_two / IPartOpt / cpu / Forward 0.000013 s 0.000011299139960101456 s 1.15
add_two / DefOpt / cpu / Forward 0.000013 s 0.000011692319976646103 s 1.11
add_two / IDefOpt / cpu / Forward 0.000013 s 0.000011732559996744386 s 1.11
add_two / JaXPipe / cpu / PreRev 0.000017 s 0.000015161400006036274 s 1.12
add_two / JaXPipe / cpu / PostRev 0.000017 s 0.00001595720002114831 s 1.07
add_two / JaXPipe / cpu / BothRev 0.000017 s 0.00001547035995827173 s 1.10
add_two / Jax / cpu / BothRev 0.000017 s 0.00001497119998020935 s 1.14
add_two / HLOOpt / cpu / PreRev 0.000017 s 0.000015616819973729433 s 1.09
add_two / HLOOpt / cpu / PostRev 0.000017999999999999997 s 0.000016793039994809077 s 1.07
add_two / HLOOpt / cpu / BothRev 0.000017 s 0.000015394640031445305 s 1.10
add_two / PartOpt / cpu / PreRev 0.000017 s 0.00001552674003505672 s 1.09
add_two / PartOpt / cpu / PostRev 0.000017 s 0.00001590923999174265 s 1.07
add_two / PartOpt / cpu / BothRev 0.000017 s 0.0000157065999792394 s 1.08
add_two / IPartOpt / cpu / PreRev 0.000017 s 0.00001480517998970754 s 1.15
add_two / IPartOpt / cpu / PostRev 0.000017 s 0.000015164839996941735 s 1.12
add_two / IPartOpt / cpu / BothRev 0.000017 s 0.000015474340025320998 s 1.10
add_two / DefOpt / cpu / PreRev 0.000016 s 0.00001505929999439104 s 1.06
add_two / DefOpt / cpu / PostRev 0.000017 s 0.00001657381996665208 s 1.03
add_two / DefOpt / cpu / BothRev 0.000017 s 0.000015267580029103555 s 1.11
add_two / IDefOpt / cpu / PreRev 0.000017 s 0.000015865159994064016 s 1.07
add_two / IDefOpt / cpu / PostRev 0.000017 s 0.00001576899999236048 s 1.08
add_two / IDefOpt / cpu / BothRev 0.000017 s 0.000015866280000409462 s 1.07
cache / JaXPipe / cpu / Primal 0.000006057979994693597 s 0.000006617600029130699 s 0.92
cache / Jax / cpu / Primal 0.0000067347599929235006 s 0.0000073614199845906115 s 0.91
cache / HLOOpt / cpu / Primal 0.0000061146600023676 s 0.000007327980010813917 s 0.83
cache / PartOpt / cpu / Primal 0.00000605648000146175 s 0.000006596160001208773 s 0.92
cache / IPartOpt / cpu / Primal 0.000006027340002674464 s 0.000006951079985810793 s 0.87
cache / DefOpt / cpu / Primal 0.000005973320012344629 s 0.000006810800023231423 s 0.88
cache / IDefOpt / cpu / Primal 0.000005934379996688221 s 0.000006719999992128578 s 0.88
cache / JaXPipe / cpu / Forward 0.000014143180003429734 s 0.00001635375998375821 s 0.86
cache / Jax / cpu / Forward 0.000013643900006172771 s 0.00001622899997528293 s 0.84
cache / HLOOpt / cpu / Forward 0.000014762459993562516 s 0.00001638741993701842 s 0.90
cache / PartOpt / cpu / Forward 0.000014526640004532965 s 0.00001637195999137475 s 0.89
cache / IPartOpt / cpu / Forward 0.000015243199979977364 s 0.000016631440003038735 s 0.92
cache / DefOpt / cpu / Forward 0.000014254939999318597 s 0.00001649958000598417 s 0.86
cache / IDefOpt / cpu / Forward 0.000014173799988839163 s 0.00001611459998457576 s 0.88
cache / JaXPipe / cpu / PreRev 0.00001473585998382987 s 0.000017117220031650505 s 0.86
cache / JaXPipe / cpu / PostRev 0.00001951246000089668 s 0.00002205438004239113 s 0.88
cache / JaXPipe / cpu / BothRev 0.000016874440000265167 s 0.00001823537998461688 s 0.93
cache / Jax / cpu / BothRev 0.000019304400000237363 s 0.00002179658003115037 s 0.89
cache / HLOOpt / cpu / PreRev 0.000015891100006228953 s 0.00001804116001949296 s 0.88
cache / HLOOpt / cpu / PostRev 0.0000175184200043077 s 0.000020775400025740963 s 0.84
cache / HLOOpt / cpu / BothRev 0.000015117539990114892 s 0.000018207139974038 s 0.83
cache / PartOpt / cpu / PreRev 0.00001538501999220898 s 0.000017113039994001157 s 0.90
cache / PartOpt / cpu / PostRev 0.00001975966000372864 s 0.000022788399955970818 s 0.87
cache / PartOpt / cpu / BothRev 0.000015390600003684086 s 0.000017054319996532286 s 0.90
cache / IPartOpt / cpu / PreRev 0.000015320739998969656 s 0.00001702454002952436 s 0.90
cache / IPartOpt / cpu / PostRev 0.000019720619995950983 s 0.00002129747997969389 s 0.93
cache / IPartOpt / cpu / BothRev 0.0000153360399963276 s 0.00001820888001930143 s 0.84
cache / DefOpt / cpu / PreRev 0.000015187940009582237 s 0.000016552119950574708 s 0.92
cache / DefOpt / cpu / PostRev 0.000015252539997163694 s 0.00001614170001630555 s 0.94
cache / DefOpt / cpu / BothRev 0.000016388180006288168 s 0.000016149760012922344 s 1.01
cache / IDefOpt / cpu / PreRev 0.000014925919995221193 s 0.000016202219967453857 s 0.92
cache / IDefOpt / cpu / PostRev 0.00001514337999651616 s 0.000015889159994912917 s 0.95
cache / IDefOpt / cpu / BothRev 0.000015097839993813975 s 0.000016051540005719288 s 0.94
cache / JaXPipe / cuda / Primal 0.0000023670000000000004 s 0.000002303 s 1.03
cache / Jax / cuda / Primal 0.000002336 s 0.000002304 s 1.01
cache / HLOOpt / cuda / Primal 0.0000023670000000000004 s 0.000002208 s 1.07
cache / PartOpt / cuda / Primal 0.000002336 s 0.000002207 s 1.06
cache / IPartOpt / cuda / Primal 0.0000023670000000000004 s 0.000002303 s 1.03
cache / DefOpt / cuda / Primal 0.000002336 s 0.000002272 s 1.03
cache / IDefOpt / cuda / Primal 0.000002336 s 0.00000224 s 1.04
cache / JaXPipe / cuda / Forward 0.0000023670000000000004 s 0.000002335 s 1.01
cache / Jax / cuda / Forward 0.000002368 s 0.000002335 s 1.01
cache / HLOOpt / cuda / Forward 0.0000023670000000000004 s 0.000002335 s 1.01
cache / PartOpt / cuda / Forward 0.000002368 s 0.000002335 s 1.01
cache / IPartOpt / cuda / Forward 0.0000023670000000000004 s 0.000002335 s 1.01
cache / DefOpt / cuda / Forward 0.0000023670000000000004 s 0.000002272 s 1.04
cache / IDefOpt / cuda / Forward 0.000002368 s 0.000002335 s 1.01
cache / JaXPipe / cuda / PreRev 0.000010431 s 0.00001088 s 0.96
cache / JaXPipe / cuda / PostRev 0.000010816 s 0.0000112 s 0.97
cache / JaXPipe / cuda / BothRev 0.000010848 s 0.0000112 s 0.97
cache / Jax / cuda / BothRev 0.000010656 s 0.000010784 s 0.99
cache / HLOOpt / cuda / PreRev 0.000013696 s 0.000013536 s 1.01
cache / HLOOpt / cuda / PostRev 0.000013664 s 0.000013505 s 1.01
cache / HLOOpt / cuda / BothRev 0.000013728 s 0.000013569 s 1.01
cache / PartOpt / cuda / PreRev 0.000010656 s 0.00001104 s 0.97
cache / PartOpt / cuda / PostRev 0.000011328 s 0.000010913 s 1.04
cache / PartOpt / cuda / BothRev 0.000011969 s 0.000010847 s 1.10
cache / IPartOpt / cuda / PreRev 0.000010624 s 0.000010976 s 0.97
cache / IPartOpt / cuda / PostRev 0.000010592 s 0.000011167 s 0.95
cache / IPartOpt / cuda / BothRev 0.000010592 s 0.000010657 s 0.99
cache / DefOpt / cuda / PreRev 0.000011136 s 0.00001104 s 1.01
cache / DefOpt / cuda / PostRev 0.000010624 s 0.000010753 s 0.99
cache / DefOpt / cuda / BothRev 0.000010752 s 0.000010815 s 0.99
cache / IDefOpt / cuda / PreRev 0.00001072 s 0.000010912 s 0.98
cache / IDefOpt / cuda / PostRev 0.000010848 s 0.000010656 s 1.02
cache / IDefOpt / cuda / BothRev 0.000010784 s 0.000010431 s 1.03
cache / JaXPipe / tpu / Primal 0.0000024664 s 0.000002477725 s 1.00
cache / Jax / tpu / Primal 0.000002463375 s 0.000002473575 s 1.00
cache / HLOOpt / tpu / Primal 0.000002466025 s 0.00000247285 s 1.00
cache / PartOpt / tpu / Primal 0.000002458425 s 0.0000024729 s 0.99
cache / IPartOpt / tpu / Primal 0.0000024613 s 0.0000024635500000000003 s 1.00
cache / DefOpt / tpu / Primal 0.0000024446 s 0.0000024609000000000004 s 0.99
cache / IDefOpt / tpu / Primal 0.000002462875 s 0.000002476775 s 0.99
cache / JaXPipe / tpu / Forward 0.0000035428 s 0.0000035618 s 0.99
cache / Jax / tpu / Forward 0.00000353855 s 0.00000355105 s 1.00
cache / HLOOpt / tpu / Forward 0.0000035641000000000003 s 0.000003544875 s 1.01
cache / PartOpt / tpu / Forward 0.000003562 s 0.000003535125 s 1.01
cache / IPartOpt / tpu / Forward 0.00000357245 s 0.000003558 s 1.00
cache / DefOpt / tpu / Forward 0.0000035285499999999994 s 0.0000035295500000000004 s 1.00
cache / IDefOpt / tpu / Forward 0.000003556675000000001 s 0.0000035562 s 1.00
cache / JaXPipe / tpu / PreRev 0.00000495595 s 0.000004968549999999999 s 1.00
cache / JaXPipe / tpu / PostRev 0.000004983274999999999 s 0.000004967025 s 1.00
cache / JaXPipe / tpu / BothRev 0.000004982825 s 0.00000499125 s 1.00
cache / Jax / tpu / BothRev 0.00000498185 s 0.000004999 s 1.00
cache / HLOOpt / tpu / PreRev 0.0000039337 s 0.0000039558750000000005 s 0.99
cache / HLOOpt / tpu / PostRev 0.00000410925 s 0.000004117675 s 1.00
cache / HLOOpt / tpu / BothRev 0.0000039428 s 0.0000039434 s 1.00
cache / PartOpt / tpu / PreRev 0.00000497965 s 0.000004983625 s 1.00
cache / PartOpt / tpu / PostRev 0.000004969125 s 0.000004951725 s 1.00
cache / PartOpt / tpu / BothRev 0.000004965975 s 0.000004973375 s 1.00
cache / IPartOpt / tpu / PreRev 0.0000049475000000000005 s 0.000004962074999999999 s 1.00
cache / IPartOpt / tpu / PostRev 0.0000049675 s 0.000004954775 s 1.00
cache / IPartOpt / tpu / BothRev 0.000004991225 s 0.00000498675 s 1.00
cache / DefOpt / tpu / PreRev 0.000004967925 s 0.000004984925 s 1.00
cache / DefOpt / tpu / PostRev 0.0000049811250000000005 s 0.000004961 s 1.00
cache / DefOpt / tpu / BothRev 0.000004954225 s 0.0000049668000000000005 s 1.00
cache / IDefOpt / tpu / PreRev 0.000004978125 s 0.00000496935 s 1.00
cache / IDefOpt / tpu / PostRev 0.0000049628 s 0.0000049644 s 1.00
cache / IDefOpt / tpu / BothRev 0.000004967650000000001 s 0.000004968675 s 1.00
cache / JaXPipe / cpu / Primal 0.00001298 s 0.000006617600029130699 s 1.96
cache / Jax / cpu / Primal 0.000012407 s 0.0000073614199845906115 s 1.69
cache / HLOOpt / cpu / Primal 0.000012689 s 0.000007327980010813917 s 1.73
cache / PartOpt / cpu / Primal 0.000012725 s 0.000006596160001208773 s 1.93
cache / IPartOpt / cpu / Primal 0.000012788 s 0.000006951079985810793 s 1.84
cache / DefOpt / cpu / Primal 0.000012636 s 0.000006810800023231423 s 1.86
cache / IDefOpt / cpu / Primal 0.000012753 s 0.000006719999992128578 s 1.90
cache / JaXPipe / cpu / Forward 0.000016933 s 0.00001635375998375821 s 1.04
cache / Jax / cpu / Forward 0.000017102 s 0.00001622899997528293 s 1.05
cache / HLOOpt / cpu / Forward 0.000017076 s 0.00001638741993701842 s 1.04
cache / PartOpt / cpu / Forward 0.000016788999999999998 s 0.00001637195999137475 s 1.03
cache / IPartOpt / cpu / Forward 0.000017152 s 0.000016631440003038735 s 1.03
cache / DefOpt / cpu / Forward 0.000017483 s 0.00001649958000598417 s 1.06
cache / IDefOpt / cpu / Forward 0.00001761 s 0.00001611459998457576 s 1.09
cache / JaXPipe / cpu / PreRev 0.000018164 s 0.000017117220031650505 s 1.06
cache / JaXPipe / cpu / PostRev 0.000019005 s 0.00002205438004239113 s 0.86
cache / JaXPipe / cpu / BothRev 0.000017859 s 0.00001823537998461688 s 0.98
cache / Jax / cpu / BothRev 0.000019958 s 0.00002179658003115037 s 0.92
cache / HLOOpt / cpu / PreRev 0.000018144 s 0.00001804116001949296 s 1.01
cache / HLOOpt / cpu / PostRev 0.000017881 s 0.000020775400025740963 s 0.86
cache / HLOOpt / cpu / BothRev 0.000017776 s 0.000018207139974038 s 0.98
cache / PartOpt / cpu / PreRev 0.000017618000000000003 s 0.000017113039994001157 s 1.03
cache / PartOpt / cpu / PostRev 0.000018767 s 0.000022788399955970818 s 0.82
cache / PartOpt / cpu / BothRev 0.000018042 s 0.000017054319996532286 s 1.06
cache / IPartOpt / cpu / PreRev 0.000017621000000000003 s 0.00001702454002952436 s 1.04
cache / IPartOpt / cpu / PostRev 0.000018843 s 0.00002129747997969389 s 0.88
cache / IPartOpt / cpu / BothRev 0.000017519 s 0.00001820888001930143 s 0.96
cache / DefOpt / cpu / PreRev 0.00002638 s 0.000016552119950574708 s 1.59
cache / DefOpt / cpu / PostRev 0.000027705 s 0.00001614170001630555 s 1.72
cache / DefOpt / cpu / BothRev 0.000028642 s 0.000016149760012922344 s 1.77
cache / IDefOpt / cpu / PreRev 0.000031254 s 0.000016202219967453857 s 1.93
cache / IDefOpt / cpu / PostRev 0.000026087 s 0.000015889159994912917 s 1.64
cache / IDefOpt / cpu / BothRev 0.000026241 s 0.000016051540005719288 s 1.63
cache / JaXPipe / cpu / Primal 0.000008999999999999999 s 0.000006617600029130699 s 1.36
cache / Jax / cpu / Primal 0.000008999999999999999 s 0.0000073614199845906115 s 1.22
cache / HLOOpt / cpu / Primal 0.000008999999999999999 s 0.000007327980010813917 s 1.23
cache / PartOpt / cpu / Primal 0.000008999999999999999 s 0.000006596160001208773 s 1.36
cache / IPartOpt / cpu / Primal 0.000008999999999999999 s 0.000006951079985810793 s 1.29
cache / DefOpt / cpu / Primal 0.000008999999999999999 s 0.000006810800023231423 s 1.32
cache / IDefOpt / cpu / Primal 0.000008999999999999999 s 0.000006719999992128578 s 1.34
cache / JaXPipe / cpu / Forward 0.000011 s 0.00001635375998375821 s 0.67
cache / Jax / cpu / Forward 0.000011 s 0.00001622899997528293 s 0.68
cache / HLOOpt / cpu / Forward 0.000011 s 0.00001638741993701842 s 0.67
cache / PartOpt / cpu / Forward 0.000011 s 0.00001637195999137475 s 0.67
cache / IPartOpt / cpu / Forward 0.000011 s 0.000016631440003038735 s 0.66
cache / DefOpt / cpu / Forward 0.000011 s 0.00001649958000598417 s 0.67
cache / IDefOpt / cpu / Forward 0.000011 s 0.00001611459998457576 s 0.68
cache / JaXPipe / cpu / PreRev 0.000011 s 0.000017117220031650505 s 0.64
cache / JaXPipe / cpu / PostRev 0.000011 s 0.00002205438004239113 s 0.50
cache / JaXPipe / cpu / BothRev 0.000014 s 0.00001823537998461688 s 0.77
cache / Jax / cpu / BothRev 0.000012 s 0.00002179658003115037 s 0.55
cache / HLOOpt / cpu / PreRev 0.000011 s 0.00001804116001949296 s 0.61
cache / HLOOpt / cpu / PostRev 0.000035999999999999994 s 0.000020775400025740963 s 1.73
cache / HLOOpt / cpu / BothRev 0.000011 s 0.000018207139974038 s 0.60
cache / PartOpt / cpu / PreRev 0.000011 s 0.000017113039994001157 s 0.64
cache / PartOpt / cpu / PostRev 0.000012 s 0.000022788399955970818 s 0.53
cache / PartOpt / cpu / BothRev 0.000011 s 0.000017054319996532286 s 0.64
cache / IPartOpt / cpu / PreRev 0.000011 s 0.00001702454002952436 s 0.65
cache / IPartOpt / cpu / PostRev 0.000012 s 0.00002129747997969389 s 0.56
cache / IPartOpt / cpu / BothRev 0.000012 s 0.00001820888001930143 s 0.66
cache / DefOpt / cpu / PreRev 0.000011 s 0.000016552119950574708 s 0.66
cache / DefOpt / cpu / PostRev 0.000011 s 0.00001614170001630555 s 0.68
cache / DefOpt / cpu / BothRev 0.000011 s 0.000016149760012922344 s 0.68
cache / IDefOpt / cpu / PreRev 0.000011 s 0.000016202219967453857 s 0.68
cache / IDefOpt / cpu / PostRev 0.000012 s 0.000015889159994912917 s 0.76
cache / IDefOpt / cpu / BothRev 0.000011 s 0.000016051540005719288 s 0.69
Concat / JaXPipe / cpu / Primal 0.000006987039998875844 s 0.000007270800006153877 s 0.96
Concat / Jax / cpu / Primal 0.000007006739995176759 s 0.000006953919983061496 s 1.01
Concat / HLOOpt / cpu / Primal 0.000006464879993473005 s 0.000007605120017615263 s 0.85
Concat / PartOpt / cpu / Primal 0.000006732580006882927 s 0.000006844539975645603 s 0.98
Concat / IPartOpt / cpu / Primal 0.000007292820012025913 s 0.000007298220016309642 s 1.00
Concat / DefOpt / cpu / Primal 0.000006477479994373425 s 0.000007233900023493334 s 0.90
Concat / IDefOpt / cpu / Primal 0.000006781359993510705 s 0.000007258480000018608 s 0.93
Concat / JaXPipe / cpu / Forward 0.000009602799991625943 s 0.000011428860007072216 s 0.84
Concat / Jax / cpu / Forward 0.000009775799999260924 s 0.00001162386000942206 s 0.84
Concat / HLOOpt / cpu / Forward 0.000009691600009773538 s 0.00001077782001630112 s 0.90
Concat / PartOpt / cpu / Forward 0.000009572519995799669 s 0.000011412619987822835 s 0.84
Concat / IPartOpt / cpu / Forward 0.00001016551999555304 s 0.000011144519994559232 s 0.91
Concat / DefOpt / cpu / Forward 0.000010006459999658546 s 0.000011024940040442745 s 0.91
Concat / IDefOpt / cpu / Forward 0.000009944120004092838 s 0.000011167399989062688 s 0.89
Concat / JaXPipe / cpu / PreRev 0.000011728599995421971 s 0.000012688919996435288 s 0.92
Concat / JaXPipe / cpu / PostRev 0.00001073239999186626 s 0.00001220757997543842 s 0.88
Concat / JaXPipe / cpu / BothRev 0.000011374120001619303 s 0.00001241400000253634 s 0.92
Concat / Jax / cpu / BothRev 0.000010680420002699977 s 0.000012415119981596944 s 0.86
Concat / HLOOpt / cpu / PreRev 0.00001205781999942701 s 0.00001305383997532772 s 0.92
Concat / HLOOpt / cpu / PostRev 0.000013610699998025666 s 0.000014871699977447862 s 0.92
Concat / HLOOpt / cpu / BothRev 0.00001129023999965284 s 0.000012056320047122426 s 0.94
Concat / PartOpt / cpu / PreRev 0.000011274859998593456 s 0.00001218095996591728 s 0.93
Concat / PartOpt / cpu / PostRev 0.00001164438000841983 s 0.00001255646002391586 s 0.93
Concat / PartOpt / cpu / BothRev 0.000012165940004251752 s 0.000012368560010145304 s 0.98
Concat / IPartOpt / cpu / PreRev 0.000011056939995341964 s 0.000013087139996059704 s 0.84
Concat / IPartOpt / cpu / PostRev 0.000011073659995872731 s 0.00001286647993765655 s 0.86
Concat / IPartOpt / cpu / BothRev 0.000010850519990981412 s 0.000012555660050566076 s 0.86
Concat / DefOpt / cpu / PreRev 0.000011033779994704671 s 0.000012576600011016123 s 0.88
Concat / DefOpt / cpu / PostRev 0.000011207579996153072 s 0.000012553639990073862 s 0.89
Concat / DefOpt / cpu / BothRev 0.000011315260001083517 s 0.00001252969997949549 s 0.90
Concat / IDefOpt / cpu / PreRev 0.000011311779983316229 s 0.000012443119985618976 s 0.91
Concat / IDefOpt / cpu / PostRev 0.00001131653999664195 s 0.0000129725800070446 s 0.87
Concat / IDefOpt / cpu / BothRev 0.000011632940002073157 s 0.00001261953995708609 s 0.92
Concat / JaXPipe / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
Concat / Jax / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
Concat / HLOOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
Concat / PartOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
Concat / IPartOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
Concat / DefOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
Concat / IDefOpt / cuda / Primal 0.0000024 s 0.0000019200000000000003 s 1.25
Concat / JaXPipe / cuda / Forward 0.000010751 s 0.00000976 s 1.10
Concat / Jax / cuda / Forward 0.000010496 s 0.000010112 s 1.04
Concat / HLOOpt / cuda / Forward 0.00001056 s 0.000010368 s 1.02
Concat / PartOpt / cuda / Forward 0.000010528 s 0.000010176 s 1.03
Concat / IPartOpt / cuda / Forward 0.000010656 s 0.000010304 s 1.03
Concat / DefOpt / cuda / Forward 0.000013535 s 0.000009952 s 1.36
Concat / IDefOpt / cuda / Forward 0.000010496 s 0.000010432 s 1.01
Concat / JaXPipe / cuda / PreRev 0.000016832 s 0.00001584 s 1.06
Concat / JaXPipe / cuda / PostRev 0.000017247999999999998 s 0.000016 s 1.08
Concat / JaXPipe / cuda / BothRev 0.000016768000000000003 s 0.000016512 s 1.02
Concat / Jax / cuda / BothRev 0.000016927999999999998 s 0.00001616 s 1.05
Concat / HLOOpt / cuda / PreRev 0.000017216 s 0.000015808 s 1.09
Concat / HLOOpt / cuda / PostRev 0.000016544 s 0.000015776 s 1.05
Concat / HLOOpt / cuda / BothRev 0.00001664 s 0.000016225 s 1.03
Concat / PartOpt / cuda / PreRev 0.000016704 s 0.000016352 s 1.02
Concat / PartOpt / cuda / PostRev 0.000016385 s 0.000016 s 1.02
Concat / PartOpt / cuda / BothRev 0.000016704 s 0.000016096 s 1.04
Concat / IPartOpt / cuda / PreRev 0.00001664 s 0.000015968 s 1.04
Concat / IPartOpt / cuda / PostRev 0.00001696 s 0.000015935999999999998 s 1.06
Concat / IPartOpt / cuda / BothRev 0.000016576000000000002 s 0.000016128 s 1.03
Concat / DefOpt / cuda / PreRev 0.000017183 s 0.000016096 s 1.07
Concat / DefOpt / cuda / PostRev 0.000016832 s 0.000016224 s 1.04
Concat / DefOpt / cuda / BothRev 0.000016929 s 0.000015809 s 1.07
Concat / IDefOpt / cuda / PreRev 0.000016958999999999998 s 0.000016224 s 1.05
Concat / IDefOpt / cuda / PostRev 0.000016704 s 0.000016448999999999998 s 1.02
Concat / IDefOpt / cuda / BothRev 0.000016958999999999998 s 0.000016383999999999998 s 1.04
Concat / JaXPipe / tpu / Primal 0.000001526775 s 0.00000152745 s 1.00
Concat / Jax / tpu / Primal 0.000001528225 s 0.0000015306 s 1.00
Concat / HLOOpt / tpu / Primal 0.00000152115 s 0.00000153055 s 0.99
Concat / PartOpt / tpu / Primal 0.00000152535 s 0.00000153385 s 0.99
Concat / IPartOpt / tpu / Primal 0.00000152805 s 0.0000015226500000000002 s 1.00
Concat / DefOpt / tpu / Primal 0.000001533825 s 0.00000152385 s 1.01
Concat / IDefOpt / tpu / Primal 0.0000015229500000000002 s 0.000001522975 s 1.00
Concat / JaXPipe / tpu / Forward 0.00000156835 s 0.000001585125 s 0.99
Concat / Jax / tpu / Forward 0.00000155665 s 0.00000155495 s 1.00
Concat / HLOOpt / tpu / Forward 0.00000156535 s 0.000001570125 s 1.00
Concat / PartOpt / tpu / Forward 0.000001562325 s 0.000001559625 s 1.00
Concat / IPartOpt / tpu / Forward 0.000001574875 s 0.0000015803249999999998 s 1.00
Concat / DefOpt / tpu / Forward 0.00000155745 s 0.00000156675 s 0.99
Concat / IDefOpt / tpu / Forward 0.000001570475 s 0.0000015851499999999998 s 0.99
Concat / JaXPipe / tpu / PreRev 0.00000201075 s 0.0000019909750000000004 s 1.01
Concat / JaXPipe / tpu / PostRev 0.00000208895 s 0.0000020868 s 1.00
Concat / JaXPipe / tpu / BothRev 0.000001992025 s 0.0000019944 s 1.00
Concat / Jax / tpu / BothRev 0.0000020739 s 0.000002074825 s 1.00
Concat / HLOOpt / tpu / PreRev 0.00000199085 s 0.00000199615 s 1.00
Concat / HLOOpt / tpu / PostRev 0.000002076275 s 0.0000020738000000000004 s 1.00
Concat / HLOOpt / tpu / BothRev 0.0000019955 s 0.000002012075 s 0.99
Concat / PartOpt / tpu / PreRev 0.000002068575 s 0.000002072 s 1.00
Concat / PartOpt / tpu / PostRev 0.0000019977 s 0.00000199355 s 1.00
Concat / PartOpt / tpu / BothRev 0.000002068425 s 0.0000020755000000000003 s 1.00
Concat / IPartOpt / tpu / PreRev 0.000001997325 s 0.000002003575 s 1.00
Concat / IPartOpt / tpu / PostRev 0.000002078 s 0.000002073975 s 1.00
Concat / IPartOpt / tpu / BothRev 0.0000019906 s 0.0000020022 s 0.99
Concat / DefOpt / tpu / PreRev 0.000002083675 s 0.00000207405 s 1.00
Concat / DefOpt / tpu / PostRev 0.0000019956 s 0.0000019901250000000003 s 1.00
Concat / DefOpt / tpu / BothRev 0.000002068 s 0.00000206865 s 1.00
Concat / IDefOpt / tpu / PreRev 0.0000019954250000000005 s 0.00000199775 s 1.00
Concat / IDefOpt / tpu / PostRev 0.00000207895 s 0.0000020728 s 1.00
Concat / IDefOpt / tpu / BothRev 0.000001995875 s 0.0000019916 s 1.00
Concat / JaXPipe / cpu / Primal 0.000012735 s 0.000007270800006153877 s 1.75
Concat / Jax / cpu / Primal 0.000012929 s 0.000006953919983061496 s 1.86
Concat / HLOOpt / cpu / Primal 0.000012748 s 0.000007605120017615263 s 1.68
Concat / PartOpt / cpu / Primal 0.000012799 s 0.000006844539975645603 s 1.87
Concat / IPartOpt / cpu / Primal 0.00001292 s 0.000007298220016309642 s 1.77
Concat / DefOpt / cpu / Primal 0.000012364 s 0.000007233900023493334 s 1.71
Concat / IDefOpt / cpu / Primal 0.000012949 s 0.000007258480000018608 s 1.78
Concat / JaXPipe / cpu / Forward 0.000017953 s 0.000011428860007072216 s 1.57
Concat / Jax / cpu / Forward 0.00001749 s 0.00001162386000942206 s 1.50
Concat / HLOOpt / cpu / Forward 0.000017533 s 0.00001077782001630112 s 1.63
Concat / PartOpt / cpu / Forward 0.000017044 s 0.000011412619987822835 s 1.49
Concat / IPartOpt / cpu / Forward 0.000017614 s 0.000011144519994559232 s 1.58
Concat / DefOpt / cpu / Forward 0.000017233999999999998 s 0.000011024940040442745 s 1.56
Concat / IDefOpt / cpu / Forward 0.00001734 s 0.000011167399989062688 s 1.55
Concat / JaXPipe / cpu / PreRev 0.000020215 s 0.000012688919996435288 s 1.59
Concat / JaXPipe / cpu / PostRev 0.000019069 s 0.00001220757997543842 s 1.56
Concat / JaXPipe / cpu / BothRev 0.000019311 s 0.00001241400000253634 s 1.56
Concat / Jax / cpu / BothRev 0.000019865 s 0.000012415119981596944 s 1.60
Concat / HLOOpt / cpu / PreRev 0.000019862 s 0.00001305383997532772 s 1.52
Concat / HLOOpt / cpu / PostRev 0.000019124 s 0.000014871699977447862 s 1.29
Concat / HLOOpt / cpu / BothRev 0.000019164 s 0.000012056320047122426 s 1.59
Concat / PartOpt / cpu / PreRev 0.000020517 s 0.00001218095996591728 s 1.68
Concat / PartOpt / cpu / PostRev 0.000019228 s 0.00001255646002391586 s 1.53
Concat / PartOpt / cpu / BothRev 0.0000194 s 0.000012368560010145304 s 1.57
Concat / IPartOpt / cpu / PreRev 0.00001927 s 0.000013087139996059704 s 1.47
Concat / IPartOpt / cpu / PostRev 0.000019469 s 0.00001286647993765655 s 1.51
Concat / IPartOpt / cpu / BothRev 0.000019469 s 0.000012555660050566076 s 1.55
Concat / DefOpt / cpu / PreRev 0.000020246 s 0.000012576600011016123 s 1.61
Concat / DefOpt / cpu / PostRev 0.000019648 s 0.000012553639990073862 s 1.57
Concat / DefOpt / cpu / BothRev 0.000019281 s 0.00001252969997949549 s 1.54
Concat / IDefOpt / cpu / PreRev 0.000019293 s 0.000012443119985618976 s 1.55
Concat / IDefOpt / cpu / PostRev 0.000019346 s 0.0000129725800070446 s 1.49
Concat / IDefOpt / cpu / BothRev 0.000019655 s 0.00001261953995708609 s 1.56
Concat / JaXPipe / cpu / Primal 0.000008999999999999999 s 0.000007270800006153877 s 1.24
Concat / Jax / cpu / Primal 0.000008999999999999999 s 0.000006953919983061496 s 1.29
Concat / HLOOpt / cpu / Primal 0.000008999999999999999 s 0.000007605120017615263 s 1.18
Concat / PartOpt / cpu / Primal 0.000008999999999999999 s 0.000006844539975645603 s 1.31
Concat / IPartOpt / cpu / Primal 0.000008999999999999999 s 0.000007298220016309642 s 1.23
Concat / DefOpt / cpu / Primal 0.000008999999999999999 s 0.000007233900023493334 s 1.24
Concat / IDefOpt / cpu / Primal 0.000008999999999999999 s 0.000007258480000018608 s 1.24
Concat / JaXPipe / cpu / Forward 0.000012 s 0.000011428860007072216 s 1.05
Concat / Jax / cpu / Forward 0.000012 s 0.00001162386000942206 s 1.03
Concat / HLOOpt / cpu / Forward 0.000012 s 0.00001077782001630112 s 1.11
Concat / PartOpt / cpu / Forward 0.000012 s 0.000011412619987822835 s 1.05
Concat / IPartOpt / cpu / Forward 0.000012 s 0.000011144519994559232 s 1.08
Concat / DefOpt / cpu / Forward 0.000012 s 0.000011024940040442745 s 1.09
Concat / IDefOpt / cpu / Forward 0.000013 s 0.000011167399989062688 s 1.16
Concat / JaXPipe / cpu / PreRev 0.000014 s 0.000012688919996435288 s 1.10
Concat / JaXPipe / cpu / PostRev 0.000015 s 0.00001220757997543842 s 1.23
Concat / JaXPipe / cpu / BothRev 0.000014 s 0.00001241400000253634 s 1.13
Concat / Jax / cpu / BothRev 0.000015 s 0.000012415119981596944 s 1.21
Concat / HLOOpt / cpu / PreRev 0.000015 s 0.00001305383997532772 s 1.15
Concat / HLOOpt / cpu / PostRev 0.000015 s 0.000014871699977447862 s 1.01
Concat / HLOOpt / cpu / BothRev 0.000015 s 0.000012056320047122426 s 1.24
Concat / PartOpt / cpu / PreRev 0.000014 s 0.00001218095996591728 s 1.15
Concat / PartOpt / cpu / PostRev 0.000015 s 0.00001255646002391586 s 1.19
Concat / PartOpt / cpu / BothRev 0.000014 s 0.000012368560010145304 s 1.13
Concat / IPartOpt / cpu / PreRev 0.000014 s 0.000013087139996059704 s 1.07
Concat / IPartOpt / cpu / PostRev 0.000015 s 0.00001286647993765655 s 1.17
Concat / IPartOpt / cpu / BothRev 0.000014 s 0.000012555660050566076 s 1.12
Concat / DefOpt / cpu / PreRev 0.000014 s 0.000012576600011016123 s 1.11
Concat / DefOpt / cpu / PostRev 0.000014 s 0.000012553639990073862 s 1.12
Concat / DefOpt / cpu / BothRev 0.000015 s 0.00001252969997949549 s 1.20
Concat / IDefOpt / cpu / PreRev 0.000015 s 0.000012443119985618976 s 1.21
Concat / IDefOpt / cpu / PostRev 0.000014 s 0.0000129725800070446 s 1.08
Concat / IDefOpt / cpu / BothRev 0.000014 s 0.00001261953995708609 s 1.11
const_scatter / JaXPipe / cpu / Primal 0.000006108559998665441 s 0.0000066701799732982185 s 0.92
const_scatter / Jax / cpu / Primal 0.000006209959999523562 s 0.000006743799976902665 s 0.92
const_scatter / HLOOpt / cpu / Primal 0.000007044020003377228 s 0.000007621859967912315 s 0.92
const_scatter / PartOpt / cpu / Primal 0.000005980339994948736 s 0.000006892500023241155 s 0.87
const_scatter / IPartOpt / cpu / Primal 0.0000066518399898996 s 0.000006738659985785489 s 0.99
const_scatter / DefOpt / cpu / Primal 0.000006993760002842464 s 0.0000071026000114216 s 0.98
const_scatter / IDefOpt / cpu / Primal 0.000007023020002634439 s 0.000007858320032028132 s 0.89
const_scatter / JaXPipe / cpu / Forward 0.000010509120004371653 s 0.000011870759990415536 s 0.89
const_scatter / Jax / cpu / Forward 0.000009151179992841208 s 0.000010619959966788884 s 0.86
const_scatter / HLOOpt / cpu / Forward 0.000010640880004757491 s 0.000011783159970946145 s 0.90
const_scatter / PartOpt / cpu / Forward 0.000010439600007430272 s 0.000011229000037928926 s 0.93
const_scatter / IPartOpt / cpu / Forward 0.000010492479996173642 s 0.000012111659980291734 s 0.87
const_scatter / DefOpt / cpu / Forward 0.00001019578000978072 s 0.000011104419972980396 s 0.92
const_scatter / IDefOpt / cpu / Forward 0.000010187319996930455 s 0.000012228739997226512 s 0.83
const_scatter / JaXPipe / cpu / PreRev 0.0002858140600096 s 0.0002900659799706 s 0.99
const_scatter / JaXPipe / cpu / PostRev 0.0002793739999947 s 0.0002828032600064 s 0.99
const_scatter / JaXPipe / cpu / BothRev 0.0002810276400077 s 0.0002831334199981 s 0.99
const_scatter / Jax / cpu / BothRev 0.000281282700007 s 0.0002821311999741 s 1.00
const_scatter / HLOOpt / cpu / PreRev 0.0002831070800039 s 0.0002847428000131 s 0.99
const_scatter / HLOOpt / cpu / PostRev 0.0002855815800012 s 0.0002864551800121 s 1.00
const_scatter / HLOOpt / cpu / BothRev 0.000281064319995 s 0.0002834857999732 s 0.99
const_scatter / PartOpt / cpu / PreRev 0.0002847579600143 s 0.000282958500029 s 1.01
const_scatter / PartOpt / cpu / PostRev 0.0002808074999916 s 0.0002846486800353 s 0.99
const_scatter / PartOpt / cpu / BothRev 0.0002822812799877 s 0.0002827322999928 s 1.00
const_scatter / IPartOpt / cpu / PreRev 0.0002822949599976 s 0.0002970088200254 s 0.95
const_scatter / IPartOpt / cpu / PostRev 0.0002825830600113 s 0.0002822358000321 s 1.00
const_scatter / IPartOpt / cpu / BothRev 0.0002826459200036 s 0.0002828837800097 s 1.00
const_scatter / DefOpt / cpu / PreRev 0.0002853437799944 s 0.0002843984800074 s 1.00
const_scatter / DefOpt / cpu / PostRev 0.0002807084799928 s 0.0002837708200422 s 0.99
const_scatter / DefOpt / cpu / BothRev 0.000286460879995 s 0.0002855213999737 s 1.00
const_scatter / IDefOpt / cpu / PreRev 0.0002846789999898 s 0.0003056161200311 s 0.93
const_scatter / IDefOpt / cpu / PostRev 0.0002851838000015 s 0.0002853217799838 s 1.00
const_scatter / IDefOpt / cpu / BothRev 0.0002815989600026 s 0.0002853424200202 s 0.99
const_scatter / JaXPipe / cuda / Primal 0.0000024 s 0.000001887 s 1.27
const_scatter / Jax / cuda / Primal 0.0000024 s 0.000001887 s 1.27
const_scatter / HLOOpt / cuda / Primal 0.0000024 s 0.000001887 s 1.27
const_scatter / PartOpt / cuda / Primal 0.0000024 s 0.000001887 s 1.27
const_scatter / IPartOpt / cuda / Primal 0.0000024 s 0.000001887 s 1.27
const_scatter / DefOpt / cuda / Primal 0.0000024 s 0.000001887 s 1.27
const_scatter / IDefOpt / cuda / Primal 0.0000024 s 0.000001887 s 1.27
const_scatter / JaXPipe / cuda / Forward 0.000010752 s 0.00000992 s 1.08
const_scatter / Jax / cuda / Forward 0.000010688 s 0.0000096 s 1.11
const_scatter / HLOOpt / cuda / Forward 0.00001056 s 0.00001008 s 1.05
const_scatter / PartOpt / cuda / Forward 0.000010528 s 0.000009857 s 1.07
const_scatter / IPartOpt / cuda / Forward 0.000010592 s 0.000010048 s 1.05
const_scatter / DefOpt / cuda / Forward 0.000010752 s 0.00001008 s 1.07
const_scatter / IDefOpt / cuda / Forward 0.0000104 s 0.000009856 s 1.06
const_scatter / JaXPipe / cuda / PreRev 0.000016992 s 0.000015968 s 1.06
const_scatter / JaXPipe / cuda / PostRev 0.000017024 s 0.000016063000000000002 s 1.06
const_scatter / JaXPipe / cuda / BothRev 0.000016864 s 0.000017984 s 0.94
const_scatter / Jax / cuda / BothRev 0.000017216 s 0.0000184 s 0.94
const_scatter / HLOOpt / cuda / PreRev 0.000016864 s 0.000016096 s 1.05
const_scatter / HLOOpt / cuda / PostRev 0.000016768000000000003 s 0.000015744 s 1.07
const_scatter / HLOOpt / cuda / BothRev 0.000016767 s 0.000016255999999999998 s 1.03
const_scatter / PartOpt / cuda / PreRev 0.000016767 s 0.000017857000000000002 s 0.94
const_scatter / PartOpt / cuda / PostRev 0.000016863 s 0.000016417000000000002 s 1.03
const_scatter / PartOpt / cuda / BothRev 0.000016767 s 0.000016416 s 1.02
const_scatter / IPartOpt / cuda / PreRev 0.000016607 s 0.00001616 s 1.03
const_scatter / IPartOpt / cuda / PostRev 0.000016799000000000003 s 0.000016544 s 1.02
const_scatter / IPartOpt / cuda / BothRev 0.000016768999999999998 s 0.000015904000000000002 s 1.05
const_scatter / DefOpt / cuda / PreRev 0.000017121 s 0.000016352 s 1.05
const_scatter / DefOpt / cuda / PostRev 0.000016896000000000002 s 0.000015935999999999998 s 1.06
const_scatter / DefOpt / cuda / BothRev 0.000016832 s 0.00001584 s 1.06
const_scatter / IDefOpt / cuda / PreRev 0.00001664 s 0.000021888 s 0.76
const_scatter / IDefOpt / cuda / PostRev 0.000017024 s 0.000015935999999999998 s 1.07
const_scatter / IDefOpt / cuda / BothRev 0.000016512 s 0.000016255999999999998 s 1.02
const_scatter / JaXPipe / tpu / Primal 0.000003805725 s 0.000003818225 s 1.00
const_scatter / Jax / tpu / Primal 0.00000384415 s 0.000003825575 s 1.00
const_scatter / HLOOpt / tpu / Primal 0.000003824675 s 0.000003793475 s 1.01
const_scatter / PartOpt / tpu / Primal 0.00000385145 s 0.000003842975 s 1.00
const_scatter / IPartOpt / tpu / Primal 0.00000381395 s 0.000003810875 s 1.00
const_scatter / DefOpt / tpu / Primal 0.000003824275 s 0.000003816975 s 1.00
const_scatter / IDefOpt / tpu / Primal 0.0000038132 s 0.000003805875 s 1.00
const_scatter / JaXPipe / tpu / Forward 0.00000647835 s 0.00000650045 s 1.00
const_scatter / Jax / tpu / Forward 0.000006475675 s 0.000006493925 s 1.00
const_scatter / HLOOpt / tpu / Forward 0.000006498150000000001 s 0.0000064629250000000005 s 1.01
const_scatter / PartOpt / tpu / Forward 0.0000064901 s 0.00000649415 s 1.00
const_scatter / IPartOpt / tpu / Forward 0.0000064710500000000005 s 0.000006451675 s 1.00
const_scatter / DefOpt / tpu / Forward 0.0000064881 s 0.00000650245 s 1.00
const_scatter / IDefOpt / tpu / Forward 0.000006466275 s 0.000006453625 s 1.00
const_scatter / JaXPipe / tpu / PreRev 0.000006669225 s 0.00000668875 s 1.00
const_scatter / JaXPipe / tpu / PostRev 0.000006671275 s 0.000006669874999999999 s 1.00
const_scatter / JaXPipe / tpu / BothRev 0.0000066567 s 0.000006674075000000001 s 1.00
const_scatter / Jax / tpu / BothRev 0.000006656525 s 0.00000666065 s 1.00
const_scatter / HLOOpt / tpu / PreRev 0.000006643825 s 0.000006668475 s 1.00
const_scatter / HLOOpt / tpu / PostRev 0.000006664125 s 0.000006640350000000001 s 1.00
const_scatter / HLOOpt / tpu / BothRev 0.0000066657750000000006 s 0.000006670825 s 1.00
const_scatter / PartOpt / tpu / PreRev 0.000006685175 s 0.000006679725 s 1.00
const_scatter / PartOpt / tpu / PostRev 0.000006661425 s 0.0000066626 s 1.00
const_scatter / PartOpt / tpu / BothRev 0.000006656875 s 0.00000665535 s 1.00
const_scatter / IPartOpt / tpu / PreRev 0.000006643550000000001 s 0.000006675325 s 1.00
const_scatter / IPartOpt / tpu / PostRev 0.00000666745 s 0.0000066698 s 1.00
const_scatter / IPartOpt / tpu / BothRev 0.000006655375 s 0.00000667635 s 1.00
const_scatter / DefOpt / tpu / PreRev 0.000006649525 s 0.0000066897 s 0.99
const_scatter / DefOpt / tpu / PostRev 0.0000066838 s 0.000006663175 s 1.00
const_scatter / DefOpt / tpu / BothRev 0.000006678575 s 0.00000667775 s 1.00
const_scatter / IDefOpt / tpu / PreRev 0.000006649175000000001 s 0.000006667025 s 1.00
const_scatter / IDefOpt / tpu / PostRev 0.000006680325 s 0.000006663125 s 1.00
const_scatter / IDefOpt / tpu / BothRev 0.000006692725 s 0.000006661175 s 1.00
const_scatter / JaXPipe / cpu / Primal 0.00001297 s 0.0000066701799732982185 s 1.94
const_scatter / Jax / cpu / Primal 0.000012837 s 0.000006743799976902665 s 1.90
const_scatter / HLOOpt / cpu / Primal 0.00001306 s 0.000007621859967912315 s 1.71
const_scatter / PartOpt / cpu / Primal 0.000012284 s 0.000006892500023241155 s 1.78
const_scatter / IPartOpt / cpu / Primal 0.000012362 s 0.000006738659985785489 s 1.83
const_scatter / DefOpt / cpu / Primal 0.000013347 s 0.0000071026000114216 s 1.88
const_scatter / IDefOpt / cpu / Primal 0.000013323 s 0.000007858320032028132 s 1.70
const_scatter / JaXPipe / cpu / Forward 0.000017816 s 0.000011870759990415536 s 1.50
const_scatter / Jax / cpu / Forward 0.000016363 s 0.000010619959966788884 s 1.54
const_scatter / HLOOpt / cpu / Forward 0.000017568000000000002 s 0.000011783159970946145 s 1.49
const_scatter / PartOpt / cpu / Forward 0.000017684 s 0.000011229000037928926 s 1.57
const_scatter / IPartOpt / cpu / Forward 0.000017641 s 0.000012111659980291734 s 1.46
const_scatter / DefOpt / cpu / Forward 0.000017715999999999998 s 0.000011104419972980396 s 1.60
const_scatter / IDefOpt / cpu / Forward 0.000018399 s 0.000012228739997226512 s 1.50
const_scatter / JaXPipe / cpu / PreRev 0.000520072 s 0.0002900659799706 s 1.79
const_scatter / JaXPipe / cpu / PostRev 0.000524602 s 0.0002828032600064 s 1.86
const_scatter / JaXPipe / cpu / BothRev 0.0005186979999999 s 0.0002831334199981 s 1.83
const_scatter / Jax / cpu / BothRev 0.000509981 s 0.0002821311999741 s 1.81
const_scatter / HLOOpt / cpu / PreRev 0.000513085 s 0.0002847428000131 s 1.80
const_scatter / HLOOpt / cpu / PostRev 0.000511103 s 0.0002864551800121 s 1.78
const_scatter / HLOOpt / cpu / BothRev 0.000524223 s 0.0002834857999732 s 1.85
const_scatter / PartOpt / cpu / PreRev 0.0005257619999999 s 0.000282958500029 s 1.86
const_scatter / PartOpt / cpu / PostRev 0.000523146 s 0.0002846486800353 s 1.84
const_scatter / PartOpt / cpu / BothRev 0.000510351 s 0.0002827322999928 s 1.81
const_scatter / IPartOpt / cpu / PreRev 0.000526735 s 0.0002970088200254 s 1.77
const_scatter / IPartOpt / cpu / PostRev 0.0005301829999999 s 0.0002822358000321 s 1.88
const_scatter / IPartOpt / cpu / BothRev 0.000522049 s 0.0002828837800097 s 1.85
const_scatter / DefOpt / cpu / PreRev 0.000521124 s 0.0002843984800074 s 1.83
const_scatter / DefOpt / cpu / PostRev 0.000504847 s 0.0002837708200422 s 1.78
const_scatter / DefOpt / cpu / BothRev 0.000519664 s 0.0002855213999737 s 1.82
const_scatter / IDefOpt / cpu / PreRev 0.000533165 s 0.0003056161200311 s 1.74
const_scatter / IDefOpt / cpu / PostRev 0.0005317049999999 s 0.0002853217799838 s 1.86
const_scatter / IDefOpt / cpu / BothRev 0.000541714 s 0.0002853424200202 s 1.90
const_scatter / JaXPipe / cpu / Primal 0.000008999999999999999 s 0.0000066701799732982185 s 1.35
const_scatter / Jax / cpu / Primal 0.000008999999999999999 s 0.000006743799976902665 s 1.33
const_scatter / HLOOpt / cpu / Primal 0.000008999999999999999 s 0.000007621859967912315 s 1.18
const_scatter / PartOpt / cpu / Primal 0.000008 s 0.000006892500023241155 s 1.16
const_scatter / IPartOpt / cpu / Primal 0.000008999999999999999 s 0.000006738659985785489 s 1.34
const_scatter / DefOpt / cpu / Primal 0.000008999999999999999 s 0.0000071026000114216 s 1.27
const_scatter / IDefOpt / cpu / Primal 0.000008999999999999999 s 0.000007858320032028132 s 1.15
const_scatter / JaXPipe / cpu / Forward 0.000013 s 0.000011870759990415536 s 1.10
const_scatter / Jax / cpu / Forward 0.000012 s 0.000010619959966788884 s 1.13
const_scatter / HLOOpt / cpu / Forward 0.000013 s 0.000011783159970946145 s 1.10
const_scatter / PartOpt / cpu / Forward 0.000013 s 0.000011229000037928926 s 1.16
const_scatter / IPartOpt / cpu / Forward 0.000013 s 0.000012111659980291734 s 1.07
const_scatter / DefOpt / cpu / Forward 0.000013 s 0.000011104419972980396 s 1.17
const_scatter / IDefOpt / cpu / Forward 0.000013 s 0.000012228739997226512 s 1.06
const_scatter / JaXPipe / cpu / PreRev 0.000336 s 0.0002900659799706 s 1.16
const_scatter / JaXPipe / cpu / PostRev 0.000348 s 0.0002828032600064 s 1.23
const_scatter / JaXPipe / cpu / BothRev 0.000327 s 0.0002831334199981 s 1.15
const_scatter / Jax / cpu / BothRev 0.000368 s 0.0002821311999741 s 1.30
const_scatter / HLOOpt / cpu / PreRev 0.000399 s 0.0002847428000131 s 1.40
const_scatter / HLOOpt / cpu / PostRev 0.0003549999999999 s 0.0002864551800121 s 1.24
const_scatter / HLOOpt / cpu / BothRev 0.000343 s 0.0002834857999732 s 1.21
const_scatter / PartOpt / cpu / PreRev 0.000384 s 0.000282958500029 s 1.36
const_scatter / PartOpt / cpu / PostRev 0.00035 s 0.0002846486800353 s 1.23
const_scatter / PartOpt / cpu / BothRev 0.00057 s 0.0002827322999928 s 2.02
const_scatter / IPartOpt / cpu / PreRev 0.000333 s 0.0002970088200254 s 1.12
const_scatter / IPartOpt / cpu / PostRev 0.000372 s 0.0002822358000321 s 1.32
const_scatter / IPartOpt / cpu / BothRev 0.000326 s 0.0002828837800097 s 1.15
const_scatter / DefOpt / cpu / PreRev 0.000388 s 0.0002843984800074 s 1.36
const_scatter / DefOpt / cpu / PostRev 0.000363 s 0.0002837708200422 s 1.28
const_scatter / DefOpt / cpu / BothRev 0.000336 s 0.0002855213999737 s 1.18
const_scatter / IDefOpt / cpu / PreRev 0.0003689999999999 s 0.0003056161200311 s 1.21
const_scatter / IDefOpt / cpu / PostRev 0.000363 s 0.0002853217799838 s 1.27
const_scatter / IDefOpt / cpu / BothRev 0.000357 s 0.0002853424200202 s 1.25
GenDot / JaXPipe / cpu / Primal 0.000007164820005982619 s 0.000008296160003737896 s 0.86
GenDot / Jax / cpu / Primal 0.000006894260000080976 s 0.000008259619971795473 s 0.83
GenDot / HLOOpt / cpu / Primal 0.000007182400011060963 s 0.000008371400035684928 s 0.86
GenDot / PartOpt / cpu / Primal 0.000006891279992942145 s 0.00000901005999367044 s 0.76
GenDot / IPartOpt / cpu / Primal 0.000007341340001403296 s 0.000008573640043323394 s 0.86
GenDot / DefOpt / cpu / Primal 0.000007125279992123978 s 0.000007987099979800406 s 0.89
GenDot / IDefOpt / cpu / Primal 0.000006793240013394097 s 0.000007670840022910853 s 0.89
GenDot / JaXPipe / cpu / Forward 0.000010680080006295611 s 0.000011564219994397715 s 0.92
GenDot / Jax / cpu / Forward 0.00001023135998593716 s 0.000011687760006680036 s 0.88
GenDot / HLOOpt / cpu / Forward 0.000010816820004038164 s 0.000012549280017992717 s 0.86
GenDot / PartOpt / cpu / Forward 0.00001070916000571742 s 0.000011167880002176387 s 0.96
GenDot / IPartOpt / cpu / Forward 0.000011368420014150616 s 0.000012273299962544117 s 0.93
GenDot / DefOpt / cpu / Forward 0.000010471499995219347 s 0.000011512759956531226 s 0.91
GenDot / IDefOpt / cpu / Forward 0.00001057081999078946 s 0.000012003320016447103 s 0.88
GenDot / JaXPipe / cpu / PreRev 0.00001126492000594226 s 0.000012277580008230873 s 0.92
GenDot / JaXPipe / cpu / PostRev 0.000010336860002553294 s 0.000011590839994823907 s 0.89
GenDot / JaXPipe / cpu / BothRev 0.00001152759998831243 s 0.000011976359992331707 s 0.96
GenDot / Jax / cpu / BothRev 0.00001020006000317153 s 0.000011309560022709776 s 0.90
GenDot / HLOOpt / cpu / PreRev 0.00001127240001096652 s 0.000012416999970810138 s 0.91
GenDot / HLOOpt / cpu / PostRev 0.00001264602000674131 s 0.00001339757995083346 s 0.94
GenDot / HLOOpt / cpu / BothRev 0.000011217700007364327 s 0.000011228120001760544 s 1.00
GenDot / PartOpt / cpu / PreRev 0.000011149759993713816 s 0.000011987920024694176 s 0.93
GenDot / PartOpt / cpu / PostRev 0.00001017342000068311 s 0.00001231975996233814 s 0.83
GenDot / PartOpt / cpu / BothRev 0.0000114205599993511 s 0.000012321479962338344 s 0.93
GenDot / IPartOpt / cpu / PreRev 0.000011031660008029576 s 0.000011272580013610423 s 0.98
GenDot / IPartOpt / cpu / PostRev 0.000010528539999086206 s 0.000012042980015394278 s 0.87
GenDot / IPartOpt / cpu / BothRev 0.000010413259990400548 s 0.000011842540016004931 s 0.88
GenDot / DefOpt / cpu / PreRev 0.000011051439998936986 s 0.000011467599997558865 s 0.96
GenDot / DefOpt / cpu / PostRev 0.000011337839994212118 s 0.000011891319991264026 s 0.95
GenDot / DefOpt / cpu / BothRev 0.000010511379994113667 s 0.000012002500052403774 s 0.88
GenDot / IDefOpt / cpu / PreRev 0.000011223220001284062 s 0.000011844639984701644 s 0.95
GenDot / IDefOpt / cpu / PostRev 0.000011027500002001031 s 0.000012110100014979252 s 0.91
GenDot / IDefOpt / cpu / BothRev 0.000011024639995866891 s 0.000012075799977537829 s 0.91
GenDot / JaXPipe / cuda / Primal 0.000002496 s 0.000002015 s 1.24
GenDot / Jax / cuda / Primal 0.000002495 s 0.000002016 s 1.24
GenDot / HLOOpt / cuda / Primal 0.000002495 s 0.000001984 s 1.26
GenDot / PartOpt / cuda / Primal 0.000002527 s 0.000002015 s 1.25
GenDot / IPartOpt / cuda / Primal 0.000002527 s 0.000002015 s 1.25
GenDot / DefOpt / cuda / Primal 0.000002496 s 0.000001984 s 1.26
GenDot / IDefOpt / cuda / Primal 0.000002495 s 0.000001984 s 1.26
GenDot / JaXPipe / cuda / Forward 0.000010176 s 0.000009792 s 1.04
GenDot / Jax / cuda / Forward 0.000010656 s 0.000010016 s 1.06
GenDot / HLOOpt / cuda / Forward 0.000010624 s 0.000009952 s 1.07
GenDot / PartOpt / cuda / Forward 0.00001056 s 0.00000992 s 1.06
GenDot / IPartOpt / cuda / Forward 0.00001072 s 0.000009792 s 1.09
GenDot / DefOpt / cuda / Forward 0.000010656 s 0.000009632 s 1.11
GenDot / IDefOpt / cuda / Forward 0.00001056 s 0.00001008 s 1.05
GenDot / JaXPipe / cuda / PreRev 0.000010496 s 0.00000992 s 1.06
GenDot / JaXPipe / cuda / PostRev 0.000010752 s 0.000009984 s 1.08
GenDot / JaXPipe / cuda / BothRev 0.000010752 s 0.000009952 s 1.08
GenDot / Jax / cuda / BothRev 0.000010591 s 0.000009952 s 1.06
GenDot / HLOOpt / cuda / PreRev 0.000010688 s 0.000009792 s 1.09
GenDot / HLOOpt / cuda / PostRev 0.000010591 s 0.000010464 s 1.01
GenDot / HLOOpt / cuda / BothRev 0.000011135 s 0.00000992 s 1.12
GenDot / PartOpt / cuda / PreRev 0.000010751 s 0.000009855 s 1.09
GenDot / PartOpt / cuda / PostRev 0.000010304 s 0.000009984 s 1.03
GenDot / PartOpt / cuda / BothRev 0.000010528 s 0.000010016 s 1.05
GenDot / IPartOpt / cuda / PreRev 0.000010528 s 0.000009888 s 1.06
GenDot / IPartOpt / cuda / PostRev 0.000009344 s 0.000010145 s 0.92
GenDot / IPartOpt / cuda / BothRev 0.00001056 s 0.000009664 s 1.09
GenDot / DefOpt / cuda / PreRev 0.000010656 s 0.000010016 s 1.06
GenDot / DefOpt / cuda / PostRev 0.000010464 s 0.000009665 s 1.08
GenDot / DefOpt / cuda / BothRev 0.00001056 s 0.000009856 s 1.07
GenDot / IDefOpt / cuda / PreRev 0.000010593 s 0.000009984 s 1.06
GenDot / IDefOpt / cuda / PostRev 0.0000104 s 0.000010016 s 1.04
GenDot / IDefOpt / cuda / BothRev 0.000010688 s 0.000009792 s 1.09
GenDot / JaXPipe / tpu / Primal 9.29575e-7 s 9.30175e-7 s 1.00
GenDot / Jax / tpu / Primal 9.2555e-7 s 9.25725e-7 s 1.00
GenDot / HLOOpt / tpu / Primal 0.0000015757750000000002 s 0.00000157525 s 1.00
GenDot / PartOpt / tpu / Primal 9.25875e-7 s 9.25625e-7 s 1.00
GenDot / IPartOpt / tpu / Primal 9.30225e-7 s 9.297e-7 s 1.00
GenDot / DefOpt / tpu / Primal 0.00000149465 s 0.00000148965 s 1.00
GenDot / IDefOpt / tpu / Primal 0.000001582175 s 0.0000015781 s 1.00
GenDot / JaXPipe / tpu / Forward 0.0000031760500000000003 s 0.000003164375 s 1.00
GenDot / Jax / tpu / Forward 0.00000232045 s 0.000002315075 s 1.00
GenDot / HLOOpt / tpu / Forward 0.0000031277500000000003 s 0.000003115875 s 1.00
GenDot / PartOpt / tpu / Forward 0.000003233825 s 0.00000322785 s 1.00
GenDot / IPartOpt / tpu / Forward 0.000003117625 s 0.000003113875 s 1.00
GenDot / DefOpt / tpu / Forward 0.000003220375 s 0.0000032144750000000003 s 1.00
GenDot / IDefOpt / tpu / Forward 0.000003118625 s 0.00000311405 s 1.00
GenDot / JaXPipe / tpu / PreRev 0.000002971175 s 0.0000029543750000000004 s 1.01
GenDot / JaXPipe / tpu / PostRev 0.00000240815 s 0.00000240305 s 1.00
GenDot / JaXPipe / tpu / BothRev 0.000002958125 s 0.0000029627 s 1.00
GenDot / Jax / tpu / BothRev 0.0000024127750000000004 s 0.000002401275 s 1.00
GenDot / HLOOpt / tpu / PreRev 0.0000029723500000000004 s 0.0000029584250000000004 s 1.00
GenDot / HLOOpt / tpu / PostRev 0.0000029275 s 0.00000294135 s 1.00
GenDot / HLOOpt / tpu / BothRev 0.00000296765 s 0.00000295935 s 1.00
GenDot / PartOpt / tpu / PreRev 0.0000029399 s 0.000002942475 s 1.00
GenDot / PartOpt / tpu / PostRev 0.000002398775 s 0.0000023939 s 1.00
GenDot / PartOpt / tpu / BothRev 0.00000294025 s 0.000002942875 s 1.00
GenDot / IPartOpt / tpu / PreRev 0.000002960475 s 0.0000029719 s 1.00
GenDot / IPartOpt / tpu / PostRev 0.000002410425 s 0.000002410775 s 1.00
GenDot / IPartOpt / tpu / BothRev 0.000002980325 s 0.000002961425 s 1.01
GenDot / DefOpt / tpu / PreRev 0.000002940425 s 0.0000029349 s 1.00
GenDot / DefOpt / tpu / PostRev 0.00000296485 s 0.000002974775 s 1.00
GenDot / DefOpt / tpu / BothRev 0.000002938225 s 0.0000029336750000000003 s 1.00
GenDot / IDefOpt / tpu / PreRev 0.00000298295 s 0.000002968575 s 1.00
GenDot / IDefOpt / tpu / PostRev 0.00000293295 s 0.0000029402250000000003 s 1.00
GenDot / IDefOpt / tpu / BothRev 0.000002980125 s 0.000002968075 s 1.00
GenDot / JaXPipe / cpu / Primal 0.000015516000000000002 s 0.000008296160003737896 s 1.87
GenDot / Jax / cpu / Primal 0.000014901 s 0.000008259619971795473 s 1.80
GenDot / HLOOpt / cpu / Primal 0.000014066 s 0.000008371400035684928 s 1.68
GenDot / PartOpt / cpu / Primal 0.000015588 s 0.00000901005999367044 s 1.73
GenDot / IPartOpt / cpu / Primal 0.00001498 s 0.000008573640043323394 s 1.75
GenDot / DefOpt / cpu / Primal 0.000014295 s 0.000007987099979800406 s 1.79
GenDot / IDefOpt / cpu / Primal 0.000014119 s 0.000007670840022910853 s 1.84
GenDot / JaXPipe / cpu / Forward 0.000019373 s 0.000011564219994397715 s 1.68
GenDot / Jax / cpu / Forward 0.000020421000000000003 s 0.000011687760006680036 s 1.75
GenDot / HLOOpt / cpu / Forward 0.000019408 s 0.000012549280017992717 s 1.55
GenDot / PartOpt / cpu / Forward 0.000019467 s 0.000011167880002176387 s 1.74
GenDot / IPartOpt / cpu / Forward 0.000019438 s 0.000012273299962544117 s 1.58
GenDot / DefOpt / cpu / Forward 0.000019345 s 0.000011512759956531226 s 1.68
GenDot / IDefOpt / cpu / Forward 0.000019934 s 0.000012003320016447103 s 1.66
GenDot / JaXPipe / cpu / PreRev 0.000019513 s 0.000012277580008230873 s 1.59
GenDot / JaXPipe / cpu / PostRev 0.000021119 s 0.000011590839994823907 s 1.82
GenDot / JaXPipe / cpu / BothRev 0.000020144 s 0.000011976359992331707 s 1.68
GenDot / Jax / cpu / BothRev 0.000019993 s 0.000011309560022709776 s 1.77
GenDot / HLOOpt / cpu / PreRev 0.00002042 s 0.000012416999970810138 s 1.64
GenDot / HLOOpt / cpu / PostRev 0.000020393000000000003 s 0.00001339757995083346 s 1.52
GenDot / HLOOpt / cpu / BothRev 0.000019157 s 0.000011228120001760544 s 1.71
GenDot / PartOpt / cpu / PreRev 0.000019171 s 0.000011987920024694176 s 1.60
GenDot / PartOpt / cpu / PostRev 0.000020405 s 0.00001231975996233814 s 1.66
GenDot / PartOpt / cpu / BothRev 0.000019198 s 0.000012321479962338344 s 1.56
GenDot / IPartOpt / cpu / PreRev 0.000019247 s 0.000011272580013610423 s 1.71
GenDot / IPartOpt / cpu / PostRev 0.000020531 s 0.000012042980015394278 s 1.70
GenDot / IPartOpt / cpu / BothRev 0.000019199 s 0.000011842540016004931 s 1.62
GenDot / DefOpt / cpu / PreRev 0.000019784 s 0.000011467599997558865 s 1.73
GenDot / DefOpt / cpu / PostRev 0.000019906 s 0.000011891319991264026 s 1.67
GenDot / DefOpt / cpu / BothRev 0.000019323 s 0.000012002500052403774 s 1.61
GenDot / IDefOpt / cpu / PreRev 0.000019563 s 0.000011844639984701644 s 1.65
GenDot / IDefOpt / cpu / PostRev 0.000018855 s 0.000012110100014979252 s 1.56
GenDot / IDefOpt / cpu / BothRev 0.000019256 s 0.000012075799977537829 s 1.59
GenDot / JaXPipe / cpu / Primal 0.000011 s 0.000008296160003737896 s 1.33
GenDot / Jax / cpu / Primal 0.000011 s 0.000008259619971795473 s 1.33
GenDot / HLOOpt / cpu / Primal 0.00001 s 0.000008371400035684928 s 1.19
GenDot / PartOpt / cpu / Primal 0.00001 s 0.00000901005999367044 s 1.11
GenDot / IPartOpt / cpu / Primal 0.000011 s 0.000008573640043323394 s 1.28
GenDot / DefOpt / cpu / Primal 0.00001 s 0.000007987099979800406 s 1.25
GenDot / IDefOpt / cpu / Primal 0.00001 s 0.000007670840022910853 s 1.30
GenDot / JaXPipe / cpu / Forward 0.000014 s 0.000011564219994397715 s 1.21
GenDot / Jax / cpu / Forward 0.000015 s 0.000011687760006680036 s 1.28
GenDot / HLOOpt / cpu / Forward 0.000014 s 0.000012549280017992717 s 1.12
GenDot / PartOpt / cpu / Forward 0.000015 s 0.000011167880002176387 s 1.34
GenDot / IPartOpt / cpu / Forward 0.000014 s 0.000012273299962544117 s 1.14
GenDot / DefOpt / cpu / Forward 0.000014 s 0.000011512759956531226 s 1.22
GenDot / IDefOpt / cpu / Forward 0.000014 s 0.000012003320016447103 s 1.17
GenDot / JaXPipe / cpu / PreRev 0.000014 s 0.000012277580008230873 s 1.14
GenDot / JaXPipe / cpu / PostRev 0.000014 s 0.000011590839994823907 s 1.21
GenDot / JaXPipe / cpu / BothRev 0.000014 s 0.000011976359992331707 s 1.17
GenDot / Jax / cpu / BothRev 0.000015 s 0.000011309560022709776 s 1.33
GenDot / HLOOpt / cpu / PreRev 0.000015 s 0.000012416999970810138 s 1.21
GenDot / HLOOpt / cpu / PostRev 0.000015 s 0.00001339757995083346 s 1.12
GenDot / HLOOpt / cpu / BothRev 0.000014 s 0.000011228120001760544 s 1.25
GenDot / PartOpt / cpu / PreRev 0.000014 s 0.000011987920024694176 s 1.17
GenDot / PartOpt / cpu / PostRev 0.000015 s 0.00001231975996233814 s 1.22
GenDot / PartOpt / cpu / BothRev 0.000014 s 0.000012321479962338344 s 1.14
GenDot / IPartOpt / cpu / PreRev 0.000014 s 0.000011272580013610423 s 1.24
GenDot / IPartOpt / cpu / PostRev 0.000015 s 0.000012042980015394278 s 1.25
GenDot / IPartOpt / cpu / BothRev 0.000014 s 0.000011842540016004931 s 1.18
GenDot / DefOpt / cpu / PreRev 0.000014 s 0.000011467599997558865 s 1.22
GenDot / DefOpt / cpu / PostRev 0.000014 s 0.000011891319991264026 s 1.18
GenDot / DefOpt / cpu / BothRev 0.000015 s 0.000012002500052403774 s 1.25
GenDot / IDefOpt / cpu / PreRev 0.000015 s 0.000011844639984701644 s 1.27
GenDot / IDefOpt / cpu / PostRev 0.000014 s 0.000012110100014979252 s 1.16
GenDot / IDefOpt / cpu / BothRev 0.000015 s 0.000012075799977537829 s 1.24
hlo_ffi / JaXPipe / cpu / Primal 0.000011069800011682672 s 0.000010163059996557424 s 1.09
hlo_ffi / Jax / cpu / Primal 0.000010401959991668264 s 0.000009343899982923176 s 1.11
hlo_ffi / HLOOpt / cpu / Primal 0.000011500960004013906 s 0.000009920599950419272 s 1.16
hlo_ffi / PartOpt / cpu / Primal 0.000010608020004383432 s 0.000009339799971712637 s 1.14
hlo_ffi / IPartOpt / cpu / Primal 0.000011773480000556448 s 0.000009727540027597567 s 1.21
hlo_ffi / DefOpt / cpu / Primal 0.00001112487999989753 s 0.000009657739974500146 s 1.15
hlo_ffi / IDefOpt / cpu / Primal 0.000011843779993796489 s 0.000009740259984027945 s 1.22
hlo_ffi / JaXPipe / cpu / Forward 0.000016607320005732618 s 0.00001370602000861254 s 1.21
hlo_ffi / Jax / cpu / Forward 0.000015605120004238416 s 0.000013844219993188744 s 1.13
hlo_ffi / HLOOpt / cpu / Forward 0.00001578803999336742 s 0.00001383915999213059 s 1.14
hlo_ffi / PartOpt / cpu / Forward 0.000015487040000152773 s 0.000013616979977086883 s 1.14
hlo_ffi / IPartOpt / cpu / Forward 0.00001607402000900038 s 0.0000135127200428542 s 1.19
hlo_ffi / DefOpt / cpu / Forward 0.000015092599999206867 s 0.000013661500024682028 s 1.10
hlo_ffi / IDefOpt / cpu / Forward 0.00001561337999874013 s 0.000013947420038675773 s 1.12
hlo_ffi / JaXPipe / cpu / PreRev 0.000016208839995215384 s 0.000013623760041809874 s 1.19
hlo_ffi / JaXPipe / cpu / PostRev 0.00001548832000707989 s 0.00001384286003485613 s 1.12
hlo_ffi / JaXPipe / cpu / BothRev 0.000015097779992174764 s 0.000014081040008022683 s 1.07
hlo_ffi / Jax / cpu / BothRev 0.000016585980001764257 s 0.000013902360005886294 s 1.19
hlo_ffi / HLOOpt / cpu / PreRev 0.000016859779998412704 s 0.000013973720015201252 s 1.21
hlo_ffi / HLOOpt / cpu / PostRev 0.00001758572001108405 s 0.00001598522000676894 s 1.10
hlo_ffi / HLOOpt / cpu / BothRev 0.000015445620001628412 s 0.000013878419968023082 s 1.11
hlo_ffi / PartOpt / cpu / PreRev 0.000016116799997689668 s 0.000013816640012009884 s 1.17
hlo_ffi / PartOpt / cpu / PostRev 0.000014931759997125482 s 0.000013948579971838624 s 1.07
hlo_ffi / PartOpt / cpu / BothRev 0.000016092899991235753 s 0.000014386660022864815 s 1.12
hlo_ffi / IPartOpt / cpu / PreRev 0.000016242600001987738 s 0.000013554640017900966 s 1.20
hlo_ffi / IPartOpt / cpu / PostRev 0.000015602879998368734 s 0.000013573120013461448 s 1.15
hlo_ffi / IPartOpt / cpu / BothRev 0.00001534966000235727 s 0.000013850839995939167 s 1.11
hlo_ffi / DefOpt / cpu / PreRev 0.000015864200001942663 s 0.000013287640031194317 s 1.19
hlo_ffi / DefOpt / cpu / PostRev 0.000015150199990330292 s 0.000013696799987883425 s 1.11
hlo_ffi / DefOpt / cpu / BothRev 0.00001562863999424735 s 0.000013824719981130328 s 1.13
hlo_ffi / IDefOpt / cpu / PreRev 0.000016428340009042584 s 0.000013894240000809077 s 1.18
hlo_ffi / IDefOpt / cpu / PostRev 0.00001582107999638538 s 0.000013869060003344204 s 1.14
hlo_ffi / IDefOpt / cpu / BothRev 0.000014694279998366256 s 0.000013805299995510722 s 1.06
hlo_ffi / JaXPipe / cuda / Primal 0.0000023670000000000004 s 0.000001983 s 1.19
hlo_ffi / Jax / cuda / Primal 0.0000023670000000000004 s 0.000001983 s 1.19
hlo_ffi / HLOOpt / cuda / Primal 0.0000023670000000000004 s 0.000001983 s 1.19
hlo_ffi / PartOpt / cuda / Primal 0.0000023670000000000004 s 0.000001984 s 1.19
hlo_ffi / IPartOpt / cuda / Primal 0.0000023670000000000004 s 0.000001983 s 1.19
hlo_ffi / DefOpt / cuda / Primal 0.0000023670000000000004 s 0.000001984 s 1.19
hlo_ffi / IDefOpt / cuda / Primal 0.0000023670000000000004 s 0.000001984 s 1.19
hlo_ffi / JaXPipe / cuda / Forward 0.000002432 s 0.00000208 s 1.17
hlo_ffi / Jax / cuda / Forward 0.000002432 s 0.000002049 s 1.19
hlo_ffi / HLOOpt / cuda / Forward 0.000002432 s 0.000002048 s 1.19
hlo_ffi / PartOpt / cuda / Forward 0.000002432 s 0.000002048 s 1.19
hlo_ffi / IPartOpt / cuda / Forward 0.000002432 s 0.00000208 s 1.17
hlo_ffi / DefOpt / cuda / Forward 0.000002431 s 0.00000208 s 1.17
hlo_ffi / IDefOpt / cuda / Forward 0.000002431 s 0.00000208 s 1.17
hlo_ffi / JaXPipe / cuda / PreRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / JaXPipe / cuda / PostRev 0.000002432 s 0.000002047 s 1.19
hlo_ffi / JaXPipe / cuda / BothRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / Jax / cuda / BothRev 0.000002431 s 0.000002048 s 1.19
hlo_ffi / HLOOpt / cuda / PreRev 0.000002431 s 0.000002048 s 1.19
hlo_ffi / HLOOpt / cuda / PostRev 0.000002432 s 0.000002047 s 1.19
hlo_ffi / HLOOpt / cuda / BothRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / PartOpt / cuda / PreRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / PartOpt / cuda / PostRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / PartOpt / cuda / BothRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / IPartOpt / cuda / PreRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / IPartOpt / cuda / PostRev 0.000002431 s 0.000002048 s 1.19
hlo_ffi / IPartOpt / cuda / BothRev 0.000002431 s 0.000002048 s 1.19
hlo_ffi / DefOpt / cuda / PreRev 0.000002431 s 0.000002048 s 1.19
hlo_ffi / DefOpt / cuda / PostRev 0.000002432 s 0.000002047 s 1.19
hlo_ffi / DefOpt / cuda / BothRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / IDefOpt / cuda / PreRev 0.000002431 s 0.000002048 s 1.19
hlo_ffi / IDefOpt / cuda / PostRev 0.000002431 s 0.000002047 s 1.19
hlo_ffi / IDefOpt / cuda / BothRev 0.000002431 s 0.000002048 s 1.19
hlo_ffi / JaXPipe / tpu / Primal 9.27925e-7 s 9.3215e-7 s 1.00
hlo_ffi / Jax / tpu / Primal 9.57425e-7 s 9.53125e-7 s 1.00
hlo_ffi / HLOOpt / tpu / Primal 9.07025e-7 s 9.07375e-7 s 1.00
hlo_ffi / PartOpt / tpu / Primal 9.55325e-7 s 9.50025e-7 s 1.01
hlo_ffi / IPartOpt / tpu / Primal 9.07475e-7 s 9.1355e-7 s 0.99
hlo_ffi / DefOpt / tpu / Primal 9.55075e-7 s 9.58125e-7 s 1.00
hlo_ffi / IDefOpt / tpu / Primal 9.074e-7 s 9.0525e-7 s 1.00
hlo_ffi / JaXPipe / tpu / Forward 9.49075e-7 s 9.49075e-7 s 1
hlo_ffi / Jax / tpu / Forward 9.822e-7 s 9.81575e-7 s 1.00
hlo_ffi / HLOOpt / tpu / Forward 9.7535e-7 s 9.73475e-7 s 1.00
hlo_ffi / PartOpt / tpu / Forward 9.34025e-7 s 9.33675e-7 s 1.00
hlo_ffi / IPartOpt / tpu / Forward 9.74275e-7 s 9.736749999999998e-7 s 1.00
hlo_ffi / DefOpt / tpu / Forward 9.34525e-7 s 9.33775e-7 s 1.00
hlo_ffi / IDefOpt / tpu / Forward 9.74825e-7 s 9.73925e-7 s 1.00
hlo_ffi / JaXPipe / tpu / PreRev 9.38475e-7 s 9.3785e-7 s 1.00
hlo_ffi / JaXPipe / tpu / PostRev 9.659e-7 s 9.64675e-7 s 1.00
hlo_ffi / JaXPipe / tpu / BothRev 9.62725e-7 s 9.62325e-7 s 1.00
hlo_ffi / Jax / tpu / BothRev 9.65625e-7 s 9.64675e-7 s 1.00
hlo_ffi / HLOOpt / tpu / PreRev 9.632499999999998e-7 s 9.62175e-7 s 1.00
hlo_ffi / HLOOpt / tpu / PostRev 9.6535e-7 s 9.64525e-7 s 1.00
hlo_ffi / HLOOpt / tpu / BothRev 9.62275e-7 s 9.62325e-7 s 1.00
hlo_ffi / PartOpt / tpu / PreRev 9.6555e-7 s 9.647e-7 s 1.00
hlo_ffi / PartOpt / tpu / PostRev 9.62075e-7 s 9.6175e-7 s 1.00
hlo_ffi / PartOpt / tpu / BothRev 9.64925e-7 s 9.64425e-7 s 1.00
hlo_ffi / IPartOpt / tpu / PreRev 9.625e-7 s 9.61625e-7 s 1.00
hlo_ffi / IPartOpt / tpu / PostRev 9.6515e-7 s 9.64925e-7 s 1.00
hlo_ffi / IPartOpt / tpu / BothRev 9.62225e-7 s 9.618500000000002e-7 s 1.00
hlo_ffi / DefOpt / tpu / PreRev 9.6525e-7 s 9.64575e-7 s 1.00
hlo_ffi / DefOpt / tpu / PostRev 9.62675e-7 s 9.6215e-7 s 1.00
hlo_ffi / DefOpt / tpu / BothRev 9.65375e-7 s 9.6475e-7 s 1.00
hlo_ffi / IDefOpt / tpu / PreRev 9.61725e-7 s 9.6165e-7 s 1.00
hlo_ffi / IDefOpt / tpu / PostRev 9.6545e-7 s 9.64825e-7 s 1.00
hlo_ffi / IDefOpt / tpu / BothRev 9.62675e-7 s 9.61925e-7 s 1.00
hlo_ffi / JaXPipe / cpu / Primal 0.000018333 s 0.000010163059996557424 s 1.80
hlo_ffi / Jax / cpu / Primal 0.000018381 s 0.000009343899982923176 s 1.97
hlo_ffi / HLOOpt / cpu / Primal 0.000018343 s 0.000009920599950419272 s 1.85
hlo_ffi / PartOpt / cpu / Primal 0.000018326 s 0.000009339799971712637 s 1.96
hlo_ffi / IPartOpt / cpu / Primal 0.000017794 s 0.000009727540027597567 s 1.83
hlo_ffi / DefOpt / cpu / Primal 0.000018222 s 0.000009657739974500146 s 1.89
hlo_ffi / IDefOpt / cpu / Primal 0.00001812 s 0.000009740259984027945 s 1.86
hlo_ffi / JaXPipe / cpu / Forward 0.000025334 s 0.00001370602000861254 s 1.85
hlo_ffi / Jax / cpu / Forward 0.000024662 s 0.000013844219993188744 s 1.78
hlo_ffi / HLOOpt / cpu / Forward 0.000024432 s 0.00001383915999213059 s 1.77
hlo_ffi / PartOpt / cpu / Forward 0.000024989 s 0.000013616979977086883 s 1.84
hlo_ffi / IPartOpt / cpu / Forward 0.000024807 s 0.0000135127200428542 s 1.84
hlo_ffi / DefOpt / cpu / Forward 0.000025618 s 0.000013661500024682028 s 1.88
hlo_ffi / IDefOpt / cpu / Forward 0.000025099 s 0.000013947420038675773 s 1.80
hlo_ffi / JaXPipe / cpu / PreRev 0.000025374 s 0.000013623760041809874 s 1.86
hlo_ffi / JaXPipe / cpu / PostRev 0.000024226 s 0.00001384286003485613 s 1.75
hlo_ffi / JaXPipe / cpu / BothRev 0.000024604 s 0.000014081040008022683 s 1.75
hlo_ffi / Jax / cpu / BothRev 0.000025102 s 0.000013902360005886294 s 1.81
hlo_ffi / HLOOpt / cpu / PreRev 0.000024745 s 0.000013973720015201252 s 1.77
hlo_ffi / HLOOpt / cpu / PostRev 0.000024858 s 0.00001598522000676894 s 1.56
hlo_ffi / HLOOpt / cpu / BothRev 0.000024576 s 0.000013878419968023082 s 1.77
hlo_ffi / PartOpt / cpu / PreRev 0.000025012 s 0.000013816640012009884 s 1.81
hlo_ffi / PartOpt / cpu / PostRev 0.000024558 s 0.000013948579971838624 s 1.76
hlo_ffi / PartOpt / cpu / BothRev 0.000025008 s 0.000014386660022864815 s 1.74
hlo_ffi / IPartOpt / cpu / PreRev 0.0000249 s 0.000013554640017900966 s 1.84
hlo_ffi / IPartOpt / cpu / PostRev 0.000024837000000000003 s 0.000013573120013461448 s 1.83
hlo_ffi / IPartOpt / cpu / BothRev 0.000024693 s 0.000013850839995939167 s 1.78
hlo_ffi / DefOpt / cpu / PreRev 0.000025106 s 0.000013287640031194317 s 1.89
hlo_ffi / DefOpt / cpu / PostRev 0.000023702 s 0.000013696799987883425 s 1.73
hlo_ffi / DefOpt / cpu / BothRev 0.000023961 s 0.000013824719981130328 s 1.73
hlo_ffi / IDefOpt / cpu / PreRev 0.00002522 s 0.000013894240000809077 s 1.82
hlo_ffi / IDefOpt / cpu / PostRev 0.000024647 s 0.000013869060003344204 s 1.78
hlo_ffi / IDefOpt / cpu / BothRev 0.000024281 s 0.000013805299995510722 s 1.76
hlo_ffi / JaXPipe / cpu / Primal 0.000013 s 0.000010163059996557424 s 1.28
hlo_ffi / Jax / cpu / Primal 0.000013 s 0.000009343899982923176 s 1.39
hlo_ffi / HLOOpt / cpu / Primal 0.000013 s 0.000009920599950419272 s 1.31
hlo_ffi / PartOpt / cpu / Primal 0.000013 s 0.000009339799971712637 s 1.39
hlo_ffi / IPartOpt / cpu / Primal 0.000012 s 0.000009727540027597567 s 1.23
hlo_ffi / DefOpt / cpu / Primal 0.000013 s 0.000009657739974500146 s 1.35
hlo_ffi / IDefOpt / cpu / Primal 0.000012 s 0.000009740259984027945 s 1.23
hlo_ffi / JaXPipe / cpu / Forward 0.000017 s 0.00001370602000861254 s 1.24
hlo_ffi / Jax / cpu / Forward 0.000017 s 0.000013844219993188744 s 1.23
hlo_ffi / HLOOpt / cpu / Forward 0.000017 s 0.00001383915999213059 s 1.23
hlo_ffi / PartOpt / cpu / Forward 0.000017 s 0.000013616979977086883 s 1.25
hlo_ffi / IPartOpt / cpu / Forward 0.000019 s 0.0000135127200428542 s 1.41
hlo_ffi / DefOpt / cpu / Forward 0.000017 s 0.000013661500024682028 s 1.24
hlo_ffi / IDefOpt / cpu / Forward 0.000017 s 0.000013947420038675773 s 1.22
hlo_ffi / JaXPipe / cpu / PreRev 0.000017 s 0.000013623760041809874 s 1.25
hlo_ffi / JaXPipe / cpu / PostRev 0.000017 s 0.00001384286003485613 s 1.23
hlo_ffi / JaXPipe / cpu / BothRev 0.000017 s 0.000014081040008022683 s 1.21
hlo_ffi / Jax / cpu / BothRev 0.000017 s 0.000013902360005886294 s 1.22
hlo_ffi / HLOOpt / cpu / PreRev 0.000017999999999999997 s 0.000013973720015201252 s 1.29
hlo_ffi / HLOOpt / cpu / PostRev 0.000017999999999999997 s 0.00001598522000676894 s 1.13
hlo_ffi / HLOOpt / cpu / BothRev 0.000017999999999999997 s 0.000013878419968023082 s 1.30
hlo_ffi / PartOpt / cpu / PreRev 0.000017 s 0.000013816640012009884 s 1.23
hlo_ffi / PartOpt / cpu / PostRev 0.000019 s 0.000013948579971838624 s 1.36
hlo_ffi / PartOpt / cpu / BothRev 0.000017999999999999997 s 0.000014386660022864815 s 1.25
hlo_ffi / IPartOpt / cpu / PreRev 0.000017999999999999997 s 0.000013554640017900966 s 1.33
hlo_ffi / IPartOpt / cpu / PostRev 0.000017999999999999997 s 0.000013573120013461448 s 1.33
hlo_ffi / IPartOpt / cpu / BothRev 0.000017 s 0.000013850839995939167 s 1.23
hlo_ffi / DefOpt / cpu / PreRev 0.000017 s 0.000013287640031194317 s 1.28
hlo_ffi / DefOpt / cpu / PostRev 0.000017 s 0.000013696799987883425 s 1.24
hlo_ffi / DefOpt / cpu / BothRev 0.000017 s 0.000013824719981130328 s 1.23
hlo_ffi / IDefOpt / cpu / PreRev 0.000017 s 0.000013894240000809077 s 1.22
hlo_ffi / IDefOpt / cpu / PostRev 0.000017 s 0.000013869060003344204 s 1.23
hlo_ffi / IDefOpt / cpu / BothRev 0.000017 s 0.000013805299995510722 s 1.23
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / Primal 0.0009213158000193 s 0.0008871002001797 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / Primal 0.0009104983999804 s 0.0008837956000206 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / Primal 0.0009805919999735 s 0.0009228236001035 s 1.06
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / Primal 0.0009094214000015 s 0.0008805019999272 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / Primal 0.0009250901999848 s 0.0008894220000001 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / Primal 0.0009860830000206 s 0.0009508617999017 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / Primal 0.0009907198000064 s 0.0009291351999308 s 1.07
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / Forward 0.002204480000023 s 0.0021042777999355 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / Forward 0.0023456586000293 s 0.0022587204000046 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / Forward 0.0022288163999746 s 0.0021500653998373 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / Forward 0.002199663000033 s 0.0022119191999081 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / Forward 0.002225231400007 s 0.0021429494000585 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / Forward 0.0022475720000102 s 0.0021400593999715 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / Forward 0.0021418807999907 s 0.0021881026000301 s 0.98
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / PreRev 0.0051802711999926 s 0.0050751822000165 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / PostRev 0.0049583580000216 s 0.0056805405999512 s 0.87
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / BothRev 0.0056441988000415 s 0.0058089991998713 s 0.97
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / BothRev 0.0055026360000056 s 0.0052328876001411 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / PreRev 0.0053289049999875 s 0.0055911849999574 s 0.95
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / PostRev 0.005597296799965 s 0.0047908924000694 s 1.17
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / BothRev 0.0070611067999834 s 0.0059099502001117 s 1.19
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / PreRev 0.0049999660000139 s 0.0036509223998109 s 1.37
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / PostRev 0.0035906825999973 s 0.0054702081999494 s 0.66
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / BothRev 0.0051593262000096 s 0.0052289548000771 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / PreRev 0.0070447117999719 s 0.0050246252000761 s 1.40
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / PostRev 0.0051492684000322 s 0.0065582606001044 s 0.79
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / BothRev 0.0032928224000215 s 0.0034128457998122 s 0.96
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / PreRev 0.0056232034000231 s 0.0054061793999608 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / PostRev 0.005711300200005 s 0.004691940000066 s 1.22
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / BothRev 0.0055255829999396 s 0.0050492209999902 s 1.09
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / PreRev 0.003636172599954 s 0.0050608083999577 s 0.72
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / PostRev 0.0050699929999836 s 0.0056577018001007 s 0.90
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / BothRev 0.0035170639999932 s 0.0047789924000426 s 0.74
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cuda / Primal 0.0002968629999999 s 0.000284001 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cuda / Primal 0.0002973109999999 s 0.000284033 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cuda / Primal 0.000304415 s 0.000291169 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cuda / Primal 0.000297503 s 0.000282881 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cuda / Primal 0.000297215 s 0.000283489 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cuda / Primal 0.0003047989999999 s 0.000290817 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cuda / Primal 0.000303359 s 0.00029104 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cuda / Forward 0.000582654 s 0.000555745 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cuda / Forward 0.000566686 s 0.000537986 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cuda / Forward 0.000582366 s 0.0005573129999999 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cuda / Forward 0.00058227 s 0.000556898 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cuda / Forward 0.0005819499999999 s 0.000556737 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cuda / Forward 0.000582109 s 0.000557762 s 1.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cuda / Forward 0.000582526 s 0.000557154 s 1.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cuda / PreRev 0.001054748 s 0.001026882 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cuda / PostRev 0.001008541 s 0.000984706 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cuda / BothRev 0.001050653 s 0.001019711 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cuda / BothRev 0.001004733 s 0.000986369 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cuda / PreRev 0.001036893 s 0.0010103089999999 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cuda / PostRev 0.0010606679999999 s 0.001035842 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cuda / BothRev 0.001037788 s 0.001006659 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cuda / PreRev 0.001050044 s 0.001025059 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cuda / PostRev 0.000998044 s 0.000975939 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cuda / BothRev 0.0010503 s 0.001024803 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cuda / PreRev 0.0010503 s 0.001022979 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cuda / PostRev 0.000997788 s 0.000974241 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cuda / BothRev 0.00105126 s 0.001024098 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cuda / PreRev 0.001052284 s 0.00101962 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cuda / PostRev 0.000985276 s 0.00095821 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cuda / BothRev 0.001048508 s 0.001023619 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cuda / PreRev 0.001052444 s 0.001021217 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cuda / PostRev 0.001053276 s 0.001017858 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cuda / BothRev 0.001053724 s 0.001020512 s 1.03
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / tpu / Primal 0.0001264599999999 s 0.00012352525 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / tpu / Primal 0.000127945 s 0.00012669775 s 1.01
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / tpu / Primal 0.00015585725 s 0.00015251875 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / tpu / Primal 0.00013539725 s 0.000134193 s 1.01
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / tpu / Primal 0.00013379625 s 0.00013077025 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / tpu / Primal 0.000148733 s 0.00014781275 s 1.01
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / tpu / Primal 0.000153813 s 0.00015106675 s 1.02
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / tpu / Forward 0.00021452475 s 0.000211938 s 1.01
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / tpu / Forward 0.00026106075 s 0.0002610617499999 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / tpu / Forward 0.00021419375 s 0.0002121707499999 s 1.01
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / tpu / Forward 0.0002174805 s 0.0002180217499999 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / tpu / Forward 0.00021441 s 0.00021172775 s 1.01
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / tpu / Forward 0.00021754825 s 0.00021815225 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / tpu / Forward 0.0002147349999999 s 0.0002117155 s 1.01
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / tpu / PreRev 0.00035538625 s 0.0003562175 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / tpu / PostRev 0.00025663025 s 0.0002588909999999 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / tpu / BothRev 0.0003546877499999 s 0.0003564275 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / tpu / BothRev 0.00025663675 s 0.0002595994999999 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / tpu / PreRev 0.00035502475 s 0.00035665025 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / tpu / PostRev 0.0002908104999999 s 0.00029234875 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / tpu / BothRev 0.00035497275 s 0.0003562995 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / tpu / PreRev 0.00035519625 s 0.00035881575 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / tpu / PostRev 0.00027154975 s 0.0002721429999999 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / tpu / BothRev 0.000355524 s 0.000358587 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / tpu / PreRev 0.0003545835 s 0.00035639775 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / tpu / PostRev 0.00027216325 s 0.0002747947499999 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / tpu / BothRev 0.0003545447499999 s 0.00035658475 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / tpu / PreRev 0.0003576357499999 s 0.0003597825 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / tpu / PostRev 0.0002831822499999 s 0.0002841942499999 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / tpu / BothRev 0.00035758025 s 0.0003598795 s 0.99
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / tpu / PreRev 0.00035697425 s 0.000357839 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / tpu / PostRev 0.00030080975 s 0.000302267 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / tpu / BothRev 0.00035766375 s 0.0003577502499999 s 1.00
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / Primal 0.001911167 s 0.0008871002001797 s 2.15
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / Primal 0.002112111 s 0.0008837956000206 s 2.39
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / Primal 0.002254026 s 0.0009228236001035 s 2.44
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / Primal 0.002523412 s 0.0008805019999272 s 2.87
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / Primal 0.002162878 s 0.0008894220000001 s 2.43
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / Primal 0.002169751 s 0.0009508617999017 s 2.28
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / Primal 0.002182197 s 0.0009291351999308 s 2.35
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / Forward 0.005627352 s 0.0021042777999355 s 2.67
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / Forward 0.005940199 s 0.0022587204000046 s 2.63
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / Forward 0.0059031529999999 s 0.0021500653998373 s 2.75
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / Forward 0.005963574 s 0.0022119191999081 s 2.70
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / Forward 0.005342604 s 0.0021429494000585 s 2.49
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / Forward 0.005899452 s 0.0021400593999715 s 2.76
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / Forward 0.0052812929999999 s 0.0021881026000301 s 2.41
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / PreRev 0.00943576 s 0.0050751822000165 s 1.86
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / PostRev 0.008678521 s 0.0056805405999512 s 1.53
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / BothRev 0.008068538 s 0.0058089991998713 s 1.39
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / BothRev 0.008703638 s 0.0052328876001411 s 1.66
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / PreRev 0.010425204 s 0.0055911849999574 s 1.86
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / PostRev 0.008165762 s 0.0047908924000694 s 1.70
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / BothRev 0.010082184 s 0.0059099502001117 s 1.71
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / PreRev 0.008714896 s 0.0036509223998109 s 2.39
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / PostRev 0.009681898 s 0.0054702081999494 s 1.77
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / BothRev 0.007816826 s 0.0052289548000771 s 1.49
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / PreRev 0.008819219 s 0.0050246252000761 s 1.76
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / PostRev 0.008366284 s 0.0065582606001044 s 1.28
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / BothRev 0.010173897 s 0.0034128457998122 s 2.98
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / PreRev 0.008575517 s 0.0054061793999608 s 1.59
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / PostRev 0.008188604 s 0.004691940000066 s 1.75
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / BothRev 0.008236665 s 0.0050492209999902 s 1.63
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / PreRev 0.00875351 s 0.0050608083999577 s 1.73
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / PostRev 0.007653495 s 0.0056577018001007 s 1.35
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / BothRev 0.0107052049999999 s 0.0047789924000426 s 2.24
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / Primal 0.002639 s 0.0008871002001797 s 2.97
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / Primal 0.002012 s 0.0008837956000206 s 2.28
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / Primal 0.002322 s 0.0009228236001035 s 2.52
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / Primal 0.002262 s 0.0008805019999272 s 2.57
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / Primal 0.002578 s 0.0008894220000001 s 2.90
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / Primal 0.002051 s 0.0009508617999017 s 2.16
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / Primal 0.001796 s 0.0009291351999308 s 1.93
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / Forward 0.004546 s 0.0021042777999355 s 2.16
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / Forward 0.004186 s 0.0022587204000046 s 1.85
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / Forward 0.004416 s 0.0021500653998373 s 2.05
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / Forward 0.004726 s 0.0022119191999081 s 2.14
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / Forward 0.004468 s 0.0021429494000585 s 2.08
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / Forward 0.004749 s 0.0021400593999715 s 2.22
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / Forward 0.004868 s 0.0021881026000301 s 2.22
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / PreRev 0.011094 s 0.0050751822000165 s 2.19
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / PostRev 0.020307 s 0.0056805405999512 s 3.57
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / JaXPipe / cpu / BothRev 0.008717 s 0.0058089991998713 s 1.50
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / Jax / cpu / BothRev 0.009814 s 0.0052328876001411 s 1.88
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / PreRev 0.0088209999999999 s 0.0055911849999574 s 1.58
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / PostRev 0.008016 s 0.0047908924000694 s 1.67
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / BothRev 0.0148379999999999 s 0.0059099502001117 s 2.51
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / PreRev 0.010439 s 0.0036509223998109 s 2.86
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / PostRev 0.014638 s 0.0054702081999494 s 2.68
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / PartOpt / cpu / BothRev 0.012754 s 0.0052289548000771 s 2.44
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / PreRev 0.0094 s 0.0050246252000761 s 1.87
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / PostRev 0.017626 s 0.0065582606001044 s 2.69
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IPartOpt / cpu / BothRev 0.0100679999999999 s 0.0034128457998122 s 2.95
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / PreRev 0.013432 s 0.0054061793999608 s 2.48
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / PostRev 0.0097899999999999 s 0.004691940000066 s 2.09
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / BothRev 0.0114749999999999 s 0.0050492209999902 s 2.27
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / PreRev 0.009778 s 0.0050608083999577 s 1.93
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / PostRev 0.011222 s 0.0056577018001007 s 1.98
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / IDefOpt / cpu / BothRev 0.012923 s 0.0047789924000426 s 2.70
scatter_sum / JaXPipe / cpu / Primal 0.000007667100005619432 s 0.000008399699972869711 s 0.91
scatter_sum / Jax / cpu / Primal 0.000007718980009485677 s 0.00000856758005284064 s 0.90
scatter_sum / HLOOpt / cpu / Primal 0.000007659139998850151 s 0.000009258660020350362 s 0.83
scatter_sum / PartOpt / cpu / Primal 0.000007713320005677815 s 0.000009661359954407087 s 0.80
scatter_sum / IPartOpt / cpu / Primal 0.000007863519997499679 s 0.000009734099976412837 s 0.81
scatter_sum / DefOpt / cpu / Primal 0.000007770459999392188 s 0.000009198239977195045 s 0.84
scatter_sum / IDefOpt / cpu / Primal 0.00000765412000419019 s 0.000008779059990047244 s 0.87
scatter_sum / JaXPipe / cpu / Forward 0.000011554239993074589 s 0.000013061339986961684 s 0.88
scatter_sum / Jax / cpu / Forward 0.000011517739997088938 s 0.000012101459969926508 s 0.95
scatter_sum / HLOOpt / cpu / Forward 0.0000116724200029239 s 0.000012741700002152357 s 0.92
scatter_sum / PartOpt / cpu / Forward 0.00001154473999349648 s 0.000012532019991340348 s 0.92
scatter_sum / IPartOpt / cpu / Forward 0.000011817240006166685 s 0.0000133172999630915 s 0.89
scatter_sum / DefOpt / cpu / Forward 0.000011767280007006775 s 0.00001335968000603316 s 0.88
scatter_sum / IDefOpt / cpu / Forward 0.000011618659993928303 s 0.000013295580029080156 s 0.87
scatter_sum / JaXPipe / cpu / PreRev 0.00001236743999470491 s 0.000012938739982928384 s 0.96
scatter_sum / JaXPipe / cpu / PostRev 0.000011306499993679609 s 0.00001292862001719186 s 0.87
scatter_sum / JaXPipe / cpu / BothRev 0.000011907039995548983 s 0.000012804760017388616 s 0.93
scatter_sum / Jax / cpu / BothRev 0.000011951460005548142 s 0.000012629419989025336 s 0.95
scatter_sum / HLOOpt / cpu / PreRev 0.000012150580000707122 s 0.000013125519999448442 s 0.93
scatter_sum / HLOOpt / cpu / PostRev 0.000013491180009168602 s 0.000014698720024171053 s 0.92
scatter_sum / HLOOpt / cpu / BothRev 0.000011304120002932905 s 0.000013126020039635478 s 0.86
scatter_sum / PartOpt / cpu / PreRev 0.000011587799995140814 s 0.000013033599962000154 s 0.89
scatter_sum / PartOpt / cpu / PostRev 0.000011495760006710042 s 0.000012915639981656568 s 0.89
scatter_sum / PartOpt / cpu / BothRev 0.000011804099999608296 s 0.000012708979975286638 s 0.93
scatter_sum / IPartOpt / cpu / PreRev 0.000011991100002433086 s 0.000012781279965565772 s 0.94
scatter_sum / IPartOpt / cpu / PostRev 0.000011454740001681784 s 0.000013472300042849384 s 0.85
scatter_sum / IPartOpt / cpu / BothRev 0.0000113740600068013 s 0.000012834419976570643 s 0.89
scatter_sum / DefOpt / cpu / PreRev 0.000011613260005560732 s 0.000013145060020178787 s 0.88
scatter_sum / DefOpt / cpu / PostRev 0.000011694960003296728 s 0.00001275038002859219 s 0.92
scatter_sum / DefOpt / cpu / BothRev 0.00001120000000582877 s 0.000012998379997952724 s 0.86
scatter_sum / IDefOpt / cpu / PreRev 0.00001162016000762378 s 0.000013091199998598312 s 0.89
scatter_sum / IDefOpt / cpu / PostRev 0.000011676740000439169 s 0.00001284074000068358 s 0.91
scatter_sum / IDefOpt / cpu / BothRev 0.00001169420000678656 s 0.000012844779957958964 s 0.91
scatter_sum / JaXPipe / cuda / Primal 0.000010784 s 0.000009665 s 1.12
scatter_sum / Jax / cuda / Primal 0.00001072 s 0.000010208 s 1.05
scatter_sum / HLOOpt / cuda / Primal 0.000010367 s 0.00000976 s 1.06
scatter_sum / PartOpt / cuda / Primal 0.000010496 s 0.000009824 s 1.07
scatter_sum / IPartOpt / cuda / Primal 0.000010688 s 0.00000992 s 1.08
scatter_sum / DefOpt / cuda / Primal 0.000010529 s 0.000009888 s 1.06
scatter_sum / IDefOpt / cuda / Primal 0.000010656 s 0.000009952 s 1.07
scatter_sum / JaXPipe / cuda / Forward 0.000017247999999999998 s 0.000016768000000000003 s 1.03
scatter_sum / Jax / cuda / Forward 0.000017216 s 0.000016255999999999998 s 1.06
scatter_sum / HLOOpt / cuda / Forward 0.000017695 s 0.000016927999999999998 s 1.05
scatter_sum / PartOpt / cuda / Forward 0.000019072 s 0.000016576000000000002 s 1.15
scatter_sum / IPartOpt / cuda / Forward 0.00001984 s 0.00001696 s 1.17
scatter_sum / DefOpt / cuda / Forward 0.000016448000000000002 s 0.000016288 s 1.01
scatter_sum / IDefOpt / cuda / Forward 0.000017184 s 0.000018336 s 0.94
scatter_sum / JaXPipe / cuda / PreRev 0.000016704 s 0.000018624000000000003 s 0.90
scatter_sum / JaXPipe / cuda / PostRev 0.000018528 s 0.000016511 s 1.12
scatter_sum / JaXPipe / cuda / BothRev 0.000018688 s 0.00001616 s 1.16
scatter_sum / Jax / cuda / BothRev 0.000017888000000000002 s 0.00001712 s 1.04
scatter_sum / HLOOpt / cuda / PreRev 0.00001952 s 0.000016704 s 1.17
scatter_sum / HLOOpt / cuda / PostRev 0.000017472 s 0.000015935999999999998 s 1.10
scatter_sum / HLOOpt / cuda / BothRev 0.000017151 s 0.000021728 s 0.79
scatter_sum / PartOpt / cuda / PreRev 0.000017024 s 0.000016832 s 1.01
scatter_sum / PartOpt / cuda / PostRev 0.000016832 s 0.000015904000000000002 s 1.06
scatter_sum / PartOpt / cuda / BothRev 0.000017056 s 0.000017568000000000002 s 0.97
scatter_sum / IPartOpt / cuda / PreRev 0.000017375999999999998 s 0.000016832 s 1.03
scatter_sum / IPartOpt / cuda / PostRev 0.000017184 s 0.000016191 s 1.06
scatter_sum / IPartOpt / cuda / BothRev 0.00001712 s 0.00001696 s 1.01
scatter_sum / DefOpt / cuda / PreRev 0.000017375999999999998 s 0.000017311 s 1.00
scatter_sum / DefOpt / cuda / PostRev 0.000017216 s 0.000016417000000000002 s 1.05
scatter_sum / DefOpt / cuda / BothRev 0.00001728 s 0.000016257 s 1.06
scatter_sum / IDefOpt / cuda / PreRev 0.000017183 s 0.00001696 s 1.01
scatter_sum / IDefOpt / cuda / PostRev 0.00001696 s 0.000021728 s 0.78
scatter_sum / IDefOpt / cuda / BothRev 0.000017728 s 0.00001712 s 1.04
scatter_sum / JaXPipe / tpu / Primal 0.000001350125 s 0.000001342625 s 1.01
scatter_sum / Jax / tpu / Primal 0.00000140485 s 0.000001403475 s 1.00
scatter_sum / HLOOpt / tpu / Primal 0.000001350875 s 0.000001342625 s 1.01
scatter_sum / PartOpt / tpu / Primal 0.0000014048 s 0.0000014036 s 1.00
scatter_sum / IPartOpt / tpu / Primal 0.000001351125 s 0.000001342725 s 1.01
scatter_sum / DefOpt / tpu / Primal 0.0000014053499999999998 s 0.00000140385 s 1.00
scatter_sum / IDefOpt / tpu / Primal 0.000001351775 s 0.0000013422999999999998 s 1.01
scatter_sum / JaXPipe / tpu / Forward 0.0000027063 s 0.0000027100000000000003 s 1.00
scatter_sum / Jax / tpu / Forward 0.000002735425 s 0.000002715175 s 1.01
scatter_sum / HLOOpt / tpu / Forward 0.000002704 s 0.0000027005750000000004 s 1.00
scatter_sum / PartOpt / tpu / Forward 0.00000271045 s 0.0000026893499999999995 s 1.01
scatter_sum / IPartOpt / tpu / Forward 0.0000027014750000000003 s 0.0000027078 s 1.00
scatter_sum / DefOpt / tpu / Forward 0.000002693675 s 0.00000269455 s 1.00
scatter_sum / IDefOpt / tpu / Forward 0.00000270625 s 0.0000026998 s 1.00
scatter_sum / JaXPipe / tpu / PreRev 0.000002695175 s 0.0000026912750000000003 s 1.00
scatter_sum / JaXPipe / tpu / PostRev 0.0000026889 s 0.000002691025 s 1.00
scatter_sum / JaXPipe / tpu / BothRev 0.00000270325 s 0.00000269995 s 1.00
scatter_sum / Jax / tpu / BothRev 0.000002742775 s 0.0000027454000000000004 s 1.00
scatter_sum / HLOOpt / tpu / PreRev 0.0000027087250000000005 s 0.0000026961750000000005 s 1.00
scatter_sum / HLOOpt / tpu / PostRev 0.000002744475 s 0.0000027491 s 1.00
scatter_sum / HLOOpt / tpu / BothRev 0.000002708375 s 0.000002698375 s 1.00
scatter_sum / PartOpt / tpu / PreRev 0.00000274675 s 0.000002740025 s 1.00
scatter_sum / PartOpt / tpu / PostRev 0.0000027046500000000004 s 0.0000026995500000000003 s 1.00
scatter_sum / PartOpt / tpu / BothRev 0.0000027474 s 0.000002742875 s 1.00
scatter_sum / IPartOpt / tpu / PreRev 0.00000271605 s 0.00000270345 s 1.00
scatter_sum / IPartOpt / tpu / PostRev 0.000002755275 s 0.000002738425 s 1.01
scatter_sum / IPartOpt / tpu / BothRev 0.00000271015 s 0.0000026955499999999995 s 1.01
scatter_sum / DefOpt / tpu / PreRev 0.0000027438000000000003 s 0.0000027395750000000004 s 1.00
scatter_sum / DefOpt / tpu / PostRev 0.0000027074749999999994 s 0.000002693525 s 1.01
scatter_sum / DefOpt / tpu / BothRev 0.000002744425 s 0.000002744675 s 1.00
scatter_sum / IDefOpt / tpu / PreRev 0.0000027122 s 0.00000269625 s 1.01
scatter_sum / IDefOpt / tpu / PostRev 0.0000027452 s 0.0000027459 s 1.00
scatter_sum / IDefOpt / tpu / BothRev 0.000002708975 s 0.0000026999 s 1.00
scatter_sum / JaXPipe / cpu / Primal 0.000016014 s 0.000008399699972869711 s 1.91
scatter_sum / Jax / cpu / Primal 0.000015633 s 0.00000856758005284064 s 1.82
scatter_sum / HLOOpt / cpu / Primal 0.000015776 s 0.000009258660020350362 s 1.70
scatter_sum / PartOpt / cpu / Primal 0.000015348999999999998 s 0.000009661359954407087 s 1.59
scatter_sum / IPartOpt / cpu / Primal 0.000015882 s 0.000009734099976412837 s 1.63
scatter_sum / DefOpt / cpu / Primal 0.000015362 s 0.000009198239977195045 s 1.67
scatter_sum / IDefOpt / cpu / Primal 0.000015368 s 0.000008779059990047244 s 1.75
scatter_sum / JaXPipe / cpu / Forward 0.000023089 s 0.000013061339986961684 s 1.77
scatter_sum / Jax / cpu / Forward 0.000021899 s 0.000012101459969926508 s 1.81
scatter_sum / HLOOpt / cpu / Forward 0.000022793 s 0.000012741700002152357 s 1.79
scatter_sum / PartOpt / cpu / Forward 0.00002255 s 0.000012532019991340348 s 1.80
scatter_sum / IPartOpt / cpu / Forward 0.000022536 s 0.0000133172999630915 s 1.69
scatter_sum / DefOpt / cpu / Forward 0.000023299 s 0.00001335968000603316 s 1.74
scatter_sum / IDefOpt / cpu / Forward 0.000022733 s 0.000013295580029080156 s 1.71
scatter_sum / JaXPipe / cpu / PreRev 0.000023226 s 0.000012938739982928384 s 1.80
scatter_sum / JaXPipe / cpu / PostRev 0.000023223 s 0.00001292862001719186 s 1.80
scatter_sum / JaXPipe / cpu / BothRev 0.000022982 s 0.000012804760017388616 s 1.79
scatter_sum / Jax / cpu / BothRev 0.000023862000000000003 s 0.000012629419989025336 s 1.89
scatter_sum / HLOOpt / cpu / PreRev 0.000022262 s 0.000013125519999448442 s 1.70
scatter_sum / HLOOpt / cpu / PostRev 0.000023463 s 0.000014698720024171053 s 1.60
scatter_sum / HLOOpt / cpu / BothRev 0.000023006 s 0.000013126020039635478 s 1.75
scatter_sum / PartOpt / cpu / PreRev 0.000023002 s 0.000013033599962000154 s 1.76
scatter_sum / PartOpt / cpu / PostRev 0.000022932 s 0.000012915639981656568 s 1.78
scatter_sum / PartOpt / cpu / BothRev 0.000022635 s 0.000012708979975286638 s 1.78
scatter_sum / IPartOpt / cpu / PreRev 0.000023855 s 0.000012781279965565772 s 1.87
scatter_sum / IPartOpt / cpu / PostRev 0.000022158 s 0.000013472300042849384 s 1.64
scatter_sum / IPartOpt / cpu / BothRev 0.000023493 s 0.000012834419976570643 s 1.83
scatter_sum / DefOpt / cpu / PreRev 0.000022268 s 0.000013145060020178787 s 1.69
scatter_sum / DefOpt / cpu / PostRev 0.000022904 s 0.00001275038002859219 s 1.80
scatter_sum / DefOpt / cpu / BothRev 0.000023208 s 0.000012998379997952724 s 1.79
scatter_sum / IDefOpt / cpu / PreRev 0.000022924 s 0.000013091199998598312 s 1.75
scatter_sum / IDefOpt / cpu / PostRev 0.000023148 s 0.00001284074000068358 s 1.80
scatter_sum / IDefOpt / cpu / BothRev 0.000022634 s 0.000012844779957958964 s 1.76
scatter_sum / JaXPipe / cpu / Primal 0.000011 s 0.000008399699972869711 s 1.31
scatter_sum / Jax / cpu / Primal 0.000011 s 0.00000856758005284064 s 1.28
scatter_sum / HLOOpt / cpu / Primal 0.000011 s 0.000009258660020350362 s 1.19
scatter_sum / PartOpt / cpu / Primal 0.000011 s 0.000009661359954407087 s 1.14
scatter_sum / IPartOpt / cpu / Primal 0.000011 s 0.000009734099976412837 s 1.13
scatter_sum / DefOpt / cpu / Primal 0.000011 s 0.000009198239977195045 s 1.20
scatter_sum / IDefOpt / cpu / Primal 0.000011 s 0.000008779059990047244 s 1.25
scatter_sum / JaXPipe / cpu / Forward 0.000017 s 0.000013061339986961684 s 1.30
scatter_sum / Jax / cpu / Forward 0.000016 s 0.000012101459969926508 s 1.32
scatter_sum / HLOOpt / cpu / Forward 0.000017 s 0.000012741700002152357 s 1.33
scatter_sum / PartOpt / cpu / Forward 0.000017 s 0.000012532019991340348 s 1.36
scatter_sum / IPartOpt / cpu / Forward 0.000016 s 0.0000133172999630915 s 1.20
scatter_sum / DefOpt / cpu / Forward 0.000017999999999999997 s 0.00001335968000603316 s 1.35
scatter_sum / IDefOpt / cpu / Forward 0.000017 s 0.000013295580029080156 s 1.28
scatter_sum / JaXPipe / cpu / PreRev 0.000017 s 0.000012938739982928384 s 1.31
scatter_sum / JaXPipe / cpu / PostRev 0.000017 s 0.00001292862001719186 s 1.31
scatter_sum / JaXPipe / cpu / BothRev 0.000017 s 0.000012804760017388616 s 1.33
scatter_sum / Jax / cpu / BothRev 0.000016 s 0.000012629419989025336 s 1.27
scatter_sum / HLOOpt / cpu / PreRev 0.000017 s 0.000013125519999448442 s 1.30
scatter_sum / HLOOpt / cpu / PostRev 0.000017 s 0.000014698720024171053 s 1.16
scatter_sum / HLOOpt / cpu / BothRev 0.000017 s 0.000013126020039635478 s 1.30
scatter_sum / PartOpt / cpu / PreRev 0.000016 s 0.000013033599962000154 s 1.23
scatter_sum / PartOpt / cpu / PostRev 0.000017 s 0.000012915639981656568 s 1.32
scatter_sum / PartOpt / cpu / BothRev 0.000017 s 0.000012708979975286638 s 1.34
scatter_sum / IPartOpt / cpu / PreRev 0.000016 s 0.000012781279965565772 s 1.25
scatter_sum / IPartOpt / cpu / PostRev 0.000017999999999999997 s 0.000013472300042849384 s 1.34
scatter_sum / IPartOpt / cpu / BothRev 0.000017 s 0.000012834419976570643 s 1.32
scatter_sum / DefOpt / cpu / PreRev 0.000017 s 0.000013145060020178787 s 1.29
scatter_sum / DefOpt / cpu / PostRev 0.000017 s 0.00001275038002859219 s 1.33
scatter_sum / DefOpt / cpu / BothRev 0.000017 s 0.000012998379997952724 s 1.31
scatter_sum / IDefOpt / cpu / PreRev 0.000017 s 0.000013091199998598312 s 1.30
scatter_sum / IDefOpt / cpu / PostRev 0.000017 s 0.00001284074000068358 s 1.32
scatter_sum / IDefOpt / cpu / BothRev 0.000017 s 0.000012844779957958964 s 1.32
slicing / JaXPipe / cpu / Primal 0.000006097719990521 s 0.000007288319993676851 s 0.84
slicing / Jax / cpu / Primal 0.0000060572400047931295 s 0.000006467119974331581 s 0.94
slicing / HLOOpt / cpu / Primal 0.000006739899999956833 s 0.000007340799984376645 s 0.92
slicing / PartOpt / cpu / Primal 0.0000061114599975553575 s 0.000006723080005031079 s 0.91
slicing / IPartOpt / cpu / Primal 0.000006765560005987936 s 0.00000703373997566814 s 0.96
slicing / DefOpt / cpu / Primal 0.000006124279998402926 s 0.000006649759980064118 s 0.92
slicing / IDefOpt / cpu / Primal 0.000006104960007178306 s 0.000006718660024489509 s 0.91
slicing / JaXPipe / cpu / Forward 0.000009634919995278325 s 0.000010664820019883336 s 0.90
slicing / Jax / cpu / Forward 0.000008973759995569709 s 0.00001085290004084527 s 0.83
slicing / HLOOpt / cpu / Forward 0.000009343960002752282 s 0.000010787459987113834 s 0.87
slicing / PartOpt / cpu / Forward 0.000009704120000151306 s 0.000009920160018737078 s 0.98
slicing / IPartOpt / cpu / Forward 0.000009749359990109953 s 0.00001104420003684936 s 0.88
slicing / DefOpt / cpu / Forward 0.000009281860002374742 s 0.00001050316004693741 s 0.88
slicing / IDefOpt / cpu / Forward 0.00000926456001252518 s 0.00001045673998305574 s 0.89
slicing / JaXPipe / cpu / PreRev 0.00001004898000473986 s 0.00001086409999516036 s 0.92
slicing / JaXPipe / cpu / PostRev 0.000010457239995957937 s 0.000011132019972137642 s 0.94
slicing / JaXPipe / cpu / BothRev 0.00000995396000917026 s 0.000010977259998981026 s 0.91
slicing / Jax / cpu / BothRev 0.000009851759987213882 s 0.00001058601999829989 s 0.93
slicing / HLOOpt / cpu / PreRev 0.000010423400010495242 s 0.000011480219991426563 s 0.91
slicing / HLOOpt / cpu / PostRev 0.000011904919990683993 s 0.000013056880015938076 s 0.91
slicing / HLOOpt / cpu / BothRev 0.00001018004000570727 s 0.000011306099986541084 s 0.90
slicing / PartOpt / cpu / PreRev 0.00000960710000299514 s 0.000011312520027786376 s 0.85
slicing / PartOpt / cpu / PostRev 0.000009860620002655196 s 0.000010987319992636912 s 0.90
slicing / PartOpt / cpu / BothRev 0.00001033418000133679 s 0.000011153960022056708 s 0.93
slicing / IPartOpt / cpu / PreRev 0.00001048506000870475 s 0.000010705020013119791 s 0.98
slicing / IPartOpt / cpu / PostRev 0.00000961794000431837 s 0.000010705700015023467 s 0.90
slicing / IPartOpt / cpu / BothRev 0.000010177359995395818 s 0.000010749940001915092 s 0.95
slicing / DefOpt / cpu / PreRev 0.000010347780003030494 s 0.000011468539933048303 s 0.90
slicing / DefOpt / cpu / PostRev 0.000009722579982280876 s 0.000010730119965955964 s 0.91
slicing / DefOpt / cpu / BothRev 0.000009908540002925292 s 0.000011325499990562091 s 0.87
slicing / IDefOpt / cpu / PreRev 0.000010187980005866848 s 0.000010818559994731911 s 0.94
slicing / IDefOpt / cpu / PostRev 0.000010396719994787418 s 0.000011040860026696464 s 0.94
slicing / IDefOpt / cpu / BothRev 0.000010105839994594134 s 0.00001058544001352857 s 0.95
slicing / JaXPipe / cuda / Primal 0.000002271 s 0.000001887 s 1.20
slicing / Jax / cuda / Primal 0.000002271 s 0.000001887 s 1.20
slicing / HLOOpt / cuda / Primal 0.000002271 s 0.000001887 s 1.20
slicing / PartOpt / cuda / Primal 0.000002271 s 0.000001887 s 1.20
slicing / IPartOpt / cuda / Primal 0.000002271 s 0.000001887 s 1.20
slicing / DefOpt / cuda / Primal 0.000002271 s 0.000001887 s 1.20
slicing / IDefOpt / cuda / Primal 0.000002271 s 0.000001887 s 1.20
slicing / JaXPipe / cuda / Forward 0.0000104 s 0.000010112 s 1.03
slicing / Jax / cuda / Forward 0.000010496 s 0.000009664 s 1.09
slicing / HLOOpt / cuda / Forward 0.00001008 s 0.000009728 s 1.04
slicing / PartOpt / cuda / Forward 0.000010144 s 0.000009312000000000002 s 1.09
slicing / IPartOpt / cuda / Forward 0.000010528 s 0.000009792 s 1.08
slicing / DefOpt / cuda / Forward 0.000010528 s 0.000010145 s 1.04
slicing / IDefOpt / cuda / Forward 0.000010752 s 0.000010048 s 1.07
slicing / JaXPipe / cuda / PreRev 0.0000104 s 0.000011232 s 0.93
slicing / JaXPipe / cuda / PostRev 0.000010239 s 0.000010976 s 0.93
slicing / JaXPipe / cuda / BothRev 0.00001024 s 0.00001008 s 1.02
slicing / Jax / cuda / BothRev 0.000010431 s 0.000009952 s 1.05
slicing / HLOOpt / cuda / PreRev 0.000010624 s 0.000009696 s 1.10
slicing / HLOOpt / cuda / PostRev 0.00001008 s 0.000009568 s 1.05
slicing / HLOOpt / cuda / BothRev 0.000010496 s 0.000009888 s 1.06
slicing / PartOpt / cuda / PreRev 0.000010432 s 0.0000096 s 1.09
slicing / PartOpt / cuda / PostRev 0.000010112 s 0.000009728 s 1.04
slicing / PartOpt / cuda / BothRev 0.000010336 s 0.000009983 s 1.04
slicing / IPartOpt / cuda / PreRev 0.000010465 s 0.000011137 s 0.94
slicing / IPartOpt / cuda / PostRev 0.000010176 s 0.000009568 s 1.06
slicing / IPartOpt / cuda / BothRev 0.000010624 s 0.000009953 s 1.07
slicing / DefOpt / cuda / PreRev 0.000010368 s 0.00000976 s 1.06
slicing / DefOpt / cuda / PostRev 0.000010304 s 0.000009633 s 1.07
slicing / DefOpt / cuda / BothRev 0.000010176 s 0.000009824 s 1.04
slicing / IDefOpt / cuda / PreRev 0.000010496 s 0.000009569 s 1.10
slicing / IDefOpt / cuda / PostRev 0.00001024 s 0.000009856 s 1.04
slicing / IDefOpt / cuda / BothRev 0.000010368 s 0.000009568 s 1.08
slicing / JaXPipe / tpu / Primal 0.0000010238 s 9.63975e-7 s 1.06
slicing / Jax / tpu / Primal 9.70575e-7 s 9.6105e-7 s 1.01
slicing / HLOOpt / tpu / Primal 0.00000102085 s 9.61225e-7 s 1.06
slicing / PartOpt / tpu / Primal 9.708e-7 s 9.65325e-7 s 1.01
slicing / IPartOpt / tpu / Primal 0.0000010202 s 9.59025e-7 s 1.06
slicing / DefOpt / tpu / Primal 9.667e-7 s 9.6145e-7 s 1.01
slicing / IDefOpt / tpu / Primal 0.0000010205 s 9.6245e-7 s 1.06
slicing / JaXPipe / tpu / Forward 0.0000014079 s 0.000001406125 s 1.00
slicing / Jax / tpu / Forward 0.000001476175 s 0.000001412075 s 1.05
slicing / HLOOpt / tpu / Forward 0.000001523725 s 0.00000151405 s 1.01
slicing / PartOpt / tpu / Forward 0.0000014935249999999998 s 0.0000014341999999999998 s 1.04
slicing / IPartOpt / tpu / Forward 0.00000151945 s 0.000001513225 s 1.00
slicing / DefOpt / tpu / Forward 0.000001493175 s 0.000001437075 s 1.04
slicing / IDefOpt / tpu / Forward 0.00000152025 s 0.00000151655 s 1.00
slicing / JaXPipe / tpu / PreRev 0.0000025657249999999995 s 0.00000238275 s 1.08
slicing / JaXPipe / tpu / PostRev 0.000002519 s 0.0000025262 s 1.00
slicing / JaXPipe / tpu / BothRev 0.000002586325 s 0.0000023967250000000003 s 1.08
slicing / Jax / tpu / BothRev 0.000002534525 s 0.000002552125 s 0.99
slicing / HLOOpt / tpu / PreRev 0.000002583475 s 0.00000239915 s 1.08
slicing / HLOOpt / tpu / PostRev 0.0000025418000000000004 s 0.000002548575 s 1.00
slicing / HLOOpt / tpu / BothRev 0.00000259175 s 0.0000024025000000000003 s 1.08
slicing / PartOpt / tpu / PreRev 0.00000253425 s 0.00000254185 s 1.00
slicing / PartOpt / tpu / PostRev 0.000002579575 s 0.0000024003 s 1.07
slicing / PartOpt / tpu / BothRev 0.000002551225 s 0.0000025502 s 1.00
slicing / IPartOpt / tpu / PreRev 0.0000025931749999999995 s 0.0000023919000000000003 s 1.08
slicing / IPartOpt / tpu / PostRev 0.0000025407000000000005 s 0.0000025419750000000003 s 1.00
slicing / IPartOpt / tpu / BothRev 0.0000025882 s 0.000002392475 s 1.08
slicing / DefOpt / tpu / PreRev 0.00000253735 s 0.0000025409 s 1.00
slicing / DefOpt / tpu / PostRev 0.0000025874 s 0.00000239555 s 1.08
slicing / DefOpt / tpu / BothRev 0.0000025396 s 0.0000025384750000000003 s 1.00
slicing / IDefOpt / tpu / PreRev 0.000002587025 s 0.0000023991 s 1.08
slicing / IDefOpt / tpu / PostRev 0.0000025399 s 0.000002548875 s 1.00
slicing / IDefOpt / tpu / BothRev 0.000002591725 s 0.00000240065 s 1.08
slicing / JaXPipe / cpu / Primal 0.000012678 s 0.000007288319993676851 s 1.74
slicing / Jax / cpu / Primal 0.000012606 s 0.000006467119974331581 s 1.95
slicing / HLOOpt / cpu / Primal 0.000012675 s 0.000007340799984376645 s 1.73
slicing / PartOpt / cpu / Primal 0.000012441 s 0.000006723080005031079 s 1.85
slicing / IPartOpt / cpu / Primal 0.000012469 s 0.00000703373997566814 s 1.77
slicing / DefOpt / cpu / Primal 0.000012446 s 0.000006649759980064118 s 1.87
slicing / IDefOpt / cpu / Primal 0.000012437 s 0.000006718660024489509 s 1.85
slicing / JaXPipe / cpu / Forward 0.000017205999999999998 s 0.000010664820019883336 s 1.61
slicing / Jax / cpu / Forward 0.000016621 s 0.00001085290004084527 s 1.53
slicing / HLOOpt / cpu / Forward 0.000016353 s 0.000010787459987113834 s 1.52
slicing / PartOpt / cpu / Forward 0.000016637 s 0.000009920160018737078 s 1.68
slicing / IPartOpt / cpu / Forward 0.000017094999999999998 s 0.00001104420003684936 s 1.55
slicing / DefOpt / cpu / Forward 0.00001681 s 0.00001050316004693741 s 1.60
slicing / IDefOpt / cpu / Forward 0.000016638 s 0.00001045673998305574 s 1.59
slicing / JaXPipe / cpu / PreRev 0.000017786 s 0.00001086409999516036 s 1.64
slicing / JaXPipe / cpu / PostRev 0.000017364000000000002 s 0.000011132019972137642 s 1.56
slicing / JaXPipe / cpu / BothRev 0.00001788 s 0.000010977259998981026 s 1.63
slicing / Jax / cpu / BothRev 0.000017517999999999997 s 0.00001058601999829989 s 1.65
slicing / HLOOpt / cpu / PreRev 0.000017375999999999998 s 0.000011480219991426563 s 1.51
slicing / HLOOpt / cpu / PostRev 0.000017748 s 0.000013056880015938076 s 1.36
slicing / HLOOpt / cpu / BothRev 0.000017328 s 0.000011306099986541084 s 1.53
slicing / PartOpt / cpu / PreRev 0.000017893 s 0.000011312520027786376 s 1.58
slicing / PartOpt / cpu / PostRev 0.000017467 s 0.000010987319992636912 s 1.59
slicing / PartOpt / cpu / BothRev 0.000017134 s 0.000011153960022056708 s 1.54
slicing / IPartOpt / cpu / PreRev 0.000017207 s 0.000010705020013119791 s 1.61
slicing / IPartOpt / cpu / PostRev 0.000017069000000000002 s 0.000010705700015023467 s 1.59
slicing / IPartOpt / cpu / BothRev 0.000017336 s 0.000010749940001915092 s 1.61
slicing / DefOpt / cpu / PreRev 0.000017749999999999998 s 0.000011468539933048303 s 1.55
slicing / DefOpt / cpu / PostRev 0.000016863 s 0.000010730119965955964 s 1.57
slicing / DefOpt / cpu / BothRev 0.000017564 s 0.000011325499990562091 s 1.55
slicing / IDefOpt / cpu / PreRev 0.000016996 s 0.000010818559994731911 s 1.57
slicing / IDefOpt / cpu / PostRev 0.000016916000000000002 s 0.000011040860026696464 s 1.53
slicing / IDefOpt / cpu / BothRev 0.000016876 s 0.00001058544001352857 s 1.59
slicing / JaXPipe / cpu / Primal 0.000008 s 0.000007288319993676851 s 1.10
slicing / Jax / cpu / Primal 0.000008999999999999999 s 0.000006467119974331581 s 1.39
slicing / HLOOpt / cpu / Primal 0.000008999999999999999 s 0.000007340799984376645 s 1.23
slicing / PartOpt / cpu / Primal 0.000008999999999999999 s 0.000006723080005031079 s 1.34
slicing / IPartOpt / cpu / Primal 0.000008999999999999999 s 0.00000703373997566814 s 1.28
slicing / DefOpt / cpu / Primal 0.000008999999999999999 s 0.000006649759980064118 s 1.35
slicing / IDefOpt / cpu / Primal 0.000008999999999999999 s 0.000006718660024489509 s 1.34
slicing / JaXPipe / cpu / Forward 0.000012 s 0.000010664820019883336 s 1.13
slicing / Jax / cpu / Forward 0.000012 s 0.00001085290004084527 s 1.11
slicing / HLOOpt / cpu / Forward 0.000012 s 0.000010787459987113834 s 1.11
slicing / PartOpt / cpu / Forward 0.000012 s 0.000009920160018737078 s 1.21
slicing / IPartOpt / cpu / Forward 0.000012 s 0.00001104420003684936 s 1.09
slicing / DefOpt / cpu / Forward 0.000013 s 0.00001050316004693741 s 1.24
slicing / IDefOpt / cpu / Forward 0.000012 s 0.00001045673998305574 s 1.15
slicing / JaXPipe / cpu / PreRev 0.000012 s 0.00001086409999516036 s 1.10
slicing / JaXPipe / cpu / PostRev 0.000013 s 0.000011132019972137642 s 1.17
slicing / JaXPipe / cpu / BothRev 0.000012 s 0.000010977259998981026 s 1.09
slicing / Jax / cpu / BothRev 0.000013 s 0.00001058601999829989 s 1.23
slicing / HLOOpt / cpu / PreRev 0.000012 s 0.000011480219991426563 s 1.05
slicing / HLOOpt / cpu / PostRev 0.000013 s 0.000013056880015938076 s 1.00
slicing / HLOOpt / cpu / BothRev 0.000012 s 0.000011306099986541084 s 1.06
slicing / PartOpt / cpu / PreRev 0.000012 s 0.000011312520027786376 s 1.06
slicing / PartOpt / cpu / PostRev 0.000012 s 0.000010987319992636912 s 1.09
slicing / PartOpt / cpu / BothRev 0.000013 s 0.000011153960022056708 s 1.17
slicing / IPartOpt / cpu / PreRev 0.000012 s 0.000010705020013119791 s 1.12
slicing / IPartOpt / cpu / PostRev 0.000012 s 0.000010705700015023467 s 1.12
slicing / IPartOpt / cpu / BothRev 0.000012 s 0.000010749940001915092 s 1.12
slicing / DefOpt / cpu / PreRev 0.000012 s 0.000011468539933048303 s 1.05
slicing / DefOpt / cpu / PostRev 0.000013 s 0.000010730119965955964 s 1.21
slicing / DefOpt / cpu / BothRev 0.000013 s 0.000011325499990562091 s 1.15
slicing / IDefOpt / cpu / PreRev 0.000012 s 0.000010818559994731911 s 1.11
slicing / IDefOpt / cpu / PostRev 0.000013 s 0.000011040860026696464 s 1.18
slicing / IDefOpt / cpu / BothRev 0.000012 s 0.00001058544001352857 s 1.13
sum / JaXPipe / cpu / Primal 0.000007667779993880685 s 0.000008742739983063074 s 0.88
sum / Jax / cpu / Primal 0.000007404020004742052 s 0.000008432739987256355 s 0.88
sum / HLOOpt / cpu / Primal 0.000007994480006345838 s 0.00000868413997523021 s 0.92
sum / PartOpt / cpu / Primal 0.00000745573999665794 s 0.000008593839993409347 s 0.87
sum / IPartOpt / cpu / Primal 0.000008130819999223604 s 0.000008873579981809598 s 0.92
sum / DefOpt / cpu / Primal 0.000007747900001504604 s 0.000008341580014530337 s 0.93
sum / IDefOpt / cpu / Primal 0.000008481119998577924 s 0.000008160259958458483 s 1.04
sum / JaXPipe / cpu / Forward 0.00001121236000699355 s 0.000012359000020296664 s 0.91
sum / Jax / cpu / Forward 0.00001148140000168496 s 0.000012516539954958715 s 0.92
sum / HLOOpt / cpu / Forward 0.00001145546000543618 s 0.000012630199980776524 s 0.91
sum / PartOpt / cpu / Forward 0.000011307780007427936 s 0.000012017819990433054 s 0.94
sum / IPartOpt / cpu / Forward 0.000011516060003486928 s 0.0000124326400236896 s 0.93
sum / DefOpt / cpu / Forward 0.000010961339992263674 s 0.000012678339962803876 s 0.86
sum / IDefOpt / cpu / Forward 0.000011057339995659276 s 0.00001217218003148446 s 0.91
sum / JaXPipe / cpu / PreRev 0.000011736080002719971 s 0.000011711759980244096 s 1.00
sum / JaXPipe / cpu / PostRev 0.000010845479998806697 s 0.0000120298000274488 s 0.90
sum / JaXPipe / cpu / BothRev 0.000010823559985055908 s 0.000012270700017324998 s 0.88
sum / Jax / cpu / BothRev 0.000011136799996620538 s 0.000012310340025578626 s 0.90
sum / HLOOpt / cpu / PreRev 0.000011556560000371974 s 0.000012216640025144444 s 0.95
sum / HLOOpt / cpu / PostRev 0.000012639439980830502 s 0.000013434619959298288 s 0.94
sum / HLOOpt / cpu / BothRev 0.000010607400010940182 s 0.000012282459956622916 s 0.86
sum / PartOpt / cpu / PreRev 0.00001056131999803256 s 0.000012226860008013318 s 0.86
sum / PartOpt / cpu / PostRev 0.000011329480000767945 s 0.000011789480022343924 s 0.96
sum / PartOpt / cpu / BothRev 0.0000106691000041792 s 0.000012224460024299332 s 0.87
sum / IPartOpt / cpu / PreRev 0.00001063957999804188 s 0.000011990420025540516 s 0.89
sum / IPartOpt / cpu / PostRev 0.000011201639999853796 s 0.000012199719985801494 s 0.92
sum / IPartOpt / cpu / BothRev 0.00001108305999878212 s 0.000011894060035047004 s 0.93
sum / DefOpt / cpu / PreRev 0.00001142704000358208 s 0.000012119699986214985 s 0.94
sum / DefOpt / cpu / PostRev 0.000010642900003858812 s 0.00001206306002131896 s 0.88
sum / DefOpt / cpu / BothRev 0.00001090210000256775 s 0.00001233915999364399 s 0.88
sum / IDefOpt / cpu / PreRev 0.000011338460003571529 s 0.000011822360029327683 s 0.96
sum / IDefOpt / cpu / PostRev 0.000010959799999454844 s 0.00001169123997897259 s 0.94
sum / IDefOpt / cpu / BothRev 0.000010769340003662364 s 0.000011358259980625007 s 0.95
sum / JaXPipe / cuda / Primal 0.000002463 s 0.000002047 s 1.20
sum / Jax / cuda / Primal 0.000002463 s 0.000002047 s 1.20
sum / HLOOpt / cuda / Primal 0.000002463 s 0.000002047 s 1.20
sum / PartOpt / cuda / Primal 0.000002463 s 0.000002047 s 1.20
sum / IPartOpt / cuda / Primal 0.000002463 s 0.000002047 s 1.20
sum / DefOpt / cuda / Primal 0.000002432 s 0.000002048 s 1.19
sum / IDefOpt / cuda / Primal 0.000002463 s 0.000002047 s 1.20
sum / JaXPipe / cuda / Forward 0.000010528 s 0.000010303 s 1.02
sum / Jax / cuda / Forward 0.000011008 s 0.000009952 s 1.11
sum / HLOOpt / cuda / Forward 0.000010656 s 0.000009728 s 1.10
sum / PartOpt / cuda / Forward 0.000010848 s 0.000009984 s 1.09
sum / IPartOpt / cuda / Forward 0.000010752 s 0.000009536 s 1.13
sum / DefOpt / cuda / Forward 0.000010719 s 0.000009824 s 1.09
sum / IDefOpt / cuda / Forward 0.0000112 s 0.00001008 s 1.11
sum / JaXPipe / cuda / PreRev 0.000010144 s 0.000009632 s 1.05
sum / JaXPipe / cuda / PostRev 0.000010143 s 0.000009951 s 1.02
sum / JaXPipe / cuda / BothRev 0.000010336 s 0.000009792 s 1.06
sum / Jax / cuda / BothRev 0.000010272 s 0.000010144 s 1.01
sum / HLOOpt / cuda / PreRev 0.000010656 s 0.000009696 s 1.10
sum / HLOOpt / cuda / PostRev 0.000009952 s 0.00000944 s 1.05
sum / HLOOpt / cuda / BothRev 0.000010113 s 0.00000976 s 1.04
sum / PartOpt / cuda / PreRev 0.000010496 s 0.000009824 s 1.07
sum / PartOpt / cuda / PostRev 0.000010017 s 0.00001024 s 0.98
sum / PartOpt / cuda / BothRev 0.000010111 s 0.000009696 s 1.04
sum / IPartOpt / cuda / PreRev 0.000012704 s 0.000009824 s 1.29
sum / IPartOpt / cuda / PostRev 0.000010176 s 0.000009984 s 1.02
sum / IPartOpt / cuda / BothRev 0.000010112 s 0.000010047 s 1.01
sum / DefOpt / cuda / PreRev 0.000010304 s 0.000010144 s 1.02
sum / DefOpt / cuda / PostRev 0.000010304 s 0.000009985 s 1.03
sum / DefOpt / cuda / BothRev 0.000010848 s 0.00000992 s 1.09
sum / IDefOpt / cuda / PreRev 0.00001088 s 0.000009824 s 1.11
sum / IDefOpt / cuda / PostRev 0.000010175 s 0.00000976 s 1.04
sum / IDefOpt / cuda / BothRev 0.000010656 s 0.000009856 s 1.08
sum / JaXPipe / tpu / Primal 5.104499999999999e-7 s 5.1025e-7 s 1.00
sum / Jax / tpu / Primal 5.473999999999999e-7 s 5.4695e-7 s 1.00
sum / HLOOpt / tpu / Primal 5.10725e-7 s 5.10625e-7 s 1.00
sum / PartOpt / tpu / Primal 5.46975e-7 s 5.471999999999999e-7 s 1.00
sum / IPartOpt / tpu / Primal 5.10175e-7 s 5.102750000000001e-7 s 1.00
sum / DefOpt / tpu / Primal 5.4705e-7 s 5.470250000000001e-7 s 1.00
sum / IDefOpt / tpu / Primal 5.10375e-7 s 5.104e-7 s 1.00
sum / JaXPipe / tpu / Forward 0.000001548525 s 0.000001546225 s 1.00
sum / Jax / tpu / Forward 0.000001498975 s 0.000001498375 s 1.00
sum / HLOOpt / tpu / Forward 0.0000015294749999999997 s 0.000001529575 s 1.00
sum / PartOpt / tpu / Forward 0.00000149805 s 0.000001495975 s 1.00
sum / IPartOpt / tpu / Forward 0.000001543575 s 0.0000015346499999999998 s 1.01
sum / DefOpt / tpu / Forward 0.00000149725 s 0.0000014994 s 1.00
sum / IDefOpt / tpu / Forward 0.0000015303249999999998 s 0.000001527925 s 1.00
sum / JaXPipe / tpu / PreRev 0.0000010508000000000002 s 0.0000010069500000000002 s 1.04
sum / JaXPipe / tpu / PostRev 0.000001090075 s 0.00000103485 s 1.05
sum / JaXPipe / tpu / BothRev 0.000001049025 s 0.000001003175 s 1.05
sum / Jax / tpu / BothRev 0.000001087375 s 0.000001038925 s 1.05
sum / HLOOpt / tpu / PreRev 0.000001047625 s 0.0000010023 s 1.05
sum / HLOOpt / tpu / PostRev 0.000001087525 s 0.000001040175 s 1.05
sum / HLOOpt / tpu / BothRev 0.000001051725 s 0.0000010055000000000002 s 1.05
sum / PartOpt / tpu / PreRev 0.000001090325 s 0.0000010352 s 1.05
sum / PartOpt / tpu / PostRev 0.0000010508000000000002 s 9.97725e-7 s 1.05
sum / PartOpt / tpu / BothRev 0.0000010896250000000002 s 0.000001040425 s 1.05
sum / IPartOpt / tpu / PreRev 0.000001057475 s 0.000001005175 s 1.05
sum / IPartOpt / tpu / PostRev 0.00000109585 s 0.00000103545 s 1.06
sum / IPartOpt / tpu / BothRev 0.00000105905 s 0.0000010101 s 1.05
sum / DefOpt / tpu / PreRev 0.0000011000749999999998 s 0.000001034625 s 1.06
sum / DefOpt / tpu / PostRev 0.0000010545749999999998 s 9.9925e-7 s 1.06
sum / DefOpt / tpu / BothRev 0.000001083125 s 0.000001041325 s 1.04
sum / IDefOpt / tpu / PreRev 0.00000104795 s 0.000001001975 s 1.05
sum / IDefOpt / tpu / PostRev 0.0000010861 s 0.0000010411 s 1.04
sum / IDefOpt / tpu / BothRev 0.000001046525 s 0.00000100035 s 1.05
sum / JaXPipe / cpu / Primal 0.000014868 s 0.000008742739983063074 s 1.70
sum / Jax / cpu / Primal 0.00001457 s 0.000008432739987256355 s 1.73
sum / HLOOpt / cpu / Primal 0.000014701 s 0.00000868413997523021 s 1.69
sum / PartOpt / cpu / Primal 0.000014327 s 0.000008593839993409347 s 1.67
sum / IPartOpt / cpu / Primal 0.000014848 s 0.000008873579981809598 s 1.67
sum / DefOpt / cpu / Primal 0.000014684 s 0.000008341580014530337 s 1.76
sum / IDefOpt / cpu / Primal 0.000014412 s 0.000008160259958458483 s 1.77
sum / JaXPipe / cpu / Forward 0.00002015 s 0.000012359000020296664 s 1.63
sum / Jax / cpu / Forward 0.000020038 s 0.000012516539954958715 s 1.60
sum / HLOOpt / cpu / Forward 0.000020079 s 0.000012630199980776524 s 1.59
sum / PartOpt / cpu / Forward 0.000020276 s 0.000012017819990433054 s 1.69
sum / IPartOpt / cpu / Forward 0.000019812 s 0.0000124326400236896 s 1.59
sum / DefOpt / cpu / Forward 0.000019123 s 0.000012678339962803876 s 1.51
sum / IDefOpt / cpu / Forward 0.000019551 s 0.00001217218003148446 s 1.61
sum / JaXPipe / cpu / PreRev 0.000018633 s 0.000011711759980244096 s 1.59
sum / JaXPipe / cpu / PostRev 0.000018572 s 0.0000120298000274488 s 1.54
sum / JaXPipe / cpu / BothRev 0.000018265 s 0.000012270700017324998 s 1.49
sum / Jax / cpu / BothRev 0.000018417 s 0.000012310340025578626 s 1.50
sum / HLOOpt / cpu / PreRev 0.00001839 s 0.000012216640025144444 s 1.51
sum / HLOOpt / cpu / PostRev 0.000018787 s 0.000013434619959298288 s 1.40
sum / HLOOpt / cpu / BothRev 0.0000186 s 0.000012282459956622916 s 1.51
sum / PartOpt / cpu / PreRev 0.000018979 s 0.000012226860008013318 s 1.55
sum / PartOpt / cpu / PostRev 0.000018780000000000003 s 0.000011789480022343924 s 1.59
sum / PartOpt / cpu / BothRev 0.000018576 s 0.000012224460024299332 s 1.52
sum / IPartOpt / cpu / PreRev 0.000018944 s 0.000011990420025540516 s 1.58
sum / IPartOpt / cpu / PostRev 0.000018881 s 0.000012199719985801494 s 1.55
sum / IPartOpt / cpu / BothRev 0.000019244 s 0.000011894060035047004 s 1.62
sum / DefOpt / cpu / PreRev 0.000018787 s 0.000012119699986214985 s 1.55
sum / DefOpt / cpu / PostRev 0.000018475 s 0.00001206306002131896 s 1.53
sum / DefOpt / cpu / BothRev 0.000018859 s 0.00001233915999364399 s 1.53
sum / IDefOpt / cpu / PreRev 0.000018386 s 0.000011822360029327683 s 1.56
sum / IDefOpt / cpu / PostRev 0.000019177 s 0.00001169123997897259 s 1.64
sum / IDefOpt / cpu / BothRev 0.000019643 s 0.000011358259980625007 s 1.73
sum / JaXPipe / cpu / Primal 0.00001 s 0.000008742739983063074 s 1.14
sum / Jax / cpu / Primal 0.00001 s 0.000008432739987256355 s 1.19
sum / HLOOpt / cpu / Primal 0.000011 s 0.00000868413997523021 s 1.27
sum / PartOpt / cpu / Primal 0.000011 s 0.000008593839993409347 s 1.28
sum / IPartOpt / cpu / Primal 0.00001 s 0.000008873579981809598 s 1.13
sum / DefOpt / cpu / Primal 0.00001 s 0.000008341580014530337 s 1.20
sum / IDefOpt / cpu / Primal 0.000011 s 0.000008160259958458483 s 1.35
sum / JaXPipe / cpu / Forward 0.000014 s 0.000012359000020296664 s 1.13
sum / Jax / cpu / Forward 0.000015 s 0.000012516539954958715 s 1.20
sum / HLOOpt / cpu / Forward 0.000015 s 0.000012630199980776524 s 1.19
sum / PartOpt / cpu / Forward 0.000015 s 0.000012017819990433054 s 1.25
sum / IPartOpt / cpu / Forward 0.000015 s 0.0000124326400236896 s 1.21
sum / DefOpt / cpu / Forward 0.000014 s 0.000012678339962803876 s 1.10
sum / IDefOpt / cpu / Forward 0.000015 s 0.00001217218003148446 s 1.23
sum / JaXPipe / cpu / PreRev 0.000014 s 0.000011711759980244096 s 1.20
sum / JaXPipe / cpu / PostRev 0.000014 s 0.0000120298000274488 s 1.16
sum / JaXPipe / cpu / BothRev 0.000014 s 0.000012270700017324998 s 1.14
sum / Jax / cpu / BothRev 0.000013 s 0.000012310340025578626 s 1.06
sum / HLOOpt / cpu / PreRev 0.000014 s 0.000012216640025144444 s 1.15
sum / HLOOpt / cpu / PostRev 0.000014 s 0.000013434619959298288 s 1.04
sum / HLOOpt / cpu / BothRev 0.000014 s 0.000012282459956622916 s 1.14
sum / PartOpt / cpu / PreRev 0.000014 s 0.000012226860008013318 s 1.15
sum / PartOpt / cpu / PostRev 0.000014 s 0.000011789480022343924 s 1.19
sum / PartOpt / cpu / BothRev 0.000014 s 0.000012224460024299332 s 1.15
sum / IPartOpt / cpu / PreRev 0.000014 s 0.000011990420025540516 s 1.17
sum / IPartOpt / cpu / PostRev 0.000014 s 0.000012199719985801494 s 1.15
sum / IPartOpt / cpu / BothRev 0.000014 s 0.000011894060035047004 s 1.18
sum / DefOpt / cpu / PreRev 0.000014 s 0.000012119699986214985 s 1.16
sum / DefOpt / cpu / PostRev 0.000013 s 0.00001206306002131896 s 1.08
sum / DefOpt / cpu / BothRev 0.000013 s 0.00001233915999364399 s 1.05
sum / IDefOpt / cpu / PreRev 0.000014 s 0.000011822360029327683 s 1.18
sum / IDefOpt / cpu / PostRev 0.000014 s 0.00001169123997897259 s 1.20
sum / IDefOpt / cpu / BothRev 0.000014 s 0.000011358259980625007 s 1.23
value_and_grad / JaXPipe / cpu / Primal 0.000014051159992050088 s 0.000015538480001850986 s 0.90
value_and_grad / Jax / cpu / Primal 0.0000131080800088057 s 0.000015070379977260016 s 0.87
value_and_grad / HLOOpt / cpu / Primal 0.000013616859996545827 s 0.000014645099963672691 s 0.93
value_and_grad / PartOpt / cpu / Primal 0.000012969880003765866 s 0.00001437700002497877 s 0.90
value_and_grad / IPartOpt / cpu / Primal 0.000013442959998428704 s 0.000014697280030304682 s 0.91
value_and_grad / DefOpt / cpu / Primal 0.00001310192000346433 s 0.00001450112002203241 s 0.90
value_and_grad / IDefOpt / cpu / Primal 0.000013357660013753048 s 0.000014610480011469918 s 0.91
value_and_grad / JaXPipe / cuda / Primal 0.000033792000000000004 s 0.000032352 s 1.04
value_and_grad / Jax / cuda / Primal 0.000032864 s 0.000032896000000000005 s 1.00
value_and_grad / HLOOpt / cuda / Primal 0.000032800000000000004 s 0.000032225 s 1.02
value_and_grad / PartOpt / cuda / Primal 0.000033632 s 0.000032385 s 1.04
value_and_grad / IPartOpt / cuda / Primal 0.000033216 s 0.000032672 s 1.02
value_and_grad / DefOpt / cuda / Primal 0.000032864 s 0.000032608 s 1.01
value_and_grad / IDefOpt / cuda / Primal 0.000033632 s 0.00003264 s 1.03
value_and_grad / JaXPipe / tpu / Primal 0 s 0 s 1
value_and_grad / Jax / tpu / Primal 0 s 0 s 1
value_and_grad / HLOOpt / tpu / Primal 0 s 0 s 1
value_and_grad / PartOpt / tpu / Primal 0 s 0 s 1
value_and_grad / IPartOpt / tpu / Primal 0 s 0 s 1
value_and_grad / DefOpt / tpu / Primal 0 s 0 s 1
value_and_grad / IDefOpt / tpu / Primal 0 s 0 s 1
value_and_grad / JaXPipe / cpu / Primal 0.000023356 s 0.000015538480001850986 s 1.50
value_and_grad / Jax / cpu / Primal 0.000022972 s 0.000015070379977260016 s 1.52
value_and_grad / HLOOpt / cpu / Primal 0.000022655 s 0.000014645099963672691 s 1.55
value_and_grad / PartOpt / cpu / Primal 0.000022765 s 0.00001437700002497877 s 1.58
value_and_grad / IPartOpt / cpu / Primal 0.000022622 s 0.000014697280030304682 s 1.54
value_and_grad / DefOpt / cpu / Primal 0.000022905 s 0.00001450112002203241 s 1.58
value_and_grad / IDefOpt / cpu / Primal 0.000022756 s 0.000014610480011469918 s 1.56
value_and_grad / JaXPipe / cpu / Primal 0.000017 s 0.000015538480001850986 s 1.09
value_and_grad / Jax / cpu / Primal 0.000017 s 0.000015070379977260016 s 1.13
value_and_grad / HLOOpt / cpu / Primal 0.000017 s 0.000014645099963672691 s 1.16
value_and_grad / PartOpt / cpu / Primal 0.000016 s 0.00001437700002497877 s 1.11
value_and_grad / IPartOpt / cpu / Primal 0.000017 s 0.000014697280030304682 s 1.16
value_and_grad / DefOpt / cpu / Primal 0.000016 s 0.00001450112002203241 s 1.10
value_and_grad / IDefOpt / cpu / Primal 0.000017 s 0.000014610480011469918 s 1.16
jaxmd20 / JaXPipe / cuda / Primal 0.001414361 s 0.001938565 s 0.73
jaxmd20 / Jax / cuda / Primal 0.0014725379999999 s 0.001496676 s 0.98
jaxmd20 / HLOOpt / cuda / Primal 0.001638937 s 0.0013105629999999 s 1.25
jaxmd20 / PartOpt / cuda / Primal 0.0013497539999999 s 0.001318019 s 1.02
jaxmd20 / IPartOpt / cuda / Primal 0.001348603 s 0.001323556 s 1.02
jaxmd20 / DefOpt / cuda / Primal 0.000938235 s 0.000916066 s 1.02
jaxmd20 / IDefOpt / cuda / Primal 0.000966332 s 0.000951714 s 1.02
jaxmd20 / JaXPipe / cuda / Forward 0.001620091 s 0.001571588 s 1.03
jaxmd20 / Jax / cuda / Forward 0.001836441 s 0.0017808059999999 s 1.03
jaxmd20 / HLOOpt / cuda / Forward 0.0017070339999999 s 0.001623371 s 1.05
jaxmd20 / PartOpt / cuda / Forward 0.001706677 s 0.001664068 s 1.03
jaxmd20 / IPartOpt / cuda / Forward 0.0017145539999999 s 0.001614468 s 1.06
jaxmd20 / DefOpt / cuda / Forward 0.001716986 s 0.001624164 s 1.06
jaxmd20 / IDefOpt / cuda / Forward 0.0017166969999999 s 0.0016119399999999 s 1.06
jaxmd20 / JaXPipe / cuda / PreRev 0.002760886 s 0.002690471 s 1.03
jaxmd20 / JaXPipe / cuda / PostRev 0.0054248329999999 s 0.005677523 s 0.96
jaxmd20 / JaXPipe / cuda / BothRev 0.002759573 s 0.002690601 s 1.03
jaxmd20 / Jax / cuda / BothRev 0.005481326 s 0.005316715 s 1.03
jaxmd20 / HLOOpt / cuda / PreRev 0.0028702299999999 s 0.002768517 s 1.04
jaxmd20 / HLOOpt / cuda / PostRev 0.005503594 s 0.005294383 s 1.04
jaxmd20 / HLOOpt / cuda / BothRev 0.002807093 s 0.002714063 s 1.03
jaxmd20 / PartOpt / cuda / PreRev 0.002877429 s 0.002801832 s 1.03
jaxmd20 / PartOpt / cuda / PostRev 0.005634188 s 0.005661502 s 1.00
jaxmd20 / PartOpt / cuda / BothRev 0.002826196 s 0.002745029 s 1.03
jaxmd20 / IPartOpt / cuda / PreRev 0.002889332 s 0.002794438 s 1.03
jaxmd20 / IPartOpt / cuda / PostRev 0.0055886509999999 s 0.0054231819999999 s 1.03
jaxmd20 / IPartOpt / cuda / BothRev 0.002815703 s 0.002751239 s 1.02
jaxmd20 / DefOpt / cuda / PreRev 0.002926036 s 0.002836042 s 1.03
jaxmd20 / DefOpt / cuda / PostRev 0.002827063 s 0.002758054 s 1.03
jaxmd20 / DefOpt / cuda / BothRev 0.002809589 s 0.0027710789999999 s 1.01
jaxmd20 / IDefOpt / cuda / PreRev 0.002926324 s 0.002802917 s 1.04
jaxmd20 / IDefOpt / cuda / PostRev 0.00236485 s 0.002298309 s 1.03
jaxmd20 / IDefOpt / cuda / BothRev 0.0028344269999999 s 0.0027711109999999 s 1.02
jaxmd20 / JaXPipe / tpu / Primal 0.00928150875 s 0.009277693125 s 1.00
jaxmd20 / Jax / tpu / Primal 0.009281055625 s 0.0092693106249999 s 1.00
jaxmd20 / HLOOpt / tpu / Primal 0.009154576875 s 0.009168335 s 1.00
jaxmd20 / PartOpt / tpu / Primal 0.0092015406249999 s 0.0092010406249999 s 1.00
jaxmd20 / IPartOpt / tpu / Primal 0.00920164125 s 0.009196949375 s 1.00
jaxmd20 / DefOpt / tpu / Primal 0.0088016868749999 s 0.008796551875 s 1.00
jaxmd20 / IDefOpt / tpu / Primal 0.00870373375 s 0.008697765625 s 1.00
jaxmd20 / JaXPipe / tpu / Forward 0.017415764375 s 0.017415084375 s 1.00
jaxmd20 / Jax / tpu / Forward 0.018736701875 s 0.018733859375 s 1.00
jaxmd20 / HLOOpt / tpu / Forward 0.01739822125 s 0.017391626875 s 1.00
jaxmd20 / PartOpt / tpu / Forward 0.017416905 s 0.017410939375 s 1.00
jaxmd20 / IPartOpt / tpu / Forward 0.01741094 s 0.017415836875 s 1.00
jaxmd20 / DefOpt / tpu / Forward 0.017426603125 s 0.017405771875 s 1.00
jaxmd20 / IDefOpt / tpu / Forward 0.01741189375 s 0.0174111487499999 s 1.00
jaxmd20 / JaXPipe / tpu / PreRev 0.025457630625 s 0.02547067625 s 1.00
jaxmd20 / JaXPipe / tpu / PostRev 0.02187902 s 0.021862679375 s 1.00
jaxmd20 / JaXPipe / tpu / BothRev 0.0254868381249999 s 0.025455455 s 1.00
jaxmd20 / Jax / tpu / BothRev 0.021871143125 s 0.021865015625 s 1.00
jaxmd20 / HLOOpt / tpu / PreRev 0.0255961368749999 s 0.02556869625 s 1.00
jaxmd20 / HLOOpt / tpu / PostRev 0.020469085 s 0.02080791125 s 0.98
jaxmd20 / HLOOpt / tpu / BothRev 0.02570186125 s 0.0256773987499999 s 1.00
jaxmd20 / PartOpt / tpu / PreRev 0.025487459375 s 0.025461623125 s 1.00
jaxmd20 / PartOpt / tpu / PostRev 0.02152813875 s 0.02152470625 s 1.00
jaxmd20 / PartOpt / tpu / BothRev 0.025567345 s 0.0255368143749999 s 1.00
jaxmd20 / IPartOpt / tpu / PreRev 0.025486080625 s 0.025459676875 s 1.00
jaxmd20 / IPartOpt / tpu / PostRev 0.021519329375 s 0.021515244375 s 1.00
jaxmd20 / IPartOpt / tpu / BothRev 0.025578833125 s 0.02555456625 s 1.00
jaxmd20 / DefOpt / tpu / PreRev 0.02548939125 s 0.02545990625 s 1.00
jaxmd20 / DefOpt / tpu / PostRev 0.018822576875 s 0.018808618125 s 1.00
jaxmd20 / DefOpt / tpu / BothRev 0.025574985 s 0.0255338281249999 s 1.00
jaxmd20 / IDefOpt / tpu / PreRev 0.0254887824999999 s 0.0254602499999999 s 1.00
jaxmd20 / IDefOpt / tpu / PostRev 0.01834349 s 0.018301281875 s 1.00
jaxmd20 / IDefOpt / tpu / BothRev 0.0255843843749999 s 0.02555139 s 1.00
jaxmd40 / JaXPipe / cpu / Primal 0.0687999389999999 s 0.0715142149999999 s 0.96
jaxmd40 / Jax / cpu / Primal 0.0710809809999999 s 0.071793129 s 0.99
jaxmd40 / HLOOpt / cpu / Primal 0.091523892 s 0.096050322 s 0.95
jaxmd40 / PartOpt / cpu / Primal 0.069485885 s 0.075406883 s 0.92
jaxmd40 / IPartOpt / cpu / Primal 0.067604632 s 0.072501867 s 0.93
jaxmd40 / DefOpt / cpu / Primal 0.093857628 s 0.093348485 s 1.01
jaxmd40 / IDefOpt / cpu / Primal 0.09189211 s 0.096174862 s 0.96
jaxmd40 / JaXPipe / cpu / Forward 0.1691235059999999 s 0.167727442 s 1.01
jaxmd40 / Jax / cpu / Forward 0.090996497 s 0.089430882 s 1.02
jaxmd40 / HLOOpt / cpu / Forward 0.1627435529999999 s 0.16773789 s 0.97
jaxmd40 / PartOpt / cpu / Forward 0.168396774 s 0.16891999 s 1.00
jaxmd40 / IPartOpt / cpu / Forward 0.161285557 s 0.170297303 s 0.95
jaxmd40 / DefOpt / cpu / Forward 0.166302271 s 0.168942012 s 0.98
jaxmd40 / IDefOpt / cpu / Forward 0.167911657 s 0.172370059 s 0.97
jaxmd40 / JaXPipe / cpu / PreRev 0.225888807 s 0.246402501 s 0.92
jaxmd40 / JaXPipe / cpu / PostRev 0.146111627 s 0.140108943 s 1.04
jaxmd40 / JaXPipe / cpu / BothRev 0.25196933 s 0.234342806 s 1.08
jaxmd40 / Jax / cpu / BothRev 0.136940089 s 0.137563376 s 1.00
jaxmd40 / HLOOpt / cpu / PreRev 0.2229825339999999 s 0.226273474 s 0.99
jaxmd40 / HLOOpt / cpu / PostRev 0.179397601 s 0.178276109 s 1.01
jaxmd40 / HLOOpt / cpu / BothRev 0.25464187 s 0.248443251 s 1.02
jaxmd40 / PartOpt / cpu / PreRev 0.242189814 s 0.227452227 s 1.06
jaxmd40 / PartOpt / cpu / PostRev 0.138690014 s 0.130284111 s 1.06
jaxmd40 / PartOpt / cpu / BothRev 0.253496024 s 0.2465847699999999 s 1.03
jaxmd40 / IPartOpt / cpu / PreRev 0.2417695919999999 s 0.22577963 s 1.07
jaxmd40 / IPartOpt / cpu / PostRev 0.134319861 s 0.14343332 s 0.94
jaxmd40 / IPartOpt / cpu / BothRev 0.254316743 s 0.2539809779999999 s 1.00
jaxmd40 / DefOpt / cpu / PreRev 0.241453073 s 0.230823172 s 1.05
jaxmd40 / DefOpt / cpu / PostRev 0.171336311 s 0.1840009849999999 s 0.93
jaxmd40 / DefOpt / cpu / BothRev 0.245049834 s 0.251601763 s 0.97
jaxmd40 / IDefOpt / cpu / PreRev 0.227253061 s 0.239430548 s 0.95
jaxmd40 / IDefOpt / cpu / PostRev 0.175590594 s 0.1760787939999999 s 1.00
jaxmd40 / IDefOpt / cpu / BothRev 0.257644724 s 0.260008438 s 0.99
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / JaXPipe / cuda / Primal 1.702872536 s 1.701965816 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / Jax / cuda / Primal 1.705362856 s 1.7047771029999998 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / HLOOpt / cuda / Primal 1.7164505449999998 s 1.714594805 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / PartOpt / cuda / Primal 1.696581195 s 1.696450168 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / IPartOpt / cuda / Primal 1.6940947000000002 s 1.694809473 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / DefOpt / cuda / Primal 1.665714621 s 1.664932468 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / IDefOpt / cuda / Primal 1.915413227 s 1.920949623 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / JaXPipe / tpu / Primal 3.038836420625 s 3.038568994375 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / Jax / tpu / Primal 3.039341376875 s 3.039189840625 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / HLOOpt / tpu / Primal 3.121594910625 s 3.1215918150000004 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / PartOpt / tpu / Primal 3.0601986081250003 s 3.05983537625 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / IPartOpt / tpu / Primal 3.06042502125 s 3.060004419375 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / DefOpt / tpu / Primal 2.102418320625 s 2.10238759125 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / IDefOpt / tpu / Primal 2.948271054375 s 2.944661759375 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / JaXPipe / cpu / Primal 6.166764331 s 6.304237769 s 0.98
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / Jax / cpu / Primal 6.188737092 s 6.35860167 s 0.97
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / HLOOpt / cpu / Primal 6.15145843 s 6.390747236999999 s 0.96
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / PartOpt / cpu / Primal 6.255778277 s 6.4081710780000005 s 0.98
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / IPartOpt / cpu / Primal 6.173246488 s 6.465016108 s 0.95
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / DefOpt / cpu / Primal 2.45970196 s 2.635569419 s 0.93
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / IDefOpt / cpu / Primal 6.712345735 s 7.045801032 s 0.95

This comment was automatically generated by workflow using github-action-benchmark.

@avik-pal avik-pal changed the title feat: remove more intermediate reshape operations [DONT MERGE] feat: remove more intermediate reshape operations Dec 29, 2025
@avik-pal avik-pal force-pushed the ap/more_reshape_simplify branch from b2739f2 to d0c24b3 Compare December 29, 2025 18:00
@avik-pal avik-pal requested a review from wsmoses December 29, 2025 18:29
@avik-pal avik-pal marked this pull request as ready for review December 29, 2025 18:29
// Topological sort: we need to process ops in dependency order
// Since we collected via DFS, we need to reverse and ensure dependencies
// come first
auto sortedOps = mlir::topologicalSort(chainOps);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

the toposort here still worries me a bit here, in contrast to a linear scan through the block [at least on large blocks]

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

let me ponder a bit, we might be able to get away without sorting if we ensure a proper insertion order into the chain

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

should now avoid the topological sort

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@wsmoses is this good to go?

@avik-pal avik-pal force-pushed the ap/more_reshape_simplify branch from d0c24b3 to 49184ad Compare December 29, 2025 21:56
@avik-pal avik-pal requested a review from wsmoses December 29, 2025 21:58
@avik-pal avik-pal force-pushed the ap/more_reshape_simplify branch 2 times, most recently from adb1748 to d1b5e26 Compare December 30, 2025 14:26
@avik-pal avik-pal linked an issue Jan 1, 2026 that may be closed by this pull request
@avik-pal avik-pal force-pushed the ap/more_reshape_simplify branch from d1b5e26 to 19db57f Compare January 1, 2026 21:14
@wsmoses wsmoses merged commit a6af9b3 into main Jan 3, 2026
26 of 28 checks passed
@wsmoses wsmoses deleted the ap/more_reshape_simplify branch January 3, 2026 22:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

break up reduce(add, add(mul(), mul())) more unwanted reshapes

3 participants