Skip to content

Conversation

@avik-pal
Copy link
Collaborator

@avik-pal avik-pal commented Jan 7, 2026

  • gather with iota indexing
    • tests

@avik-pal avik-pal force-pushed the ap/scatter_gather_with_iota_indexing branch 6 times, most recently from 710321c to 768ce4d Compare January 9, 2026 01:22
@avik-pal avik-pal changed the title feat: more optimizations feat: gather iota indexing optimizations Jan 9, 2026
@avik-pal avik-pal force-pushed the ap/scatter_gather_with_iota_indexing branch from 768ce4d to b6bdb77 Compare January 9, 2026 02:13
@avik-pal avik-pal marked this pull request as ready for review January 9, 2026 02:18
@avik-pal avik-pal requested a review from wsmoses January 9, 2026 02:18
@avik-pal avik-pal force-pushed the ap/scatter_gather_with_iota_indexing branch from b6bdb77 to 0018a68 Compare January 9, 2026 02:20
@wsmoses wsmoses merged commit 6b24d4a into main Jan 9, 2026
21 of 25 checks passed
@wsmoses wsmoses deleted the ap/scatter_gather_with_iota_indexing branch January 9, 2026 04:36
@avik-pal
Copy link
Collaborator Author

avik-pal commented Jan 9, 2026

the failure is real. fixing in a follow up

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: 0018a68 Previous: 9749c28 Ratio
actmtch / JaXPipe / cpu / Primal 0.000006480840029325918 s 0.000007178499872679822 s 0.90
actmtch / Jax / cpu / Primal 0.000006448240037570941 s 0.0000069215799157973376 s 0.93
actmtch / HLOOpt / cpu / Primal 0.0000069831399468967 s 0.00000738472012017155 s 0.95
actmtch / PartOpt / cpu / Primal 0.000006246259972613188 s 0.000006676359917037189 s 0.94
actmtch / IPartOpt / cpu / Primal 0.000006322779991023708 s 0.0000066632801463129 s 0.95
actmtch / DefOpt / cpu / Primal 0.0000071219199890038 s 0.0000070232199868769385 s 1.01
actmtch / IDefOpt / cpu / Primal 0.000007055779997244827 s 0.0000071157399361254646 s 0.99
actmtch / JaXPipe / cpu / Forward 0.000010433720008222736 s 0.00001071821985533461 s 0.97
actmtch / Jax / cpu / Forward 0.000009771019949766924 s 0.000010183720041823107 s 0.96
actmtch / HLOOpt / cpu / Forward 0.000010915839957306162 s 0.000011394539869797882 s 0.96
actmtch / PartOpt / cpu / Forward 0.000010746339958132012 s 0.000010248900107399096 s 1.05
actmtch / IPartOpt / cpu / Forward 0.00001084615999388916 s 0.000010885099873121365 s 1.00
actmtch / DefOpt / cpu / Forward 0.000010668199993233429 s 0.000010958880120597312 s 0.97
actmtch / IDefOpt / cpu / Forward 0.000010595879984975907 s 0.000010758259886642918 s 0.98
actmtch / JaXPipe / cpu / PreRev 0.00001059808001627971 s 0.00001075559983291896 s 0.99
actmtch / JaXPipe / cpu / PostRev 0.0000099785200018232 s 0.000010429819885757752 s 0.96
actmtch / JaXPipe / cpu / BothRev 0.00001115823998588894 s 0.000011315619995002636 s 0.99
actmtch / Jax / cpu / BothRev 0.000009109599968724067 s 0.000009239179926225915 s 0.99
actmtch / HLOOpt / cpu / PreRev 0.000010915519987975133 s 0.000010561299859546123 s 1.03
actmtch / HLOOpt / cpu / PostRev 0.000012069239955962985 s 0.000013761459958914202 s 0.88
actmtch / HLOOpt / cpu / BothRev 0.000010682180000003428 s 0.000011175479849043769 s 0.96
actmtch / PartOpt / cpu / PreRev 0.00001047332001689938 s 0.000010472680114617106 s 1.00
actmtch / PartOpt / cpu / PostRev 0.000009623299993108958 s 0.000009276260025217198 s 1.04
actmtch / PartOpt / cpu / BothRev 0.000011377439986972604 s 0.000011865220039908307 s 0.96
actmtch / IPartOpt / cpu / PreRev 0.000010196500015808852 s 0.000011052040244976524 s 0.92
actmtch / IPartOpt / cpu / PostRev 0.000009512899996479972 s 0.000010075359750771894 s 0.94
actmtch / IPartOpt / cpu / BothRev 0.000011297240016574506 s 0.00001128298001276562 s 1.00
actmtch / DefOpt / cpu / PreRev 0.000010503240046091378 s 0.000011099379953520838 s 0.95
actmtch / DefOpt / cpu / PostRev 0.00001128756003708986 s 0.0000108453000211739 s 1.04
actmtch / DefOpt / cpu / BothRev 0.000010898359978455118 s 0.000012118520025978796 s 0.90
actmtch / IDefOpt / cpu / PreRev 0.000010891380015891628 s 0.00001125396007410018 s 0.97
actmtch / IDefOpt / cpu / PostRev 0.000011162919990965747 s 0.000011109960032626988 s 1.00
actmtch / IDefOpt / cpu / BothRev 0.000011074839985667497 s 0.000011985499986622017 s 0.92
actmtch / JaXPipe / cuda / Primal 0.0000024 s 0.0000024 s 1
actmtch / Jax / cuda / Primal 0.0000024 s 0.000002431 s 0.99
actmtch / HLOOpt / cuda / Primal 0.0000024 s 0.000002431 s 0.99
actmtch / PartOpt / cuda / Primal 0.0000024 s 0.0000024 s 1
actmtch / IPartOpt / cuda / Primal 0.0000024 s 0.0000024 s 1
actmtch / DefOpt / cuda / Primal 0.0000024 s 0.0000024 s 1
actmtch / IDefOpt / cuda / Primal 0.0000024 s 0.0000024 s 1
actmtch / JaXPipe / cuda / Forward 0.00001056 s 0.00001104 s 0.96
actmtch / Jax / cuda / Forward 0.000010112 s 0.000010465 s 0.97
actmtch / HLOOpt / cuda / Forward 0.000010656 s 0.000010527 s 1.01
actmtch / PartOpt / cuda / Forward 0.000010495 s 0.000010912 s 0.96
actmtch / IPartOpt / cuda / Forward 0.000010432 s 0.00001056 s 0.99
actmtch / DefOpt / cuda / Forward 0.000010976 s 0.000010592 s 1.04
actmtch / IDefOpt / cuda / Forward 0.000010527 s 0.00001088 s 0.97
actmtch / JaXPipe / cuda / PreRev 0.000010912 s 0.000010592 s 1.03
actmtch / JaXPipe / cuda / PostRev 0.00001104 s 0.000010848 s 1.02
actmtch / JaXPipe / cuda / BothRev 0.000010369 s 0.000010688 s 0.97
actmtch / Jax / cuda / BothRev 0.00001072 s 0.0000104 s 1.03
actmtch / HLOOpt / cuda / PreRev 0.000011904 s 0.000011007 s 1.08
actmtch / HLOOpt / cuda / PostRev 0.000011071 s 0.00001072 s 1.03
actmtch / HLOOpt / cuda / BothRev 0.000010656 s 0.000010751 s 0.99
actmtch / PartOpt / cuda / PreRev 0.00001088 s 0.000010848 s 1.00
actmtch / PartOpt / cuda / PostRev 0.000010528 s 0.000011008 s 0.96
actmtch / PartOpt / cuda / BothRev 0.000010784 s 0.000010656 s 1.01
actmtch / IPartOpt / cuda / PreRev 0.00001072 s 0.000010816 s 0.99
actmtch / IPartOpt / cuda / PostRev 0.000010752 s 0.000010912 s 0.99
actmtch / IPartOpt / cuda / BothRev 0.000010847 s 0.000010847 s 1
actmtch / DefOpt / cuda / PreRev 0.000016414999999999998 s 0.000011168 s 1.47
actmtch / DefOpt / cuda / PostRev 0.000011264 s 0.000011136 s 1.01
actmtch / DefOpt / cuda / BothRev 0.000010848 s 0.000011169 s 0.97
actmtch / IDefOpt / cuda / PreRev 0.00000992 s 0.000011104 s 0.89
actmtch / IDefOpt / cuda / PostRev 0.000010912 s 0.000010304 s 1.06
actmtch / IDefOpt / cuda / BothRev 0.000010592 s 0.000010528 s 1.01
actmtch / JaXPipe / tpu / Primal 5.6365e-7 s 5.823e-7 s 0.97
actmtch / Jax / tpu / Primal 5.967e-7 s 5.63425e-7 s 1.06
actmtch / HLOOpt / tpu / Primal 0.000002115525 s 0.00000216355 s 0.98
actmtch / PartOpt / tpu / Primal 5.971000000000001e-7 s 5.63175e-7 s 1.06
actmtch / IPartOpt / tpu / Primal 5.528e-7 s 5.7545e-7 s 0.96
actmtch / DefOpt / tpu / Primal 0.000002174575 s 0.000002061875 s 1.05
actmtch / IDefOpt / tpu / Primal 0.000002107 s 0.00000216755 s 0.97
actmtch / JaXPipe / tpu / Forward 0.000003833274999999999 s 0.000003854575 s 0.99
actmtch / Jax / tpu / Forward 0.000001214475 s 0.00000123035 s 0.99
actmtch / HLOOpt / tpu / Forward 0.000003931875 s 0.0000036632 s 1.07
actmtch / PartOpt / tpu / Forward 0.000003914025 s 0.000003898975 s 1.00
actmtch / IPartOpt / tpu / Forward 0.0000039373250000000005 s 0.000003676975000000001 s 1.07
actmtch / DefOpt / tpu / Forward 0.0000039175 s 0.000003894825 s 1.01
actmtch / IDefOpt / tpu / Forward 0.000003935675 s 0.0000036601 s 1.08
actmtch / JaXPipe / tpu / PreRev 0.000003481625 s 0.00000374785 s 0.93
actmtch / JaXPipe / tpu / PostRev 0.0000016329 s 0.0000016244500000000002 s 1.01
actmtch / JaXPipe / tpu / BothRev 0.00000347945 s 0.00000374185 s 0.93
actmtch / Jax / tpu / BothRev 0.000001641775 s 0.0000016273499999999998 s 1.01
actmtch / HLOOpt / tpu / PreRev 0.0000034899 s 0.000003769575 s 0.93
actmtch / HLOOpt / tpu / PostRev 0.00000340945 s 0.000003456725 s 0.99
actmtch / HLOOpt / tpu / BothRev 0.000003499275 s 0.000003745975 s 0.93
actmtch / PartOpt / tpu / PreRev 0.0000034107500000000003 s 0.000003455375 s 0.99
actmtch / PartOpt / tpu / PostRev 0.000001583175 s 0.00000167475 s 0.95
actmtch / PartOpt / tpu / BothRev 0.000003422175 s 0.0000034385 s 1.00
actmtch / IPartOpt / tpu / PreRev 0.000003486325 s 0.00000374925 s 0.93
actmtch / IPartOpt / tpu / PostRev 0.000001638625 s 0.000001629225 s 1.01
actmtch / IPartOpt / tpu / BothRev 0.000003466625 s 0.00000374965 s 0.92
actmtch / DefOpt / tpu / PreRev 0.000003415425 s 0.00000344795 s 0.99
actmtch / DefOpt / tpu / PostRev 0.000003414125 s 0.00000366565 s 0.93
actmtch / DefOpt / tpu / BothRev 0.000003403625 s 0.000003442575 s 0.99
actmtch / IDefOpt / tpu / PreRev 0.0000034718500000000005 s 0.000003745325 s 0.93
actmtch / IDefOpt / tpu / PostRev 0.00000341105 s 0.0000034381 s 0.99
actmtch / IDefOpt / tpu / BothRev 0.0000034642 s 0.0000037527500000000006 s 0.92
actmtch / JaXPipe / cpu / Primal 0.000013295 s 0.000007178499872679822 s 1.85
actmtch / Jax / cpu / Primal 0.000013471 s 0.0000069215799157973376 s 1.95
actmtch / HLOOpt / cpu / Primal 0.000013977 s 0.00000738472012017155 s 1.89
actmtch / PartOpt / cpu / Primal 0.000013289 s 0.000006676359917037189 s 1.99
actmtch / IPartOpt / cpu / Primal 0.000013722 s 0.0000066632801463129 s 2.06
actmtch / DefOpt / cpu / Primal 0.000014444 s 0.0000070232199868769385 s 2.06
actmtch / IDefOpt / cpu / Primal 0.00001408 s 0.0000071157399361254646 s 1.98
actmtch / JaXPipe / cpu / Forward 0.00001995 s 0.00001071821985533461 s 1.86
actmtch / Jax / cpu / Forward 0.000018292 s 0.000010183720041823107 s 1.80
actmtch / HLOOpt / cpu / Forward 0.000019565 s 0.000011394539869797882 s 1.72
actmtch / PartOpt / cpu / Forward 0.000019646 s 0.000010248900107399096 s 1.92
actmtch / IPartOpt / cpu / Forward 0.000019301 s 0.000010885099873121365 s 1.77
actmtch / DefOpt / cpu / Forward 0.000019767 s 0.000010958880120597312 s 1.80
actmtch / IDefOpt / cpu / Forward 0.000019396 s 0.000010758259886642918 s 1.80
actmtch / JaXPipe / cpu / PreRev 0.000019509 s 0.00001075559983291896 s 1.81
actmtch / JaXPipe / cpu / PostRev 0.000017755 s 0.000010429819885757752 s 1.70
actmtch / JaXPipe / cpu / BothRev 0.000019495 s 0.000011315619995002636 s 1.72
actmtch / Jax / cpu / BothRev 0.000018214 s 0.000009239179926225915 s 1.97
actmtch / HLOOpt / cpu / PreRev 0.00001873 s 0.000010561299859546123 s 1.77
actmtch / HLOOpt / cpu / PostRev 0.000019458 s 0.000013761459958914202 s 1.41
actmtch / HLOOpt / cpu / BothRev 0.000019206 s 0.000011175479849043769 s 1.72
actmtch / PartOpt / cpu / PreRev 0.000019491 s 0.000010472680114617106 s 1.86
actmtch / PartOpt / cpu / PostRev 0.000017958 s 0.000009276260025217198 s 1.94
actmtch / PartOpt / cpu / BothRev 0.000019243 s 0.000011865220039908307 s 1.62
actmtch / IPartOpt / cpu / PreRev 0.000019074 s 0.000011052040244976524 s 1.73
actmtch / IPartOpt / cpu / PostRev 0.000018032 s 0.000010075359750771894 s 1.79
actmtch / IPartOpt / cpu / BothRev 0.000019142 s 0.00001128298001276562 s 1.70
actmtch / DefOpt / cpu / PreRev 0.000018969 s 0.000011099379953520838 s 1.71
actmtch / DefOpt / cpu / PostRev 0.000019113 s 0.0000108453000211739 s 1.76
actmtch / DefOpt / cpu / BothRev 0.000019417 s 0.000012118520025978796 s 1.60
actmtch / IDefOpt / cpu / PreRev 0.000019257 s 0.00001125396007410018 s 1.71
actmtch / IDefOpt / cpu / PostRev 0.000019382 s 0.000011109960032626988 s 1.74
actmtch / IDefOpt / cpu / BothRev 0.000019575 s 0.000011985499986622017 s 1.63
add_one / JaXPipe / cpu / Primal 0.000006749640024281689 s 0.000006788360224163625 s 0.99
add_one / Jax / cpu / Primal 0.000006460519998654491 s 0.000006639620187343098 s 0.97
add_one / HLOOpt / cpu / Primal 0.000006847339964224375 s 0.000007346800084633287 s 0.93
add_one / PartOpt / cpu / Primal 0.000006429439990824904 s 0.000006595760023628827 s 0.97
add_one / IPartOpt / cpu / Primal 0.00000661434001813177 s 0.000006949459930183366 s 0.95
add_one / DefOpt / cpu / Primal 0.000006262640008571907 s 0.0000064497800485696645 s 0.97
add_one / IDefOpt / cpu / Primal 0.000006418959983420791 s 0.000006538660054502543 s 0.98
add_one / JaXPipe / cpu / Forward 0.000009555560000080732 s 0.000010474200062162708 s 0.91
add_one / Jax / cpu / Forward 0.000011702879983204183 s 0.00001017146001686342 s 1.15
add_one / HLOOpt / cpu / Forward 0.00001026835998345632 s 0.000010015480002039113 s 1.03
add_one / PartOpt / cpu / Forward 0.000010552519997872876 s 0.000010248180115013384 s 1.03
add_one / IPartOpt / cpu / Forward 0.000009876400008579367 s 0.000010215500260528645 s 0.97
add_one / DefOpt / cpu / Forward 0.00000971111998296692 s 0.000010283020164933987 s 0.94
add_one / IDefOpt / cpu / Forward 0.000009708180032248492 s 0.000009787819981283976 s 0.99
add_one / JaXPipe / cpu / PreRev 0.000011194739990969535 s 0.000011898379962076432 s 0.94
add_one / JaXPipe / cpu / PostRev 0.000010815499990712851 s 0.000011255620083829854 s 0.96
add_one / JaXPipe / cpu / BothRev 0.000011414879982112324 s 0.000011512960045365616 s 0.99
add_one / Jax / cpu / BothRev 0.000010819220005942042 s 0.00001132955996581586 s 0.95
add_one / HLOOpt / cpu / PreRev 0.000011095739955635508 s 0.00001178636000986444 s 0.94
add_one / HLOOpt / cpu / PostRev 0.00001317952000135847 s 0.00001347255994915031 s 0.98
add_one / HLOOpt / cpu / BothRev 0.000011253259963268648 s 0.0000117777798732277 s 0.96
add_one / PartOpt / cpu / PreRev 0.000011011799924744992 s 0.000011587020017032046 s 0.95
add_one / PartOpt / cpu / PostRev 0.000010766780042104072 s 0.00001128248004533816 s 0.95
add_one / PartOpt / cpu / BothRev 0.000011455880021458142 s 0.000012154820069554262 s 0.94
add_one / IPartOpt / cpu / PreRev 0.00001155569999355066 s 0.000011255660210736095 s 1.03
add_one / IPartOpt / cpu / PostRev 0.000011315200026729144 s 0.000011514540019561535 s 0.98
add_one / IPartOpt / cpu / BothRev 0.000010946200000034878 s 0.000011263940286880823 s 0.97
add_one / DefOpt / cpu / PreRev 0.000011241059974054224 s 0.00001126040006056428 s 1.00
add_one / DefOpt / cpu / PostRev 0.000011185120047230158 s 0.000011456360080046578 s 0.98
add_one / DefOpt / cpu / BothRev 0.00001101718000427354 s 0.00001145955997344572 s 0.96
add_one / IDefOpt / cpu / PreRev 0.000010762839956441894 s 0.00001153671990323346 s 0.93
add_one / IDefOpt / cpu / PostRev 0.000010986199995386416 s 0.000011387340055080131 s 0.96
add_one / IDefOpt / cpu / BothRev 0.000011265280008956324 s 0.00001162638007372152 s 0.97
add_one / JaXPipe / cuda / Primal 0.000002304 s 0.000002303 s 1.00
add_one / Jax / cuda / Primal 0.000002304 s 0.000002303 s 1.00
add_one / HLOOpt / cuda / Primal 0.000002304 s 0.000002303 s 1.00
add_one / PartOpt / cuda / Primal 0.000002303 s 0.000002303 s 1
add_one / IPartOpt / cuda / Primal 0.000002304 s 0.000002303 s 1.00
add_one / DefOpt / cuda / Primal 0.000002304 s 0.000002303 s 1.00
add_one / IDefOpt / cuda / Primal 0.000002304 s 0.000002303 s 1.00
add_one / JaXPipe / cuda / Forward 0.000011328 s 0.000010848 s 1.04
add_one / Jax / cuda / Forward 0.00001104 s 0.000010688 s 1.03
add_one / HLOOpt / cuda / Forward 0.00001072 s 0.000010816 s 0.99
add_one / PartOpt / cuda / Forward 0.000010847 s 0.000009952 s 1.09
add_one / IPartOpt / cuda / Forward 0.000010592 s 0.000010816 s 0.98
add_one / DefOpt / cuda / Forward 0.00001072 s 0.000010688 s 1.00
add_one / IDefOpt / cuda / Forward 0.000011009 s 0.00001024 s 1.08
add_one / JaXPipe / cuda / PreRev 0.00002544 s 0.000024704 s 1.03
add_one / JaXPipe / cuda / PostRev 0.000025536 s 0.0000256 s 1.00
add_one / JaXPipe / cuda / BothRev 0.000025313 s 0.000025824 s 0.98
add_one / Jax / cuda / BothRev 0.000025888 s 0.00002592 s 1.00
add_one / HLOOpt / cuda / PreRev 0.000025312 s 0.000026048 s 0.97
add_one / HLOOpt / cuda / PostRev 0.000025472000000000003 s 0.000025696 s 0.99
add_one / HLOOpt / cuda / BothRev 0.000025088 s 0.000025953 s 0.97
add_one / PartOpt / cuda / PreRev 0.000025952 s 0.000025696 s 1.01
add_one / PartOpt / cuda / PostRev 0.000025408 s 0.000025728 s 0.99
add_one / PartOpt / cuda / BothRev 0.000025312 s 0.000025599 s 0.99
add_one / IPartOpt / cuda / PreRev 0.000025632 s 0.000026016 s 0.99
add_one / IPartOpt / cuda / PostRev 0.000025504 s 0.000025664 s 0.99
add_one / IPartOpt / cuda / BothRev 0.000025632 s 0.000026144 s 0.98
add_one / DefOpt / cuda / PreRev 0.00002528 s 0.000025312 s 1.00
add_one / DefOpt / cuda / PostRev 0.000025728 s 0.000026432 s 0.97
add_one / DefOpt / cuda / BothRev 0.000025216 s 0.000025792 s 0.98
add_one / IDefOpt / cuda / PreRev 0.000025536 s 0.000026496 s 0.96
add_one / IDefOpt / cuda / PostRev 0.000025568 s 0.000025632 s 1.00
add_one / IDefOpt / cuda / BothRev 0.00002608 s 0.000025952 s 1.00
add_one / JaXPipe / tpu / Primal 0.0000014332749999999998 s 0.000001445175 s 0.99
add_one / Jax / tpu / Primal 0.0000014069 s 0.000001447025 s 0.97
add_one / HLOOpt / tpu / Primal 0.0000014319 s 0.00000145205 s 0.99
add_one / PartOpt / tpu / Primal 0.0000014080999999999997 s 0.0000014558 s 0.97
add_one / IPartOpt / tpu / Primal 0.000001424575 s 0.0000014553 s 0.98
add_one / DefOpt / tpu / Primal 0.0000014011000000000002 s 0.000001455125 s 0.96
add_one / IDefOpt / tpu / Primal 0.000001423725 s 0.000001453475 s 0.98
add_one / JaXPipe / tpu / Forward 0.00000186385 s 0.0000019106 s 0.98
add_one / Jax / tpu / Forward 0.0000018374 s 0.000001863875 s 0.99
add_one / HLOOpt / tpu / Forward 0.00000184355 s 0.000001903025 s 0.97
add_one / PartOpt / tpu / Forward 0.000001847 s 0.000001868175 s 0.99
add_one / IPartOpt / tpu / Forward 0.000001848475 s 0.0000019037 s 0.97
add_one / DefOpt / tpu / Forward 0.0000018374 s 0.000001863 s 0.99
add_one / IDefOpt / tpu / Forward 0.000001844125 s 0.0000019137 s 0.96
add_one / JaXPipe / tpu / PreRev 0.0000022361 s 0.0000022631 s 0.99
add_one / JaXPipe / tpu / PostRev 0.00000223975 s 0.000002302475 s 0.97
add_one / JaXPipe / tpu / BothRev 0.0000022392750000000003 s 0.0000022539 s 0.99
add_one / Jax / tpu / BothRev 0.000002247575 s 0.0000022978499999999995 s 0.98
add_one / HLOOpt / tpu / PreRev 0.000002236275 s 0.000002257775 s 0.99
add_one / HLOOpt / tpu / PostRev 0.000002242225 s 0.000002294675 s 0.98
add_one / HLOOpt / tpu / BothRev 0.0000022413 s 0.00000225465 s 0.99
add_one / PartOpt / tpu / PreRev 0.0000022377 s 0.000002291325 s 0.98
add_one / PartOpt / tpu / PostRev 0.000002236475 s 0.00000225305 s 0.99
add_one / PartOpt / tpu / BothRev 0.000002235825 s 0.000002290925 s 0.98
add_one / IPartOpt / tpu / PreRev 0.0000022377500000000003 s 0.000002252175 s 0.99
add_one / IPartOpt / tpu / PostRev 0.000002241075 s 0.000002288375 s 0.98
add_one / IPartOpt / tpu / BothRev 0.00000224 s 0.00000225975 s 0.99
add_one / DefOpt / tpu / PreRev 0.000002240075 s 0.000002292425 s 0.98
add_one / DefOpt / tpu / PostRev 0.00000224615 s 0.000002253675 s 1.00
add_one / DefOpt / tpu / BothRev 0.000002238 s 0.0000022969 s 0.97
add_one / IDefOpt / tpu / PreRev 0.00000224155 s 0.000002254075 s 0.99
add_one / IDefOpt / tpu / PostRev 0.000002245125 s 0.0000022885749999999995 s 0.98
add_one / IDefOpt / tpu / BothRev 0.0000022363 s 0.000002253675 s 0.99
add_one / JaXPipe / cpu / Primal 0.00001317 s 0.000006788360224163625 s 1.94
add_one / Jax / cpu / Primal 0.000013115 s 0.000006639620187343098 s 1.98
add_one / HLOOpt / cpu / Primal 0.000013024 s 0.000007346800084633287 s 1.77
add_one / PartOpt / cpu / Primal 0.00001302 s 0.000006595760023628827 s 1.97
add_one / IPartOpt / cpu / Primal 0.00001333 s 0.000006949459930183366 s 1.92
add_one / DefOpt / cpu / Primal 0.00001308 s 0.0000064497800485696645 s 2.03
add_one / IDefOpt / cpu / Primal 0.000013126 s 0.000006538660054502543 s 2.01
add_one / JaXPipe / cpu / Forward 0.000018068 s 0.000010474200062162708 s 1.73
add_one / Jax / cpu / Forward 0.000018076 s 0.00001017146001686342 s 1.78
add_one / HLOOpt / cpu / Forward 0.000017963 s 0.000010015480002039113 s 1.79
add_one / PartOpt / cpu / Forward 0.000018022 s 0.000010248180115013384 s 1.76
add_one / IPartOpt / cpu / Forward 0.000017843000000000002 s 0.000010215500260528645 s 1.75
add_one / DefOpt / cpu / Forward 0.000017901 s 0.000010283020164933987 s 1.74
add_one / IDefOpt / cpu / Forward 0.000018004 s 0.000009787819981283976 s 1.84
add_one / JaXPipe / cpu / PreRev 0.000020385 s 0.000011898379962076432 s 1.71
add_one / JaXPipe / cpu / PostRev 0.000020104 s 0.000011255620083829854 s 1.79
add_one / JaXPipe / cpu / BothRev 0.000019999 s 0.000011512960045365616 s 1.74
add_one / Jax / cpu / BothRev 0.000019641 s 0.00001132955996581586 s 1.73
add_one / HLOOpt / cpu / PreRev 0.000019817 s 0.00001178636000986444 s 1.68
add_one / HLOOpt / cpu / PostRev 0.000019691 s 0.00001347255994915031 s 1.46
add_one / HLOOpt / cpu / BothRev 0.000020182 s 0.0000117777798732277 s 1.71
add_one / PartOpt / cpu / PreRev 0.00002 s 0.000011587020017032046 s 1.73
add_one / PartOpt / cpu / PostRev 0.000019811 s 0.00001128248004533816 s 1.76
add_one / PartOpt / cpu / BothRev 0.000019871 s 0.000012154820069554262 s 1.63
add_one / IPartOpt / cpu / PreRev 0.000019627 s 0.000011255660210736095 s 1.74
add_one / IPartOpt / cpu / PostRev 0.000019969 s 0.000011514540019561535 s 1.73
add_one / IPartOpt / cpu / BothRev 0.000019676 s 0.000011263940286880823 s 1.75
add_one / DefOpt / cpu / PreRev 0.000019593 s 0.00001126040006056428 s 1.74
add_one / DefOpt / cpu / PostRev 0.000033159 s 0.000011456360080046578 s 2.89
add_one / DefOpt / cpu / BothRev 0.000019797 s 0.00001145955997344572 s 1.73
add_one / IDefOpt / cpu / PreRev 0.000019587 s 0.00001153671990323346 s 1.70
add_one / IDefOpt / cpu / PostRev 0.000019802 s 0.000011387340055080131 s 1.74
add_one / IDefOpt / cpu / BothRev 0.000019737 s 0.00001162638007372152 s 1.70
add_two / JaXPipe / cpu / Primal 0.000006744399961462477 s 0.000007508040071115829 s 0.90
add_two / Jax / cpu / Primal 0.00000725338002666831 s 0.000007063660050334875 s 1.03
add_two / HLOOpt / cpu / Primal 0.000007474579997506226 s 0.000006841000285930932 s 1.09
add_two / PartOpt / cpu / Primal 0.000006766079968656413 s 0.00000713966008333955 s 0.95
add_two / IPartOpt / cpu / Primal 0.000007332519980991492 s 0.000007327299972530454 s 1.00
add_two / DefOpt / cpu / Primal 0.000006640560031883069 s 0.0000073373199484194626 s 0.91
add_two / IDefOpt / cpu / Primal 0.000006745600012436626 s 0.00000649896013783291 s 1.04
add_two / JaXPipe / cpu / Forward 0.000009772819994395832 s 0.000010095840043504722 s 0.97
add_two / Jax / cpu / Forward 0.000010164500044993474 s 0.000010318800086679402 s 0.99
add_two / HLOOpt / cpu / Forward 0.00001001997998173465 s 0.000010510679967410397 s 0.95
add_two / PartOpt / cpu / Forward 0.000009908600031849344 s 0.000010098139973706566 s 0.98
add_two / IPartOpt / cpu / Forward 0.00000988983999377524 s 0.000010444340186950284 s 0.95
add_two / DefOpt / cpu / Forward 0.000009689580019767164 s 0.000010026939889939968 s 0.97
add_two / IDefOpt / cpu / Forward 0.000010364240006310866 s 0.000010137019962712655 s 1.02
add_two / JaXPipe / cpu / PreRev 0.00001313253998887376 s 0.00001420562002749648 s 0.92
add_two / JaXPipe / cpu / PostRev 0.000014123119990472331 s 0.000013980979929328896 s 1.01
add_two / JaXPipe / cpu / BothRev 0.00001397969997015025 s 0.00001418544001353439 s 0.99
add_two / Jax / cpu / BothRev 0.000013415639969025506 s 0.000013970599975436923 s 0.96
add_two / HLOOpt / cpu / PreRev 0.000013454099989758108 s 0.000013817140097671652 s 0.97
add_two / HLOOpt / cpu / PostRev 0.000015322980043492862 s 0.000016204279963858424 s 0.95
add_two / HLOOpt / cpu / BothRev 0.000014095079968683422 s 0.000014592959960282316 s 0.97
add_two / PartOpt / cpu / PreRev 0.000013547700000344775 s 0.000014299439899332355 s 0.95
add_two / PartOpt / cpu / PostRev 0.00001392940002915566 s 0.000013849920032953375 s 1.01
add_two / PartOpt / cpu / BothRev 0.0000136702400050126 s 0.000014024859992787242 s 0.97
add_two / IPartOpt / cpu / PreRev 0.000013101460026518908 s 0.00001402448000590084 s 0.93
add_two / IPartOpt / cpu / PostRev 0.000014376560020537 s 0.000014053559934836813 s 1.02
add_two / IPartOpt / cpu / BothRev 0.000013773699984085396 s 0.000014002400166646112 s 0.98
add_two / DefOpt / cpu / PreRev 0.000013236279983175336 s 0.000014023519870534074 s 0.94
add_two / DefOpt / cpu / PostRev 0.000013807539999106664 s 0.000013425140059553086 s 1.03
add_two / DefOpt / cpu / BothRev 0.000013923879987487452 s 0.000014358260050357786 s 0.97
add_two / IDefOpt / cpu / PreRev 0.00001337076004347182 s 0.000014406740083359182 s 0.93
add_two / IDefOpt / cpu / PostRev 0.000014014559974384613 s 0.0000142485800824943 s 0.98
add_two / IDefOpt / cpu / BothRev 0.000014098019955781635 s 0.000014178099991113414 s 0.99
add_two / JaXPipe / cuda / Primal 0.000002432 s 0.0000024 s 1.01
add_two / Jax / cuda / Primal 0.000002431 s 0.0000024 s 1.01
add_two / HLOOpt / cuda / Primal 0.000002432 s 0.000002431 s 1.00
add_two / PartOpt / cuda / Primal 0.000002431 s 0.000002431 s 1
add_two / IPartOpt / cuda / Primal 0.000002431 s 0.0000024 s 1.01
add_two / DefOpt / cuda / Primal 0.000002431 s 0.0000024 s 1.01
add_two / IDefOpt / cuda / Primal 0.000002431 s 0.0000024 s 1.01
add_two / JaXPipe / cuda / Forward 0.000010847 s 0.00001104 s 0.98
add_two / Jax / cuda / Forward 0.000010624 s 0.0000104 s 1.02
add_two / HLOOpt / cuda / Forward 0.000010367 s 0.00001072 s 0.97
add_two / PartOpt / cuda / Forward 0.000010336 s 0.000010335 s 1.00
add_two / IPartOpt / cuda / Forward 0.000010689 s 0.000010368 s 1.03
add_two / DefOpt / cuda / Forward 0.000010752 s 0.000010017 s 1.07
add_two / IDefOpt / cuda / Forward 0.000010912 s 0.000010113 s 1.08
add_two / JaXPipe / cuda / PreRev 0.000032992 s 0.000033152000000000004 s 1.00
add_two / JaXPipe / cuda / PostRev 0.000033344 s 0.000033376 s 1.00
add_two / JaXPipe / cuda / BothRev 0.000032512 s 0.000033088 s 0.98
add_two / Jax / cuda / BothRev 0.000033119999999999995 s 0.000033023 s 1.00
add_two / HLOOpt / cuda / PreRev 0.000033729 s 0.000033824 s 1.00
add_two / HLOOpt / cuda / PostRev 0.000032705 s 0.000032576 s 1.00
add_two / HLOOpt / cuda / BothRev 0.00003344 s 0.000033472 s 1.00
add_two / PartOpt / cuda / PreRev 0.000033088 s 0.000033791 s 0.98
add_two / PartOpt / cuda / PostRev 0.000038016 s 0.000033568 s 1.13
add_two / PartOpt / cuda / BothRev 0.000032863 s 0.000033088 s 0.99
add_two / IPartOpt / cuda / PreRev 0.000034687 s 0.000033696 s 1.03
add_two / IPartOpt / cuda / PostRev 0.000032864 s 0.00003296 s 1.00
add_two / IPartOpt / cuda / BothRev 0.000033856 s 0.000033376 s 1.01
add_two / DefOpt / cuda / PreRev 0.000033856 s 0.000032704 s 1.04
add_two / DefOpt / cuda / PostRev 0.000033407 s 0.000033089 s 1.01
add_two / DefOpt / cuda / BothRev 0.000033248 s 0.000033536000000000006 s 0.99
add_two / IDefOpt / cuda / PreRev 0.000033664 s 0.000032896000000000005 s 1.02
add_two / IDefOpt / cuda / PostRev 0.000033216 s 0.000032032 s 1.04
add_two / IDefOpt / cuda / BothRev 0.000033504 s 0.000034016 s 0.98
add_two / JaXPipe / tpu / Primal 0.000001429525 s 0.0000013973499999999998 s 1.02
add_two / Jax / tpu / Primal 0.000001475975 s 0.00000146445 s 1.01
add_two / HLOOpt / tpu / Primal 0.000001440375 s 0.00000143 s 1.01
add_two / PartOpt / tpu / Primal 0.00000147945 s 0.00000144125 s 1.03
add_two / IPartOpt / tpu / Primal 0.000001431275 s 0.00000139755 s 1.02
add_two / DefOpt / tpu / Primal 0.00000147455 s 0.000001449025 s 1.02
add_two / IDefOpt / tpu / Primal 0.000001429975 s 0.0000013983750000000002 s 1.02
add_two / JaXPipe / tpu / Forward 0.000001828375 s 0.000001801375 s 1.01
add_two / Jax / tpu / Forward 0.00000183035 s 0.000001790175 s 1.02
add_two / HLOOpt / tpu / Forward 0.000001825 s 0.000001801875 s 1.01
add_two / PartOpt / tpu / Forward 0.00000184285 s 0.000001786925 s 1.03
add_two / IPartOpt / tpu / Forward 0.000001832175 s 0.00000180065 s 1.02
add_two / DefOpt / tpu / Forward 0.000001826175 s 0.0000017977000000000002 s 1.02
add_two / IDefOpt / tpu / Forward 0.0000018305 s 0.00000180145 s 1.02
add_two / JaXPipe / tpu / PreRev 0.000002835625 s 0.0000027991 s 1.01
add_two / JaXPipe / tpu / PostRev 0.00000274515 s 0.000002727475 s 1.01
add_two / JaXPipe / tpu / BothRev 0.000002838075 s 0.0000027946250000000004 s 1.02
add_two / Jax / tpu / BothRev 0.0000027635 s 0.0000027328 s 1.01
add_two / HLOOpt / tpu / PreRev 0.000002850675 s 0.000002806575 s 1.02
add_two / HLOOpt / tpu / PostRev 0.000002753975 s 0.00000273025 s 1.01
add_two / HLOOpt / tpu / BothRev 0.000002825975 s 0.00000280145 s 1.01
add_two / PartOpt / tpu / PreRev 0.00000275775 s 0.000002737475 s 1.01
add_two / PartOpt / tpu / PostRev 0.000002830375 s 0.000002805375 s 1.01
add_two / PartOpt / tpu / BothRev 0.0000027642 s 0.000002730425 s 1.01
add_two / IPartOpt / tpu / PreRev 0.00000283665 s 0.000002812925 s 1.01
add_two / IPartOpt / tpu / PostRev 0.0000027553 s 0.0000027271 s 1.01
add_two / IPartOpt / tpu / BothRev 0.0000028352250000000005 s 0.0000027982 s 1.01
add_two / DefOpt / tpu / PreRev 0.000002768925 s 0.00000273005 s 1.01
add_two / DefOpt / tpu / PostRev 0.0000028393000000000005 s 0.000002803275 s 1.01
add_two / DefOpt / tpu / BothRev 0.00000274915 s 0.000002721425 s 1.01
add_two / IDefOpt / tpu / PreRev 0.00000283195 s 0.00000279425 s 1.01
add_two / IDefOpt / tpu / PostRev 0.0000027539 s 0.0000027338000000000004 s 1.01
add_two / IDefOpt / tpu / BothRev 0.00000285205 s 0.000002794225 s 1.02
add_two / JaXPipe / cpu / Primal 0.000013656 s 0.000007508040071115829 s 1.82
add_two / Jax / cpu / Primal 0.000013588 s 0.000007063660050334875 s 1.92
add_two / HLOOpt / cpu / Primal 0.000013473 s 0.000006841000285930932 s 1.97
add_two / PartOpt / cpu / Primal 0.00001361 s 0.00000713966008333955 s 1.91
add_two / IPartOpt / cpu / Primal 0.000013457 s 0.000007327299972530454 s 1.84
add_two / DefOpt / cpu / Primal 0.000013344 s 0.0000073373199484194626 s 1.82
add_two / IDefOpt / cpu / Primal 0.000013304 s 0.00000649896013783291 s 2.05
add_two / JaXPipe / cpu / Forward 0.00001855 s 0.000010095840043504722 s 1.84
add_two / Jax / cpu / Forward 0.000018043 s 0.000010318800086679402 s 1.75
add_two / HLOOpt / cpu / Forward 0.000018073 s 0.000010510679967410397 s 1.72
add_two / PartOpt / cpu / Forward 0.000018139 s 0.000010098139973706566 s 1.80
add_two / IPartOpt / cpu / Forward 0.000017959 s 0.000010444340186950284 s 1.72
add_two / DefOpt / cpu / Forward 0.000018378 s 0.000010026939889939968 s 1.83
add_two / IDefOpt / cpu / Forward 0.000018292 s 0.000010137019962712655 s 1.80
add_two / JaXPipe / cpu / PreRev 0.000023766 s 0.00001420562002749648 s 1.67
add_two / JaXPipe / cpu / PostRev 0.000023392 s 0.000013980979929328896 s 1.67
add_two / JaXPipe / cpu / BothRev 0.000023323 s 0.00001418544001353439 s 1.64
add_two / Jax / cpu / BothRev 0.000023279 s 0.000013970599975436923 s 1.67
add_two / HLOOpt / cpu / PreRev 0.000023423 s 0.000013817140097671652 s 1.70
add_two / HLOOpt / cpu / PostRev 0.000023699 s 0.000016204279963858424 s 1.46
add_two / HLOOpt / cpu / BothRev 0.000023553 s 0.000014592959960282316 s 1.61
add_two / PartOpt / cpu / PreRev 0.000023469 s 0.000014299439899332355 s 1.64
add_two / PartOpt / cpu / PostRev 0.000023513 s 0.000013849920032953375 s 1.70
add_two / PartOpt / cpu / BothRev 0.000023701 s 0.000014024859992787242 s 1.69
add_two / IPartOpt / cpu / PreRev 0.000023258 s 0.00001402448000590084 s 1.66
add_two / IPartOpt / cpu / PostRev 0.000023695 s 0.000014053559934836813 s 1.69
add_two / IPartOpt / cpu / BothRev 0.000023846 s 0.000014002400166646112 s 1.70
add_two / DefOpt / cpu / PreRev 0.000023455 s 0.000014023519870534074 s 1.67
add_two / DefOpt / cpu / PostRev 0.000023561 s 0.000013425140059553086 s 1.75
add_two / DefOpt / cpu / BothRev 0.000023595 s 0.000014358260050357786 s 1.64
add_two / IDefOpt / cpu / PreRev 0.000023333 s 0.000014406740083359182 s 1.62
add_two / IDefOpt / cpu / PostRev 0.000023667 s 0.0000142485800824943 s 1.66
add_two / IDefOpt / cpu / BothRev 0.000023166 s 0.000014178099991113414 s 1.63
cache / JaXPipe / cpu / Primal 0.000006524520022139768 s 0.0000069787199026905 s 0.93
cache / Jax / cpu / Primal 0.000006105239990574774 s 0.0000063129400223260745 s 0.97
cache / HLOOpt / cpu / Primal 0.000005932719996053493 s 0.000006742240257153753 s 0.88
cache / PartOpt / cpu / Primal 0.000006358220016409177 s 0.000005976620013825596 s 1.06
cache / IPartOpt / cpu / Primal 0.000006492120010079816 s 0.000006029200048942584 s 1.08
cache / DefOpt / cpu / Primal 0.000006659280034000403 s 0.000006931579846423119 s 0.96
cache / IDefOpt / cpu / Primal 0.000006107260005592252 s 0.000006147040076029953 s 0.99
cache / JaXPipe / cpu / Forward 0.000014796080022279056 s 0.000015323639890993944 s 0.97
cache / Jax / cpu / Forward 0.000014785600005779998 s 0.00001486237993958639 s 0.99
cache / HLOOpt / cpu / Forward 0.00001533258002382354 s 0.000015309420123230667 s 1.00
cache / PartOpt / cpu / Forward 0.000014471960002993 s 0.000015313760195567737 s 0.95
cache / IPartOpt / cpu / Forward 0.00001546555995446397 s 0.00001572162003867561 s 0.98
cache / DefOpt / cpu / Forward 0.000015041760034364416 s 0.000015381539778900333 s 0.98
cache / IDefOpt / cpu / Forward 0.000014735720014869004 s 0.000015069819964992347 s 0.98
cache / JaXPipe / cpu / PreRev 0.00001574644001266279 s 0.000015864039996813518 s 0.99
cache / JaXPipe / cpu / PostRev 0.00002089581999825896 s 0.000020334559703769628 s 1.03
cache / JaXPipe / cpu / BothRev 0.000016230299970629857 s 0.00001632793999306159 s 0.99
cache / Jax / cpu / BothRev 0.00002066409995677532 s 0.000021568059783021453 s 0.96
cache / HLOOpt / cpu / PreRev 0.00001645104005547182 s 0.00001589912008057581 s 1.03
cache / HLOOpt / cpu / PostRev 0.000019381060019441067 s 0.00001920374001201708 s 1.01
cache / HLOOpt / cpu / BothRev 0.00001641673997255566 s 0.00001632200008316431 s 1.01
cache / PartOpt / cpu / PreRev 0.000016237680001722764 s 0.000016088679876702373 s 1.01
cache / PartOpt / cpu / PostRev 0.000020833939988733615 s 0.00002114913997502299 s 0.99
cache / PartOpt / cpu / BothRev 0.00001675482003520301 s 0.00001660899994021747 s 1.01
cache / IPartOpt / cpu / PreRev 0.000016451159999633092 s 0.00001628128004085738 s 1.01
cache / IPartOpt / cpu / PostRev 0.000020962800008419435 s 0.00002147131999663543 s 0.98
cache / IPartOpt / cpu / BothRev 0.000015446720008185368 s 0.000015473139937967063 s 1.00
cache / DefOpt / cpu / PreRev 0.000015555799973299144 s 0.000016298939954140224 s 0.95
cache / DefOpt / cpu / PostRev 0.000015160580005613156 s 0.00001588609997270396 s 0.95
cache / DefOpt / cpu / BothRev 0.0000157176199627429 s 0.000016655639992677607 s 0.94
cache / IDefOpt / cpu / PreRev 0.000015464740008610534 s 0.000016662720081512817 s 0.93
cache / IDefOpt / cpu / PostRev 0.000015548019955531345 s 0.000015710999869043007 s 0.99
cache / IDefOpt / cpu / BothRev 0.000015369940037999185 s 0.00001563078003528062 s 0.98
cache / JaXPipe / cuda / Primal 0.000002335 s 0.000002336 s 1.00
cache / Jax / cuda / Primal 0.000002335 s 0.000002336 s 1.00
cache / HLOOpt / cuda / Primal 0.000002335 s 0.000002336 s 1.00
cache / PartOpt / cuda / Primal 0.000002335 s 0.000002336 s 1.00
cache / IPartOpt / cuda / Primal 0.000002304 s 0.000002335 s 0.99
cache / DefOpt / cuda / Primal 0.000002335 s 0.000002336 s 1.00
cache / IDefOpt / cuda / Primal 0.000002335 s 0.000002336 s 1.00
cache / JaXPipe / cuda / Forward 0.0000023670000000000004 s 0.000002368 s 1.00
cache / Jax / cuda / Forward 0.0000023670000000000004 s 0.000002368 s 1.00
cache / HLOOpt / cuda / Forward 0.0000023670000000000004 s 0.0000023670000000000004 s 1
cache / PartOpt / cuda / Forward 0.0000023670000000000004 s 0.0000023670000000000004 s 1
cache / IPartOpt / cuda / Forward 0.0000023670000000000004 s 0.000002368 s 1.00
cache / DefOpt / cuda / Forward 0.0000023670000000000004 s 0.000002368 s 1.00
cache / IDefOpt / cuda / Forward 0.0000023670000000000004 s 0.0000023670000000000004 s 1
cache / JaXPipe / cuda / PreRev 0.00001072 s 0.000011488 s 0.93
cache / JaXPipe / cuda / PostRev 0.000010816 s 0.000010816 s 1
cache / JaXPipe / cuda / BothRev 0.00001088 s 0.000011296 s 0.96
cache / Jax / cuda / BothRev 0.000015745 s 0.000011071 s 1.42
cache / HLOOpt / cuda / PreRev 0.00001376 s 0.000013727 s 1.00
cache / HLOOpt / cuda / PostRev 0.000013696 s 0.000013696 s 1
cache / HLOOpt / cuda / BothRev 0.00001376 s 0.000013728 s 1.00
cache / PartOpt / cuda / PreRev 0.000010432 s 0.0000112 s 0.93
cache / PartOpt / cuda / PostRev 0.000010752 s 0.000010976 s 0.98
cache / PartOpt / cuda / BothRev 0.000010752 s 0.000010592 s 1.02
cache / IPartOpt / cuda / PreRev 0.000011007 s 0.000010848 s 1.01
cache / IPartOpt / cuda / PostRev 0.000011008 s 0.000010752 s 1.02
cache / IPartOpt / cuda / BothRev 0.000011071 s 0.000010785 s 1.03
cache / DefOpt / cuda / PreRev 0.000011136 s 0.000010912 s 1.02
cache / DefOpt / cuda / PostRev 0.000011007 s 0.00001088 s 1.01
cache / DefOpt / cuda / BothRev 0.000010784 s 0.000010752 s 1.00
cache / IDefOpt / cuda / PreRev 0.000010592 s 0.000011008 s 0.96
cache / IDefOpt / cuda / PostRev 0.000010976 s 0.000011264 s 0.97
cache / IDefOpt / cuda / BothRev 0.000010783 s 0.000011232 s 0.96
cache / JaXPipe / tpu / Primal 0.0000024626 s 0.000002465675 s 1.00
cache / Jax / tpu / Primal 0.000002488075 s 0.00000248255 s 1.00
cache / HLOOpt / tpu / Primal 0.000002462975 s 0.0000024704750000000003 s 1.00
cache / PartOpt / tpu / Primal 0.0000024467500000000003 s 0.00000246885 s 0.99
cache / IPartOpt / tpu / Primal 0.000002471 s 0.000002460525 s 1.00
cache / DefOpt / tpu / Primal 0.000002467825 s 0.000002455425 s 1.01
cache / IDefOpt / tpu / Primal 0.0000024628 s 0.000002467525 s 1.00
cache / JaXPipe / tpu / Forward 0.0000035347250000000003 s 0.000003551475 s 1.00
cache / Jax / tpu / Forward 0.000003525425 s 0.00000355505 s 0.99
cache / HLOOpt / tpu / Forward 0.000003556325 s 0.00000356615 s 1.00
cache / PartOpt / tpu / Forward 0.000003523275 s 0.000003547725 s 0.99
cache / IPartOpt / tpu / Forward 0.00000356355 s 0.0000035691 s 1.00
cache / DefOpt / tpu / Forward 0.0000035270249999999995 s 0.00000354025 s 1.00
cache / IDefOpt / tpu / Forward 0.000003555475 s 0.000003548275 s 1.00
cache / JaXPipe / tpu / PreRev 0.000005004425 s 0.000004952975 s 1.01
cache / JaXPipe / tpu / PostRev 0.000005020075 s 0.00000499115 s 1.01
cache / JaXPipe / tpu / BothRev 0.0000050448 s 0.000004966925 s 1.02
cache / Jax / tpu / BothRev 0.000005026 s 0.00000498985 s 1.01
cache / HLOOpt / tpu / PreRev 0.000003972825 s 0.0000041277750000000005 s 0.96
cache / HLOOpt / tpu / PostRev 0.000004145175 s 0.000004155575 s 1.00
cache / HLOOpt / tpu / BothRev 0.000003968475 s 0.0000041377 s 0.96
cache / PartOpt / tpu / PreRev 0.000005037125 s 0.000005022225 s 1.00
cache / PartOpt / tpu / PostRev 0.000005030425 s 0.000004973099999999999 s 1.01
cache / PartOpt / tpu / BothRev 0.000005015825 s 0.000005009025 s 1.00
cache / IPartOpt / tpu / PreRev 0.0000050363 s 0.000004991375 s 1.01
cache / IPartOpt / tpu / PostRev 0.000005014475000000001 s 0.000005010125 s 1.00
cache / IPartOpt / tpu / BothRev 0.000005016375 s 0.00000499415 s 1.00
cache / DefOpt / tpu / PreRev 0.000005016925 s 0.000005018575 s 1.00
cache / DefOpt / tpu / PostRev 0.000005009675 s 0.000004991775 s 1.00
cache / DefOpt / tpu / BothRev 0.000005021849999999999 s 0.0000050015 s 1.00
cache / IDefOpt / tpu / PreRev 0.000005022075 s 0.00000499745 s 1.00
cache / IDefOpt / tpu / PostRev 0.000005037175000000001 s 0.0000050139500000000005 s 1.00
cache / IDefOpt / tpu / BothRev 0.000005025875 s 0.0000049622 s 1.01
cache / JaXPipe / cpu / Primal 0.000013006 s 0.0000069787199026905 s 1.86
cache / Jax / cpu / Primal 0.000013179 s 0.0000063129400223260745 s 2.09
cache / HLOOpt / cpu / Primal 0.000012862 s 0.000006742240257153753 s 1.91
cache / PartOpt / cpu / Primal 0.000012926 s 0.000005976620013825596 s 2.16
cache / IPartOpt / cpu / Primal 0.00001283 s 0.000006029200048942584 s 2.13
cache / DefOpt / cpu / Primal 0.000013358 s 0.000006931579846423119 s 1.93
cache / IDefOpt / cpu / Primal 0.000012958 s 0.000006147040076029953 s 2.11
cache / JaXPipe / cpu / Forward 0.000024319 s 0.000015323639890993944 s 1.59
cache / Jax / cpu / Forward 0.000023024 s 0.00001486237993958639 s 1.55
cache / HLOOpt / cpu / Forward 0.000024887 s 0.000015309420123230667 s 1.63
cache / PartOpt / cpu / Forward 0.000026892 s 0.000015313760195567737 s 1.76
cache / IPartOpt / cpu / Forward 0.000023056 s 0.00001572162003867561 s 1.47
cache / DefOpt / cpu / Forward 0.000023201 s 0.000015381539778900333 s 1.51
cache / IDefOpt / cpu / Forward 0.00002603 s 0.000015069819964992347 s 1.73
cache / JaXPipe / cpu / PreRev 0.000018183 s 0.000015864039996813518 s 1.15
cache / JaXPipe / cpu / PostRev 0.000019487 s 0.000020334559703769628 s 0.96
cache / JaXPipe / cpu / BothRev 0.00001841 s 0.00001632793999306159 s 1.13
cache / Jax / cpu / BothRev 0.000020272 s 0.000021568059783021453 s 0.94
cache / HLOOpt / cpu / PreRev 0.000018374 s 0.00001589912008057581 s 1.16
cache / HLOOpt / cpu / PostRev 0.000018352 s 0.00001920374001201708 s 0.96
cache / HLOOpt / cpu / BothRev 0.000017967 s 0.00001632200008316431 s 1.10
cache / PartOpt / cpu / PreRev 0.000018041 s 0.000016088679876702373 s 1.12
cache / PartOpt / cpu / PostRev 0.000019922 s 0.00002114913997502299 s 0.94
cache / PartOpt / cpu / BothRev 0.000017879 s 0.00001660899994021747 s 1.08
cache / IPartOpt / cpu / PreRev 0.00001817 s 0.00001628128004085738 s 1.12
cache / IPartOpt / cpu / PostRev 0.000020734 s 0.00002147131999663543 s 0.97
cache / IPartOpt / cpu / BothRev 0.000017749999999999998 s 0.000015473139937967063 s 1.15
cache / DefOpt / cpu / PreRev 0.000017511 s 0.000016298939954140224 s 1.07
cache / DefOpt / cpu / PostRev 0.000018119 s 0.00001588609997270396 s 1.14
cache / DefOpt / cpu / BothRev 0.000017336 s 0.000016655639992677607 s 1.04
cache / IDefOpt / cpu / PreRev 0.000017749999999999998 s 0.000016662720081512817 s 1.07
cache / IDefOpt / cpu / PostRev 0.000025914 s 0.000015710999869043007 s 1.65
cache / IDefOpt / cpu / BothRev 0.000034429 s 0.00001563078003528062 s 2.20
Concat / JaXPipe / cpu / Primal 0.000006781759993828018 s 0.000006810739942011423 s 1.00
Concat / Jax / cpu / Primal 0.000006882879988552304 s 0.00000690752003720263 s 1.00
Concat / HLOOpt / cpu / Primal 0.000006927140047991998 s 0.000006830919955973513 s 1.01
Concat / PartOpt / cpu / Primal 0.0000066143200274382255 s 0.000006688800240226555 s 0.99
Concat / IPartOpt / cpu / Primal 0.000006680340020466247 s 0.000006490459891210776 s 1.03
Concat / DefOpt / cpu / Primal 0.000006325279991870048 s 0.000007036420065560378 s 0.90
Concat / IDefOpt / cpu / Primal 0.000006676539969703299 s 0.0000067557199145085175 s 0.99
Concat / JaXPipe / cpu / Forward 0.000010183839949604587 s 0.000010362099965277594 s 0.98
Concat / Jax / cpu / Forward 0.0000101644799906353 s 0.000009953659937309568 s 1.02
Concat / HLOOpt / cpu / Forward 0.000009553699992466136 s 0.00000976933995843865 s 0.98
Concat / PartOpt / cpu / Forward 0.000009925180029313196 s 0.000009767280134838076 s 1.02
Concat / IPartOpt / cpu / Forward 0.000010219480036539608 s 0.000010285560092597734 s 0.99
Concat / DefOpt / cpu / Forward 0.000009551540015309002 s 0.000009621979988878594 s 0.99
Concat / IDefOpt / cpu / Forward 0.000009999760022765258 s 0.000009680279981694185 s 1.03
Concat / JaXPipe / cpu / PreRev 0.00001126398004089424 s 0.000012034819956170395 s 0.94
Concat / JaXPipe / cpu / PostRev 0.000010704339965741384 s 0.000011304079926048871 s 0.95
Concat / JaXPipe / cpu / BothRev 0.000011370240044925595 s 0.000011185099720023572 s 1.02
Concat / Jax / cpu / BothRev 0.000011071879980590892 s 0.000011534079858392944 s 0.96
Concat / HLOOpt / cpu / PreRev 0.000011769080001613477 s 0.000011462079819466452 s 1.03
Concat / HLOOpt / cpu / PostRev 0.000013192899987188866 s 0.000013215480030339678 s 1.00
Concat / HLOOpt / cpu / BothRev 0.00001161439999123104 s 0.000011233880068175494 s 1.03
Concat / PartOpt / cpu / PreRev 0.000011185499952262034 s 0.000011222520042792892 s 1.00
Concat / PartOpt / cpu / PostRev 0.000011117320027551614 s 0.000011747020071197769 s 0.95
Concat / PartOpt / cpu / BothRev 0.000011325219975333312 s 0.000012005360076727811 s 0.94
Concat / IPartOpt / cpu / PreRev 0.000011885480007549632 s 0.000011245579880778678 s 1.06
Concat / IPartOpt / cpu / PostRev 0.000011170639982083233 s 0.00001154162015154725 s 0.97
Concat / IPartOpt / cpu / BothRev 0.000011304319959890563 s 0.000011165699870616665 s 1.01
Concat / DefOpt / cpu / PreRev 0.00001165049992778222 s 0.000011291160044493152 s 1.03
Concat / DefOpt / cpu / PostRev 0.000011115980023532756 s 0.00001136698014306603 s 0.98
Concat / DefOpt / cpu / BothRev 0.000010888340011661057 s 0.00001205559976369841 s 0.90
Concat / IDefOpt / cpu / PreRev 0.000011905620012839793 s 0.000011354359921824652 s 1.05
Concat / IDefOpt / cpu / PostRev 0.000011301979966447109 s 0.00001100361994758714 s 1.03
Concat / IDefOpt / cpu / BothRev 0.000011669060004351197 s 0.000011541079984453971 s 1.01
Concat / JaXPipe / cuda / Primal 0.000002463 s 0.000002431 s 1.01
Concat / Jax / cuda / Primal 0.000002464 s 0.000002431 s 1.01
Concat / HLOOpt / cuda / Primal 0.000002432 s 0.000002431 s 1.00
Concat / PartOpt / cuda / Primal 0.000002463 s 0.000002431 s 1.01
Concat / IPartOpt / cuda / Primal 0.000002431 s 0.000002432 s 1.00
Concat / DefOpt / cuda / Primal 0.000002463 s 0.000002431 s 1.01
Concat / IDefOpt / cuda / Primal 0.000002432 s 0.000002431 s 1.00
Concat / JaXPipe / cuda / Forward 0.000010688 s 0.000010912 s 0.98
Concat / Jax / cuda / Forward 0.000010656 s 0.000010944 s 0.97
Concat / HLOOpt / cuda / Forward 0.000011136 s 0.000010496 s 1.06
Concat / PartOpt / cuda / Forward 0.000010784 s 0.000010752 s 1.00
Concat / IPartOpt / cuda / Forward 0.000010848 s 0.000010464 s 1.04
Concat / DefOpt / cuda / Forward 0.00001104 s 0.000010719 s 1.03
Concat / IDefOpt / cuda / Forward 0.000010816 s 0.000010848 s 1.00
Concat / JaXPipe / cuda / PreRev 0.00001728 s 0.000016896000000000002 s 1.02
Concat / JaXPipe / cuda / PostRev 0.00001712 s 0.000016896000000000002 s 1.01
Concat / JaXPipe / cuda / BothRev 0.0000168 s 0.00001696 s 0.99
Concat / Jax / cuda / BothRev 0.000017345 s 0.0000168 s 1.03
Concat / HLOOpt / cuda / PreRev 0.000017472 s 0.000016832 s 1.04
Concat / HLOOpt / cuda / PostRev 0.000017088 s 0.000016958999999999998 s 1.01
Concat / HLOOpt / cuda / BothRev 0.000017024 s 0.000016670999999999997 s 1.02
Concat / PartOpt / cuda / PreRev 0.00001712 s 0.000017184 s 1.00
Concat / PartOpt / cuda / PostRev 0.000017408 s 0.000016672 s 1.04
Concat / PartOpt / cuda / BothRev 0.000017344 s 0.000016927999999999998 s 1.02
Concat / IPartOpt / cuda / PreRev 0.000017184 s 0.000017088 s 1.01
Concat / IPartOpt / cuda / PostRev 0.00001728 s 0.000016607 s 1.04
Concat / IPartOpt / cuda / BothRev 0.000017312 s 0.000016768000000000003 s 1.03
Concat / DefOpt / cuda / PreRev 0.00001712 s 0.000016896000000000002 s 1.01
Concat / DefOpt / cuda / PostRev 0.000017152 s 0.000016608 s 1.03
Concat / DefOpt / cuda / BothRev 0.000016929 s 0.000016832 s 1.01
Concat / IDefOpt / cuda / PreRev 0.000017216 s 0.000017312 s 0.99
Concat / IDefOpt / cuda / PostRev 0.000016768000000000003 s 0.000017408 s 0.96
Concat / IDefOpt / cuda / BothRev 0.00001728 s 0.00001696 s 1.02
Concat / JaXPipe / tpu / Primal 0.00000152335 s 0.00000151785 s 1.00
Concat / Jax / tpu / Primal 0.000001542425 s 0.00000151125 s 1.02
Concat / HLOOpt / tpu / Primal 0.000001531725 s 0.000001529975 s 1.00
Concat / PartOpt / tpu / Primal 0.000001529025 s 0.00000151725 s 1.01
Concat / IPartOpt / tpu / Primal 0.0000015254499999999995 s 0.0000015205 s 1.00
Concat / DefOpt / tpu / Primal 0.00000152395 s 0.0000015108750000000002 s 1.01
Concat / IDefOpt / tpu / Primal 0.000001522575 s 0.0000015263 s 1.00
Concat / JaXPipe / tpu / Forward 0.0000015812999999999998 s 0.000001552875 s 1.02
Concat / Jax / tpu / Forward 0.0000015659 s 0.000001551275 s 1.01
Concat / HLOOpt / tpu / Forward 0.00000158265 s 0.000001546475 s 1.02
Concat / PartOpt / tpu / Forward 0.00000155405 s 0.000001565425 s 0.99
Concat / IPartOpt / tpu / Forward 0.000001572875 s 0.000001556725 s 1.01
Concat / DefOpt / tpu / Forward 0.00000156155 s 0.0000015515499999999998 s 1.01
Concat / IDefOpt / tpu / Forward 0.0000015759 s 0.000001565875 s 1.01
Concat / JaXPipe / tpu / PreRev 0.00000200615 s 0.000002029675 s 0.99
Concat / JaXPipe / tpu / PostRev 0.00000206715 s 0.000002016 s 1.03
Concat / JaXPipe / tpu / BothRev 0.0000019972750000000003 s 0.000002037175 s 0.98
Concat / Jax / tpu / BothRev 0.000002062275 s 0.000001993125 s 1.03
Concat / HLOOpt / tpu / PreRev 0.000001999275 s 0.0000020260500000000003 s 0.99
Concat / HLOOpt / tpu / PostRev 0.00000206455 s 0.000002003875 s 1.03
Concat / HLOOpt / tpu / BothRev 0.0000020139250000000004 s 0.000002037225 s 0.99
Concat / PartOpt / tpu / PreRev 0.00000206385 s 0.000002002225 s 1.03
Concat / PartOpt / tpu / PostRev 0.0000020095 s 0.0000020269 s 0.99
Concat / PartOpt / tpu / BothRev 0.0000020593 s 0.000002004175 s 1.03
Concat / IPartOpt / tpu / PreRev 0.000002004825 s 0.000002021725 s 0.99
Concat / IPartOpt / tpu / PostRev 0.0000020631250000000003 s 0.000001998375 s 1.03
Concat / IPartOpt / tpu / BothRev 0.000002001675 s 0.0000020281750000000003 s 0.99
Concat / DefOpt / tpu / PreRev 0.0000020721 s 0.000001993525 s 1.04
Concat / DefOpt / tpu / PostRev 0.000001994075 s 0.0000020290250000000004 s 0.98
Concat / DefOpt / tpu / BothRev 0.000002064275 s 0.0000019899 s 1.04
Concat / IDefOpt / tpu / PreRev 0.0000019979 s 0.0000020158250000000003 s 0.99
Concat / IDefOpt / tpu / PostRev 0.000002064975 s 0.0000020024 s 1.03
Concat / IDefOpt / tpu / BothRev 0.000002001275 s 0.0000020226 s 0.99
Concat / JaXPipe / cpu / Primal 0.00001284 s 0.000006810739942011423 s 1.89
Concat / Jax / cpu / Primal 0.00001278 s 0.00000690752003720263 s 1.85
Concat / HLOOpt / cpu / Primal 0.000012964 s 0.000006830919955973513 s 1.90
Concat / PartOpt / cpu / Primal 0.000013193 s 0.000006688800240226555 s 1.97
Concat / IPartOpt / cpu / Primal 0.000013143 s 0.000006490459891210776 s 2.02
Concat / DefOpt / cpu / Primal 0.000013458 s 0.000007036420065560378 s 1.91
Concat / IDefOpt / cpu / Primal 0.000012896 s 0.0000067557199145085175 s 1.91
Concat / JaXPipe / cpu / Forward 0.000018213 s 0.000010362099965277594 s 1.76
Concat / Jax / cpu / Forward 0.000017612 s 0.000009953659937309568 s 1.77
Concat / HLOOpt / cpu / Forward 0.000017639 s 0.00000976933995843865 s 1.81
Concat / PartOpt / cpu / Forward 0.000017678 s 0.000009767280134838076 s 1.81
Concat / IPartOpt / cpu / Forward 0.000017766 s 0.000010285560092597734 s 1.73
Concat / DefOpt / cpu / Forward 0.00001746 s 0.000009621979988878594 s 1.81
Concat / IDefOpt / cpu / Forward 0.000017569 s 0.000009680279981694185 s 1.81
Concat / JaXPipe / cpu / PreRev 0.000020523 s 0.000012034819956170395 s 1.71
Concat / JaXPipe / cpu / PostRev 0.000019097 s 0.000011304079926048871 s 1.69
Concat / JaXPipe / cpu / BothRev 0.000020148 s 0.000011185099720023572 s 1.80
Concat / Jax / cpu / BothRev 0.000020085 s 0.000011534079858392944 s 1.74
Concat / HLOOpt / cpu / PreRev 0.000019797 s 0.000011462079819466452 s 1.73
Concat / HLOOpt / cpu / PostRev 0.000019392 s 0.000013215480030339678 s 1.47
Concat / HLOOpt / cpu / BothRev 0.000020144 s 0.000011233880068175494 s 1.79
Concat / PartOpt / cpu / PreRev 0.000020108 s 0.000011222520042792892 s 1.79
Concat / PartOpt / cpu / PostRev 0.000019754 s 0.000011747020071197769 s 1.68
Concat / PartOpt / cpu / BothRev 0.000019933 s 0.000012005360076727811 s 1.66
Concat / IPartOpt / cpu / PreRev 0.000019959 s 0.000011245579880778678 s 1.77
Concat / IPartOpt / cpu / PostRev 0.000019629 s 0.00001154162015154725 s 1.70
Concat / IPartOpt / cpu / BothRev 0.000019989 s 0.000011165699870616665 s 1.79
Concat / DefOpt / cpu / PreRev 0.000019958 s 0.000011291160044493152 s 1.77
Concat / DefOpt / cpu / PostRev 0.000032514 s 0.00001136698014306603 s 2.86
Concat / DefOpt / cpu / BothRev 0.000020309 s 0.00001205559976369841 s 1.68
Concat / IDefOpt / cpu / PreRev 0.000020037 s 0.000011354359921824652 s 1.76
Concat / IDefOpt / cpu / PostRev 0.000019829 s 0.00001100361994758714 s 1.80
Concat / IDefOpt / cpu / BothRev 0.000020216 s 0.000011541079984453971 s 1.75
const_scatter / JaXPipe / cpu / Primal 0.000006475779982793028 s 0.000006286120042204857 s 1.03
const_scatter / Jax / cpu / Primal 0.000006203679959071451 s 0.000006213160122570116 s 1.00
const_scatter / HLOOpt / cpu / Primal 0.000007251440028994693 s 0.000007482039800379425 s 0.97
const_scatter / PartOpt / cpu / Primal 0.0000063402399973711 s 0.000006517260080727283 s 0.97
const_scatter / IPartOpt / cpu / Primal 0.00000644616004137788 s 0.0000066745399453793655 s 0.97
const_scatter / DefOpt / cpu / Primal 0.000006890359954923042 s 0.000007434799845214002 s 0.93
const_scatter / IDefOpt / cpu / Primal 0.0000071125799786386776 s 0.000007135140112950466 s 1.00
const_scatter / JaXPipe / cpu / Forward 0.000010735819951150917 s 0.000010798659968713763 s 0.99
const_scatter / Jax / cpu / Forward 0.000009027119958773257 s 0.000008980739839898889 s 1.01
const_scatter / HLOOpt / cpu / Forward 0.000010950879941447056 s 0.000010923260015260894 s 1.00
const_scatter / PartOpt / cpu / Forward 0.00001035103997310216 s 0.00001075406002200907 s 0.96
const_scatter / IPartOpt / cpu / Forward 0.000010979480039168266 s 0.00001102800011722138 s 1.00
const_scatter / DefOpt / cpu / Forward 0.0000106474999756756 s 0.000010180959907302168 s 1.05
const_scatter / IDefOpt / cpu / Forward 0.0000105772000097204 s 0.000010711239956435748 s 0.99
const_scatter / JaXPipe / cpu / PreRev 0.0002891748000274 s 0.0002903115599838 s 1.00
const_scatter / JaXPipe / cpu / PostRev 0.0002829711799859 s 0.0002942251400963 s 0.96
const_scatter / JaXPipe / cpu / BothRev 0.0002858953199938 s 0.0002862006801296 s 1.00
const_scatter / Jax / cpu / BothRev 0.0002804655800173 s 0.0002829291398666 s 0.99
const_scatter / HLOOpt / cpu / PreRev 0.000283232699985 s 0.0002868426199711 s 0.99
const_scatter / HLOOpt / cpu / PostRev 0.0003201737000108 s 0.0002869060600642 s 1.12
const_scatter / HLOOpt / cpu / BothRev 0.000283079860028 s 0.0003001028399376 s 0.94
const_scatter / PartOpt / cpu / PreRev 0.000282718760036 s 0.0002874072000849 s 0.98
const_scatter / PartOpt / cpu / PostRev 0.0002814333599781 s 0.0002842899800089 s 0.99
const_scatter / PartOpt / cpu / BothRev 0.0002842633000182 s 0.0002859568200437 s 0.99
const_scatter / IPartOpt / cpu / PreRev 0.0002845302599689 s 0.0002868205001141 s 0.99
const_scatter / IPartOpt / cpu / PostRev 0.0002792865399806 s 0.0002999288201317 s 0.93
const_scatter / IPartOpt / cpu / BothRev 0.0002827281999634 s 0.0002863051198437 s 0.99
const_scatter / DefOpt / cpu / PreRev 0.0002973312000267 s 0.000284554440077 s 1.04
const_scatter / DefOpt / cpu / PostRev 0.0002825952600414 s 0.000285246339954 s 0.99
const_scatter / DefOpt / cpu / BothRev 0.0002797493400339 s 0.0002855436599202 s 0.98
const_scatter / IDefOpt / cpu / PreRev 0.0002827880599943 s 0.0002843137198942 s 0.99
const_scatter / IDefOpt / cpu / PostRev 0.0002823460400031 s 0.0002881772199543 s 0.98
const_scatter / IDefOpt / cpu / BothRev 0.0002871784999661 s 0.0002883722399201 s 1.00
const_scatter / JaXPipe / cuda / Primal 0.000002463 s 0.000002431 s 1.01
const_scatter / Jax / cuda / Primal 0.000002463 s 0.000002431 s 1.01
const_scatter / HLOOpt / cuda / Primal 0.000002463 s 0.000002432 s 1.01
const_scatter / PartOpt / cuda / Primal 0.000002463 s 0.000002432 s 1.01
const_scatter / IPartOpt / cuda / Primal 0.000002463 s 0.000002431 s 1.01
const_scatter / DefOpt / cuda / Primal 0.000002463 s 0.000002431 s 1.01
const_scatter / IDefOpt / cuda / Primal 0.000002463 s 0.000002431 s 1.01
const_scatter / JaXPipe / cuda / Forward 0.00001088 s 0.000010624 s 1.02
const_scatter / Jax / cuda / Forward 0.000011072 s 0.000010847 s 1.02
const_scatter / HLOOpt / cuda / Forward 0.000011232 s 0.000010496 s 1.07
const_scatter / PartOpt / cuda / Forward 0.000010496 s 0.000010464 s 1.00
const_scatter / IPartOpt / cuda / Forward 0.000011168 s 0.000011168 s 1
const_scatter / DefOpt / cuda / Forward 0.00001088 s 0.000010687 s 1.02
const_scatter / IDefOpt / cuda / Forward 0.000010656 s 0.00001088 s 0.98
const_scatter / JaXPipe / cuda / PreRev 0.000017503999999999997 s 0.00001696 s 1.03
const_scatter / JaXPipe / cuda / PostRev 0.000017056 s 0.000017024 s 1.00
const_scatter / JaXPipe / cuda / BothRev 0.000017664 s 0.0000184 s 0.96
const_scatter / Jax / cuda / BothRev 0.000017344 s 0.000016639 s 1.04
const_scatter / HLOOpt / cuda / PreRev 0.000016895 s 0.000016831 s 1.00
const_scatter / HLOOpt / cuda / PostRev 0.000017216 s 0.000017344 s 0.99
const_scatter / HLOOpt / cuda / BothRev 0.000019232 s 0.000017024 s 1.13
const_scatter / PartOpt / cuda / PreRev 0.00001744 s 0.000017152 s 1.02
const_scatter / PartOpt / cuda / PostRev 0.000017503999999999997 s 0.00001728 s 1.01
const_scatter / PartOpt / cuda / BothRev 0.000018304 s 0.000017632 s 1.04
const_scatter / IPartOpt / cuda / PreRev 0.0000176 s 0.000016544 s 1.06
const_scatter / IPartOpt / cuda / PostRev 0.000017152 s 0.000016673 s 1.03
const_scatter / IPartOpt / cuda / BothRev 0.000017056 s 0.000016992 s 1.00
const_scatter / DefOpt / cuda / PreRev 0.000017632 s 0.000016832 s 1.05
const_scatter / DefOpt / cuda / PostRev 0.00001728 s 0.00001728 s 1
const_scatter / DefOpt / cuda / BothRev 0.000017344 s 0.000017536 s 0.99
const_scatter / IDefOpt / cuda / PreRev 0.000017632 s 0.000017534999999999997 s 1.01
const_scatter / IDefOpt / cuda / PostRev 0.000017249 s 0.000017536 s 0.98
const_scatter / IDefOpt / cuda / BothRev 0.00001744 s 0.0000176 s 0.99
const_scatter / JaXPipe / tpu / Primal 0.0000038054 s 0.00000379845 s 1.00
const_scatter / Jax / tpu / Primal 0.000003818625 s 0.000003824975 s 1.00
const_scatter / HLOOpt / tpu / Primal 0.000003782175 s 0.00000380035 s 1.00
const_scatter / PartOpt / tpu / Primal 0.000003810975 s 0.000003851725 s 0.99
const_scatter / IPartOpt / tpu / Primal 0.000003799325 s 0.000003791925 s 1.00
const_scatter / DefOpt / tpu / Primal 0.000003830575 s 0.000003819925 s 1.00
const_scatter / IDefOpt / tpu / Primal 0.000003814100000000001 s 0.00000378655 s 1.01
const_scatter / JaXPipe / tpu / Forward 0.0000064681 s 0.000006479 s 1.00
const_scatter / Jax / tpu / Forward 0.000006506775 s 0.00000648265 s 1.00
const_scatter / HLOOpt / tpu / Forward 0.000006485125 s 0.0000064919000000000005 s 1.00
const_scatter / PartOpt / tpu / Forward 0.0000064809 s 0.00000646155 s 1.00
const_scatter / IPartOpt / tpu / Forward 0.00000647215 s 0.000006509575 s 0.99
const_scatter / DefOpt / tpu / Forward 0.000006481725000000001 s 0.00000646095 s 1.00
const_scatter / IDefOpt / tpu / Forward 0.000006473125 s 0.00000648355 s 1.00
const_scatter / JaXPipe / tpu / PreRev 0.0000066978000000000005 s 0.000006650225 s 1.01
const_scatter / JaXPipe / tpu / PostRev 0.000006666975000000001 s 0.0000066379250000000005 s 1.00
const_scatter / JaXPipe / tpu / BothRev 0.0000066444 s 0.000006636999999999999 s 1.00
const_scatter / Jax / tpu / BothRev 0.000006644099999999999 s 0.000006643 s 1.00
const_scatter / HLOOpt / tpu / PreRev 0.000006678575 s 0.00000664195 s 1.01
const_scatter / HLOOpt / tpu / PostRev 0.000006644949999999999 s 0.00000664285 s 1.00
const_scatter / HLOOpt / tpu / BothRev 0.000006663125 s 0.000006624949999999999 s 1.01
const_scatter / PartOpt / tpu / PreRev 0.00000665045 s 0.000006633075 s 1.00
const_scatter / PartOpt / tpu / PostRev 0.000006662850000000001 s 0.00000661795 s 1.01
const_scatter / PartOpt / tpu / BothRev 0.00000667025 s 0.0000066514 s 1.00
const_scatter / IPartOpt / tpu / PreRev 0.000006641025 s 0.000006649350000000001 s 1.00
const_scatter / IPartOpt / tpu / PostRev 0.000006663900000000001 s 0.0000066319 s 1.00
const_scatter / IPartOpt / tpu / BothRev 0.000006649825 s 0.000006648775 s 1.00
const_scatter / DefOpt / tpu / PreRev 0.0000066514 s 0.00000664805 s 1.00
const_scatter / DefOpt / tpu / PostRev 0.0000066691 s 0.000006630575 s 1.01
const_scatter / DefOpt / tpu / BothRev 0.000006646025 s 0.000006650775 s 1.00
const_scatter / IDefOpt / tpu / PreRev 0.0000066491 s 0.000006623449999999999 s 1.00
const_scatter / IDefOpt / tpu / PostRev 0.000006645425 s 0.000006627374999999999 s 1.00
const_scatter / IDefOpt / tpu / BothRev 0.000006681824999999999 s 0.000006617875 s 1.01
const_scatter / JaXPipe / cpu / Primal 0.000013155 s 0.000006286120042204857 s 2.09
const_scatter / Jax / cpu / Primal 0.000012858 s 0.000006213160122570116 s 2.07
const_scatter / HLOOpt / cpu / Primal 0.000013416000000000002 s 0.000007482039800379425 s 1.79
const_scatter / PartOpt / cpu / Primal 0.000012822 s 0.000006517260080727283 s 1.97
const_scatter / IPartOpt / cpu / Primal 0.000013096999999999998 s 0.0000066745399453793655 s 1.96
const_scatter / DefOpt / cpu / Primal 0.000013619 s 0.000007434799845214002 s 1.83
const_scatter / IDefOpt / cpu / Primal 0.00001335 s 0.000007135140112950466 s 1.87
const_scatter / JaXPipe / cpu / Forward 0.000018037 s 0.000010798659968713763 s 1.67
const_scatter / Jax / cpu / Forward 0.00001698 s 0.000008980739839898889 s 1.89
const_scatter / HLOOpt / cpu / Forward 0.000018241 s 0.000010923260015260894 s 1.67
const_scatter / PartOpt / cpu / Forward 0.000017745 s 0.00001075406002200907 s 1.65
const_scatter / IPartOpt / cpu / Forward 0.000017915 s 0.00001102800011722138 s 1.62
const_scatter / DefOpt / cpu / Forward 0.000018022 s 0.000010180959907302168 s 1.77
const_scatter / IDefOpt / cpu / Forward 0.000017807 s 0.000010711239956435748 s 1.66
const_scatter / JaXPipe / cpu / PreRev 0.000542264 s 0.0002903115599838 s 1.87
const_scatter / JaXPipe / cpu / PostRev 0.00052428 s 0.0002942251400963 s 1.78
const_scatter / JaXPipe / cpu / BothRev 0.0005249979999999 s 0.0002862006801296 s 1.83
const_scatter / Jax / cpu / BothRev 0.000528863 s 0.0002829291398666 s 1.87
const_scatter / HLOOpt / cpu / PreRev 0.000524897 s 0.0002868426199711 s 1.83
const_scatter / HLOOpt / cpu / PostRev 0.000524954 s 0.0002869060600642 s 1.83
const_scatter / HLOOpt / cpu / BothRev 0.000525206 s 0.0003001028399376 s 1.75
const_scatter / PartOpt / cpu / PreRev 0.000535813 s 0.0002874072000849 s 1.86
const_scatter / PartOpt / cpu / PostRev 0.0005064889999999 s 0.0002842899800089 s 1.78
const_scatter / PartOpt / cpu / BothRev 0.0005216239999999 s 0.0002859568200437 s 1.82
const_scatter / IPartOpt / cpu / PreRev 0.000535832 s 0.0002868205001141 s 1.87
const_scatter / IPartOpt / cpu / PostRev 0.0005254529999999 s 0.0002999288201317 s 1.75
const_scatter / IPartOpt / cpu / BothRev 0.000515669 s 0.0002863051198437 s 1.80
const_scatter / DefOpt / cpu / PreRev 0.000507963 s 0.000284554440077 s 1.79
const_scatter / DefOpt / cpu / PostRev 0.000526204 s 0.000285246339954 s 1.84
const_scatter / DefOpt / cpu / BothRev 0.000518865 s 0.0002855436599202 s 1.82
const_scatter / IDefOpt / cpu / PreRev 0.000531266 s 0.0002843137198942 s 1.87
const_scatter / IDefOpt / cpu / PostRev 0.000525134 s 0.0002881772199543 s 1.82
const_scatter / IDefOpt / cpu / BothRev 0.00053395 s 0.0002883722399201 s 1.85
GenDot / JaXPipe / cpu / Primal 0.000007058640039758757 s 0.000007314039976336062 s 0.97
GenDot / Jax / cpu / Primal 0.000007158180023907335 s 0.00000663925995468162 s 1.08
GenDot / HLOOpt / cpu / Primal 0.000007439219980369671 s 0.000007459320186171681 s 1.00
GenDot / PartOpt / cpu / Primal 0.000006639479970544926 s 0.000006850139907328412 s 0.97
GenDot / IPartOpt / cpu / Primal 0.000007283680006366922 s 0.0000070201801645453086 s 1.04
GenDot / DefOpt / cpu / Primal 0.000007094319980751606 s 0.00000712750006641727 s 1.00
GenDot / IDefOpt / cpu / Primal 0.000007065799982228782 s 0.000007662799871468451 s 0.92
GenDot / JaXPipe / cpu / Forward 0.000011142140001538792 s 0.000011181799964106176 s 1.00
GenDot / Jax / cpu / Forward 0.000010192559984716356 s 0.000010393080046924296 s 0.98
GenDot / HLOOpt / cpu / Forward 0.000011260099981882376 s 0.000011051219880755523 s 1.02
GenDot / PartOpt / cpu / Forward 0.000010684959997888654 s 0.000010293439991073684 s 1.04
GenDot / IPartOpt / cpu / Forward 0.00001114266000513453 s 0.000011467720323707908 s 0.97
GenDot / DefOpt / cpu / Forward 0.000010522500033403049 s 0.000010869840079976712 s 0.97
GenDot / IDefOpt / cpu / Forward 0.000010230900024907895 s 0.00001090998004656285 s 0.94
GenDot / JaXPipe / cpu / PreRev 0.00001133010000557988 s 0.00001122697998653166 s 1.01
GenDot / JaXPipe / cpu / PostRev 0.00000938481997764029 s 0.000010410219911136664 s 0.90
GenDot / JaXPipe / cpu / BothRev 0.000010959040009765886 s 0.000010769860055006574 s 1.02
GenDot / Jax / cpu / BothRev 0.000010252279962514876 s 0.000010883420036407188 s 0.94
GenDot / HLOOpt / cpu / PreRev 0.000011009320023731562 s 0.000010963120112137404 s 1.00
GenDot / HLOOpt / cpu / PostRev 0.00001289864003410912 s 0.00001305949997913558 s 0.99
GenDot / HLOOpt / cpu / BothRev 0.00001051152002219169 s 0.000011048319975088816 s 0.95
GenDot / PartOpt / cpu / PreRev 0.000010857540037250146 s 0.000011295960139250382 s 0.96
GenDot / PartOpt / cpu / PostRev 0.000009799860008570249 s 0.000010292840015608817 s 0.95
GenDot / PartOpt / cpu / BothRev 0.000010963799977616871 s 0.000010914960075751878 s 1.00
GenDot / IPartOpt / cpu / PreRev 0.000010451199968883885 s 0.000011207640054635704 s 0.93
GenDot / IPartOpt / cpu / PostRev 0.000009857459999693676 s 0.000010758139869722073 s 0.92
GenDot / IPartOpt / cpu / BothRev 0.000011122760024591117 s 0.000010854679858312011 s 1.02
GenDot / DefOpt / cpu / PreRev 0.000010702600047807209 s 0.000010493860172573476 s 1.02
GenDot / DefOpt / cpu / PostRev 0.000010777740017147152 s 0.000011142780094814952 s 0.97
GenDot / DefOpt / cpu / BothRev 0.000010920439999608788 s 0.000010720419995777774 s 1.02
GenDot / IDefOpt / cpu / PreRev 0.000010280499973305268 s 0.00001113910027925158 s 0.92
GenDot / IDefOpt / cpu / PostRev 0.000010718200046540004 s 0.000011556879871932325 s 0.93
GenDot / IDefOpt / cpu / BothRev 0.000010665100035112118 s 0.000010867720084206666 s 0.98
GenDot / JaXPipe / cuda / Primal 0.000002528 s 0.000002528 s 1
GenDot / Jax / cuda / Primal 0.000002527 s 0.000002527 s 1
GenDot / HLOOpt / cuda / Primal 0.000002496 s 0.000002527 s 0.99
GenDot / PartOpt / cuda / Primal 0.000002528 s 0.000002527 s 1.00
GenDot / IPartOpt / cuda / Primal 0.000002528 s 0.000002528 s 1
GenDot / DefOpt / cuda / Primal 0.000002527 s 0.000002527 s 1
GenDot / IDefOpt / cuda / Primal 0.000002527 s 0.000002527 s 1
GenDot / JaXPipe / cuda / Forward 0.000010624 s 0.000012224 s 0.87
GenDot / Jax / cuda / Forward 0.000010848 s 0.000010528 s 1.03
GenDot / HLOOpt / cuda / Forward 0.000010689 s 0.000010687 s 1.00
GenDot / PartOpt / cuda / Forward 0.000011520000000000002 s 0.000010464 s 1.10
GenDot / IPartOpt / cuda / Forward 0.00000992 s 0.000010752 s 0.92
GenDot / DefOpt / cuda / Forward 0.000010592 s 0.000012064 s 0.88
GenDot / IDefOpt / cuda / Forward 0.000010944 s 0.000012224 s 0.90
GenDot / JaXPipe / cuda / PreRev 0.000010624 s 0.000012288 s 0.86
GenDot / JaXPipe / cuda / PostRev 0.000010688 s 0.000010656 s 1.00
GenDot / JaXPipe / cuda / BothRev 0.00001072 s 0.00001072 s 1
GenDot / Jax / cuda / BothRev 0.0000112 s 0.000010688 s 1.05
GenDot / HLOOpt / cuda / PreRev 0.000010784 s 0.000010752 s 1.00
GenDot / HLOOpt / cuda / PostRev 0.00001072 s 0.000010496 s 1.02
GenDot / HLOOpt / cuda / BothRev 0.000010432 s 0.000010849 s 0.96
GenDot / PartOpt / cuda / PreRev 0.000010848 s 0.000011071 s 0.98
GenDot / PartOpt / cuda / PostRev 0.000010624 s 0.000010624 s 1
GenDot / PartOpt / cuda / BothRev 0.000010784 s 0.000010592 s 1.02
GenDot / IPartOpt / cuda / PreRev 0.000010592 s 0.000010848 s 0.98
GenDot / IPartOpt / cuda / PostRev 0.000010912 s 0.00001072 s 1.02
GenDot / IPartOpt / cuda / BothRev 0.000010912 s 0.00001088 s 1.00
GenDot / DefOpt / cuda / PreRev 0.000011168 s 0.000010784 s 1.04
GenDot / DefOpt / cuda / PostRev 0.000010847 s 0.000010784 s 1.01
GenDot / DefOpt / cuda / BothRev 0.000010816 s 0.000010656 s 1.02
GenDot / IDefOpt / cuda / PreRev 0.000010623 s 0.000010623 s 1
GenDot / IDefOpt / cuda / PostRev 0.000010688 s 0.000010848 s 0.99
GenDot / IDefOpt / cuda / BothRev 0.000011039 s 0.000010656 s 1.04
GenDot / JaXPipe / tpu / Primal 9.2685e-7 s 9.43175e-7 s 0.98
GenDot / Jax / tpu / Primal 9.25575e-7 s 9.3e-7 s 1.00
GenDot / HLOOpt / tpu / Primal 0.000001557925 s 0.0000016040750000000002 s 0.97
GenDot / PartOpt / tpu / Primal 9.25925e-7 s 9.303e-7 s 1.00
GenDot / IPartOpt / tpu / Primal 9.26075e-7 s 9.436e-7 s 0.98
GenDot / DefOpt / tpu / Primal 0.0000014994 s 0.0000014983 s 1.00
GenDot / IDefOpt / tpu / Primal 0.000001557325 s 0.000001602275 s 0.97
GenDot / JaXPipe / tpu / Forward 0.000003170625 s 0.0000030512 s 1.04
GenDot / Jax / tpu / Forward 0.000002311125 s 0.000002273475 s 1.02
GenDot / HLOOpt / tpu / Forward 0.000003124975 s 0.0000031171500000000003 s 1.00
GenDot / PartOpt / tpu / Forward 0.000003226675 s 0.0000031349 s 1.03
GenDot / IPartOpt / tpu / Forward 0.000003113575 s 0.00000311855 s 1.00
GenDot / DefOpt / tpu / Forward 0.00000321945 s 0.000003145825 s 1.02
GenDot / IDefOpt / tpu / Forward 0.0000031173 s 0.000003114475 s 1.00
GenDot / JaXPipe / tpu / PreRev 0.000002957 s 0.0000030383 s 0.97
GenDot / JaXPipe / tpu / PostRev 0.000002416325 s 0.0000023743000000000003 s 1.02
GenDot / JaXPipe / tpu / BothRev 0.0000029558 s 0.00000302605 s 0.98
GenDot / Jax / tpu / BothRev 0.000002405025 s 0.000002377425 s 1.01
GenDot / HLOOpt / tpu / PreRev 0.00000296925 s 0.000003033975 s 0.98
GenDot / HLOOpt / tpu / PostRev 0.0000029382750000000003 s 0.000002945275 s 1.00
GenDot / HLOOpt / tpu / BothRev 0.000002951775 s 0.00000302865 s 0.97
GenDot / PartOpt / tpu / PreRev 0.0000029365999999999995 s 0.0000029443000000000003 s 1.00
GenDot / PartOpt / tpu / PostRev 0.00000239175 s 0.000002416025 s 0.99
GenDot / PartOpt / tpu / BothRev 0.0000029344 s 0.000002944575 s 1.00
GenDot / IPartOpt / tpu / PreRev 0.0000029613250000000004 s 0.000003024925 s 0.98
GenDot / IPartOpt / tpu / PostRev 0.0000024087000000000003 s 0.00000238245 s 1.01
GenDot / IPartOpt / tpu / BothRev 0.0000029529 s 0.00000302485 s 0.98
GenDot / DefOpt / tpu / PreRev 0.0000029382 s 0.00000297045 s 0.99
GenDot / DefOpt / tpu / PostRev 0.00000295465 s 0.0000030201 s 0.98
GenDot / DefOpt / tpu / BothRev 0.000002934975 s 0.000002939575 s 1.00
GenDot / IDefOpt / tpu / PreRev 0.0000029626 s 0.00000301605 s 0.98
GenDot / IDefOpt / tpu / PostRev 0.000002929725 s 0.000002966875 s 0.99
GenDot / IDefOpt / tpu / BothRev 0.00000296775 s 0.000003017975 s 0.98
GenDot / JaXPipe / cpu / Primal 0.000014994 s 0.000007314039976336062 s 2.05
GenDot / Jax / cpu / Primal 0.000015317 s 0.00000663925995468162 s 2.31
GenDot / HLOOpt / cpu / Primal 0.000014223 s 0.000007459320186171681 s 1.91
GenDot / PartOpt / cpu / Primal 0.000015079 s 0.000006850139907328412 s 2.20
GenDot / IPartOpt / cpu / Primal 0.000014777 s 0.0000070201801645453086 s 2.10
GenDot / DefOpt / cpu / Primal 0.000014128 s 0.00000712750006641727 s 1.98
GenDot / IDefOpt / cpu / Primal 0.000014334 s 0.000007662799871468451 s 1.87
GenDot / JaXPipe / cpu / Forward 0.000019377 s 0.000011181799964106176 s 1.73
GenDot / Jax / cpu / Forward 0.000020238 s 0.000010393080046924296 s 1.95
GenDot / HLOOpt / cpu / Forward 0.000019147 s 0.000011051219880755523 s 1.73
GenDot / PartOpt / cpu / Forward 0.000019468 s 0.000010293439991073684 s 1.89
GenDot / IPartOpt / cpu / Forward 0.000019482 s 0.000011467720323707908 s 1.70
GenDot / DefOpt / cpu / Forward 0.000019869 s 0.000010869840079976712 s 1.83
GenDot / IDefOpt / cpu / Forward 0.000019446 s 0.00001090998004656285 s 1.78
GenDot / JaXPipe / cpu / PreRev 0.000019747 s 0.00001122697998653166 s 1.76
GenDot / JaXPipe / cpu / PostRev 0.000020508 s 0.000010410219911136664 s 1.97
GenDot / JaXPipe / cpu / BothRev 0.000019463 s 0.000010769860055006574 s 1.81
GenDot / Jax / cpu / BothRev 0.000021255 s 0.000010883420036407188 s 1.95
GenDot / HLOOpt / cpu / PreRev 0.000019659 s 0.000010963120112137404 s 1.79
GenDot / HLOOpt / cpu / PostRev 0.000019426 s 0.00001305949997913558 s 1.49
GenDot / HLOOpt / cpu / BothRev 0.000019623 s 0.000011048319975088816 s 1.78
GenDot / PartOpt / cpu / PreRev 0.000019093 s 0.000011295960139250382 s 1.69
GenDot / PartOpt / cpu / PostRev 0.00002089 s 0.000010292840015608817 s 2.03
GenDot / PartOpt / cpu / BothRev 0.000019675 s 0.000010914960075751878 s 1.80
GenDot / IPartOpt / cpu / PreRev 0.000019429 s 0.000011207640054635704 s 1.73
GenDot / IPartOpt / cpu / PostRev 0.000020497 s 0.000010758139869722073 s 1.91
GenDot / IPartOpt / cpu / BothRev 0.000019604 s 0.000010854679858312011 s 1.81
GenDot / DefOpt / cpu / PreRev 0.000019313 s 0.000010493860172573476 s 1.84
GenDot / DefOpt / cpu / PostRev 0.000019731 s 0.000011142780094814952 s 1.77
GenDot / DefOpt / cpu / BothRev 0.00001981 s 0.000010720419995777774 s 1.85
GenDot / IDefOpt / cpu / PreRev 0.000019317 s 0.00001113910027925158 s 1.73
GenDot / IDefOpt / cpu / PostRev 0.000019795 s 0.000011556879871932325 s 1.71
GenDot / IDefOpt / cpu / BothRev 0.000019768 s 0.000010867720084206666 s 1.82
hlo_ffi / JaXPipe / cpu / Primal 0.000011248860027990304 s 0.000011581580110942014 s 0.97
hlo_ffi / Jax / cpu / Primal 0.000010887139997066697 s 0.000011033399787265809 s 0.99
hlo_ffi / HLOOpt / cpu / Primal 0.000011678839946398512 s 0.000011139259877381846 s 1.05
hlo_ffi / PartOpt / cpu / Primal 0.000011032420015908428 s 0.00001092610018531559 s 1.01
hlo_ffi / IPartOpt / cpu / Primal 0.000011074700041717734 s 0.000011699480055540334 s 0.95
hlo_ffi / DefOpt / cpu / Primal 0.00001087645998268272 s 0.000011188660319021437 s 0.97
hlo_ffi / IDefOpt / cpu / Primal 0.00001081174003047636 s 0.000011229419942537787 s 0.96
hlo_ffi / JaXPipe / cpu / Forward 0.000015464319976672413 s 0.00001648509998631198 s 0.94
hlo_ffi / Jax / cpu / Forward 0.000016157620011654215 s 0.0000156106400754652 s 1.04
hlo_ffi / HLOOpt / cpu / Forward 0.000015860419989621732 s 0.000016109379939734935 s 0.98
hlo_ffi / PartOpt / cpu / Forward 0.00001607590001185599 s 0.00001583156001288444 s 1.02
hlo_ffi / IPartOpt / cpu / Forward 0.000016245139995589853 s 0.00001557716004754184 s 1.04
hlo_ffi / DefOpt / cpu / Forward 0.000015877420009928757 s 0.00001562955996632809 s 1.02
hlo_ffi / IDefOpt / cpu / Forward 0.000017157500005851036 s 0.000015224300077534282 s 1.13
hlo_ffi / JaXPipe / cpu / PreRev 0.00001564812001561222 s 0.00001584945996000897 s 0.99
hlo_ffi / JaXPipe / cpu / PostRev 0.000014765619989702828 s 0.00001478201989812078 s 1.00
hlo_ffi / JaXPipe / cpu / BothRev 0.000014673519999632844 s 0.000015412480024679098 s 0.95
hlo_ffi / Jax / cpu / BothRev 0.000015155280016188044 s 0.000016434639983344825 s 0.92
hlo_ffi / HLOOpt / cpu / PreRev 0.000016049400001065807 s 0.000015840260020922868 s 1.01
hlo_ffi / HLOOpt / cpu / PostRev 0.000016979560014078744 s 0.00001723093995678937 s 0.99
hlo_ffi / HLOOpt / cpu / BothRev 0.000014861960034977528 s 0.000015616500058968084 s 0.95
hlo_ffi / PartOpt / cpu / PreRev 0.00001615130001482612 s 0.000017233819999091793 s 0.94
hlo_ffi / PartOpt / cpu / PostRev 0.000015035579999675972 s 0.00001562794001074508 s 0.96
hlo_ffi / PartOpt / cpu / BothRev 0.00001524263998362585 s 0.000015728639846201985 s 0.97
hlo_ffi / IPartOpt / cpu / PreRev 0.00001541059998999117 s 0.000016466500019305384 s 0.94
hlo_ffi / IPartOpt / cpu / PostRev 0.000014859400007480872 s 0.000015306339992093854 s 0.97
hlo_ffi / IPartOpt / cpu / BothRev 0.000014832500000920846 s 0.000015431739957421086 s 0.96
hlo_ffi / DefOpt / cpu / PreRev 0.00001579862007019983 s 0.000016104480055219027 s 0.98
hlo_ffi / DefOpt / cpu / PostRev 0.000015516859939452843 s 0.0000156216798859532 s 0.99
hlo_ffi / DefOpt / cpu / BothRev 0.000015102860015758778 s 0.000015211660211207345 s 0.99
hlo_ffi / IDefOpt / cpu / PreRev 0.00001593401998434274 s 0.00001618880000023637 s 0.98
hlo_ffi / IDefOpt / cpu / PostRev 0.000015209540060823202 s 0.000015539559790340718 s 0.98
hlo_ffi / IDefOpt / cpu / BothRev 0.00001575933999447443 s 0.00001603133998287376 s 0.98
hlo_ffi / JaXPipe / cuda / Primal 0.0000023670000000000004 s 0.000002368 s 1.00
hlo_ffi / Jax / cuda / Primal 0.0000023670000000000004 s 0.000002368 s 1.00
hlo_ffi / HLOOpt / cuda / Primal 0.0000023670000000000004 s 0.000002368 s 1.00
hlo_ffi / PartOpt / cuda / Primal 0.000002368 s 0.000002368 s 1
hlo_ffi / IPartOpt / cuda / Primal 0.0000023670000000000004 s 0.000002368 s 1.00
hlo_ffi / DefOpt / cuda / Primal 0.000002368 s 0.0000023670000000000004 s 1.00
hlo_ffi / IDefOpt / cuda / Primal 0.0000023670000000000004 s 0.000002369 s 1.00
hlo_ffi / JaXPipe / cuda / Forward 0.000002463 s 0.000002463 s 1
hlo_ffi / Jax / cuda / Forward 0.000002463 s 0.000002463 s 1
hlo_ffi / HLOOpt / cuda / Forward 0.000002463 s 0.000002463 s 1
hlo_ffi / PartOpt / cuda / Forward 0.000002463 s 0.000002463 s 1
hlo_ffi / IPartOpt / cuda / Forward 0.000002463 s 0.000002463 s 1
hlo_ffi / DefOpt / cuda / Forward 0.000002463 s 0.000002463 s 1
hlo_ffi / IDefOpt / cuda / Forward 0.000002463 s 0.000002464 s 1.00
hlo_ffi / JaXPipe / cuda / PreRev 0.000002432 s 0.000002432 s 1
hlo_ffi / JaXPipe / cuda / PostRev 0.000002432 s 0.000002463 s 0.99
hlo_ffi / JaXPipe / cuda / BothRev 0.000002431 s 0.000002463 s 0.99
hlo_ffi / Jax / cuda / BothRev 0.000002432 s 0.000002432 s 1
hlo_ffi / HLOOpt / cuda / PreRev 0.000002432 s 0.000002433 s 1.00
hlo_ffi / HLOOpt / cuda / PostRev 0.000002432 s 0.000002463 s 0.99
hlo_ffi / HLOOpt / cuda / BothRev 0.000002432 s 0.000002463 s 0.99
hlo_ffi / PartOpt / cuda / PreRev 0.000002431 s 0.000002432 s 1.00
hlo_ffi / PartOpt / cuda / PostRev 0.000002431 s 0.000002463 s 0.99
hlo_ffi / PartOpt / cuda / BothRev 0.000002432 s 0.000002432 s 1
hlo_ffi / IPartOpt / cuda / PreRev 0.000002432 s 0.000002463 s 0.99
hlo_ffi / IPartOpt / cuda / PostRev 0.000002432 s 0.000002432 s 1
hlo_ffi / IPartOpt / cuda / BothRev 0.000002431 s 0.000002463 s 0.99
hlo_ffi / DefOpt / cuda / PreRev 0.000002432 s 0.000002432 s 1
hlo_ffi / DefOpt / cuda / PostRev 0.000002432 s 0.000002432 s 1
hlo_ffi / DefOpt / cuda / BothRev 0.000002432 s 0.000002432 s 1
hlo_ffi / IDefOpt / cuda / PreRev 0.000002431 s 0.000002432 s 1.00
hlo_ffi / IDefOpt / cuda / PostRev 0.000002433 s 0.000002432 s 1.00
hlo_ffi / IDefOpt / cuda / BothRev 0.000002432 s 0.000002432 s 1
hlo_ffi / JaXPipe / tpu / Primal 9.27725e-7 s 9.07525e-7 s 1.02
hlo_ffi / Jax / tpu / Primal 9.533e-7 s 9.7945e-7 s 0.97
hlo_ffi / HLOOpt / tpu / Primal 9.0905e-7 s 9.32225e-7 s 0.98
hlo_ffi / PartOpt / tpu / Primal 9.50925e-7 s 9.72375e-7 s 0.98
hlo_ffi / IPartOpt / tpu / Primal 9.09225e-7 s 9.4025e-7 s 0.97
hlo_ffi / DefOpt / tpu / Primal 9.571e-7 s 9.73025e-7 s 0.98
hlo_ffi / IDefOpt / tpu / Primal 9.08225e-7 s 9.38325e-7 s 0.97
hlo_ffi / JaXPipe / tpu / Forward 9.48975e-7 s 9.494e-7 s 1.00
hlo_ffi / Jax / tpu / Forward 9.8135e-7 s 9.8205e-7 s 1.00
hlo_ffi / HLOOpt / tpu / Forward 9.743499999999998e-7 s 9.744e-7 s 1.00
hlo_ffi / PartOpt / tpu / Forward 9.3415e-7 s 9.5955e-7 s 0.97
hlo_ffi / IPartOpt / tpu / Forward 9.74025e-7 s 9.74075e-7 s 1.00
hlo_ffi / DefOpt / tpu / Forward 9.335e-7 s 9.59675e-7 s 0.97
hlo_ffi / IDefOpt / tpu / Forward 9.7445e-7 s 9.74475e-7 s 1.00
hlo_ffi / JaXPipe / tpu / PreRev 9.38475e-7 s 9.481e-7 s 0.99
hlo_ffi / JaXPipe / tpu / PostRev 9.6545e-7 s 9.65425e-7 s 1.00
hlo_ffi / JaXPipe / tpu / BothRev 9.62225e-7 s 9.942499999999998e-7 s 0.97
hlo_ffi / Jax / tpu / BothRev 9.64425e-7 s 9.64775e-7 s 1.00
hlo_ffi / HLOOpt / tpu / PreRev 9.6195e-7 s 9.950249999999998e-7 s 0.97
hlo_ffi / HLOOpt / tpu / PostRev 9.648e-7 s 9.65125e-7 s 1.00
hlo_ffi / HLOOpt / tpu / BothRev 9.61375e-7 s 9.94975e-7 s 0.97
hlo_ffi / PartOpt / tpu / PreRev 9.649e-7 s 9.64675e-7 s 1.00
hlo_ffi / PartOpt / tpu / PostRev 9.612999999999998e-7 s 9.9515e-7 s 0.97
hlo_ffi / PartOpt / tpu / BothRev 9.648e-7 s 9.64725e-7 s 1.00
hlo_ffi / IPartOpt / tpu / PreRev 9.61725e-7 s 9.946249999999998e-7 s 0.97
hlo_ffi / IPartOpt / tpu / PostRev 9.651e-7 s 9.6535e-7 s 1.00
hlo_ffi / IPartOpt / tpu / BothRev 9.625e-7 s 9.95075e-7 s 0.97
hlo_ffi / DefOpt / tpu / PreRev 9.64825e-7 s 9.64825e-7 s 1
hlo_ffi / DefOpt / tpu / PostRev 9.618500000000002e-7 s 9.948e-7 s 0.97
hlo_ffi / DefOpt / tpu / BothRev 9.64825e-7 s 9.654e-7 s 1.00
hlo_ffi / IDefOpt / tpu / PreRev 9.6155e-7 s 9.9475e-7 s 0.97
hlo_ffi / IDefOpt / tpu / PostRev 9.64675e-7 s 9.646e-7 s 1.00
hlo_ffi / IDefOpt / tpu / BothRev 9.62175e-7 s 9.949e-7 s 0.97
hlo_ffi / JaXPipe / cpu / Primal 0.000018173 s 0.000011581580110942014 s 1.57
hlo_ffi / Jax / cpu / Primal 0.000017526 s 0.000011033399787265809 s 1.59
hlo_ffi / HLOOpt / cpu / Primal 0.000017415 s 0.000011139259877381846 s 1.56
hlo_ffi / PartOpt / cpu / Primal 0.000017746999999999998 s 0.00001092610018531559 s 1.62
hlo_ffi / IPartOpt / cpu / Primal 0.000017413 s 0.000011699480055540334 s 1.49
hlo_ffi / DefOpt / cpu / Primal 0.000017848 s 0.000011188660319021437 s 1.60
hlo_ffi / IDefOpt / cpu / Primal 0.000017776 s 0.000011229419942537787 s 1.58
hlo_ffi / JaXPipe / cpu / Forward 0.000025084 s 0.00001648509998631198 s 1.52
hlo_ffi / Jax / cpu / Forward 0.00002387 s 0.0000156106400754652 s 1.53
hlo_ffi / HLOOpt / cpu / Forward 0.000023866 s 0.000016109379939734935 s 1.48
hlo_ffi / PartOpt / cpu / Forward 0.000024555 s 0.00001583156001288444 s 1.55
hlo_ffi / IPartOpt / cpu / Forward 0.000023656 s 0.00001557716004754184 s 1.52
hlo_ffi / DefOpt / cpu / Forward 0.00002411 s 0.00001562955996632809 s 1.54
hlo_ffi / IDefOpt / cpu / Forward 0.000023745 s 0.000015224300077534282 s 1.56
hlo_ffi / JaXPipe / cpu / PreRev 0.00002368 s 0.00001584945996000897 s 1.49
hlo_ffi / JaXPipe / cpu / PostRev 0.000023558 s 0.00001478201989812078 s 1.59
hlo_ffi / JaXPipe / cpu / BothRev 0.000023434 s 0.000015412480024679098 s 1.52
hlo_ffi / Jax / cpu / BothRev 0.000023831000000000003 s 0.000016434639983344825 s 1.45
hlo_ffi / HLOOpt / cpu / PreRev 0.000023874 s 0.000015840260020922868 s 1.51
hlo_ffi / HLOOpt / cpu / PostRev 0.000024168 s 0.00001723093995678937 s 1.40
hlo_ffi / HLOOpt / cpu / BothRev 0.000024111 s 0.000015616500058968084 s 1.54
hlo_ffi / PartOpt / cpu / PreRev 0.000023619 s 0.000017233819999091793 s 1.37
hlo_ffi / PartOpt / cpu / PostRev 0.000023751 s 0.00001562794001074508 s 1.52
hlo_ffi / PartOpt / cpu / BothRev 0.000023997 s 0.000015728639846201985 s 1.53
hlo_ffi / IPartOpt / cpu / PreRev 0.000023674 s 0.000016466500019305384 s 1.44
hlo_ffi / IPartOpt / cpu / PostRev 0.000023993 s 0.000015306339992093854 s 1.57
hlo_ffi / IPartOpt / cpu / BothRev 0.000024231 s 0.000015431739957421086 s 1.57
hlo_ffi / DefOpt / cpu / PreRev 0.000023674 s 0.000016104480055219027 s 1.47
hlo_ffi / DefOpt / cpu / PostRev 0.000023736 s 0.0000156216798859532 s 1.52
hlo_ffi / DefOpt / cpu / BothRev 0.000023625 s 0.000015211660211207345 s 1.55
hlo_ffi / IDefOpt / cpu / PreRev 0.000024441 s 0.00001618880000023637 s 1.51
hlo_ffi / IDefOpt / cpu / PostRev 0.000023735 s 0.000015539559790340718 s 1.53
hlo_ffi / IDefOpt / cpu / BothRev 0.000024226 s 0.00001603133998287376 s 1.51
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 0.0008727956001166 s 0.0008776109996688 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 / Jax / cpu / Primal 0.0009022051999636 s 0.0008988255998701 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 / cpu / Primal 0.0009360922001178 s 0.0009498715997324 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 / cpu / Primal 0.0008805690000372 s 0.0008953810000093 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 / IPartOpt / cpu / Primal 0.0008812449999823 s 0.0009070822001376 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 / DefOpt / cpu / Primal 0.0009272497999518 s 0.0009808615999645 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 / IDefOpt / cpu / Primal 0.0009334880001006 s 0.000965224200263 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 / JaXPipe / cpu / Forward 0.0021853263999219 s 0.0021659400001226 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 / cpu / Forward 0.0022222202000193 s 0.0022857151994685 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 / HLOOpt / cpu / Forward 0.0022111282000878 s 0.002169238800343 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 / cpu / Forward 0.002196060999995 s 0.002201275599873 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 / cpu / Forward 0.0021652914000696 s 0.0022307005994662 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 / DefOpt / cpu / Forward 0.0022615739999309 s 0.0022027423998224 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 / cpu / Forward 0.0021877831999518 s 0.0021937772002274 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 / PreRev 0.0050008354000965 s 0.0054776116001448 s 0.91
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 0.004994894200081 s 0.0051365493993216 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 / JaXPipe / cpu / BothRev 0.0048707785999795 s 0.0044106954002927 s 1.10
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 0.0058284432000618 s 0.0037571628003206 s 1.55
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 0.0047370393999699 s 0.0032911693997448 s 1.44
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 0.0030746492000616 s 0.0034393395995721 s 0.89
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 0.0058646501999646 s 0.0051362879999942 s 1.14
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 0.0034704194000369 s 0.0041970395999669 s 0.83
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 0.0047814321999794 s 0.003264855600355 s 1.46
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 0.0048789258000397 s 0.0033160025999677 s 1.47
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 0.0032065564000731 s 0.0032031649996497 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 / cpu / PostRev 0.006050337599936 s 0.0036078554003324 s 1.68
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 0.0038809566000963 s 0.0031930009998177 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 / PreRev 0.0046246357999734 s 0.0035410561995377 s 1.31
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 0.0047534366000036 s 0.0030184160001226 s 1.57
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 0.0031858473999818 s 0.0034935592000692 s 0.91
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 0.004561921000004 s 0.0032312440001987 s 1.41
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 0.004580826000074 s 0.003169360999891 s 1.45
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 0.0047023729999637 s 0.0039165231999504 s 1.20
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 0.000299263 s 0.0002999339999999 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 / cuda / Primal 0.000299487 s 0.00030019 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 / cuda / Primal 0.000305535 s 0.000306686 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 / cuda / Primal 0.000299231 s 0.00029987 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 / cuda / Primal 0.000299071 s 0.000300703 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 / cuda / Primal 0.000306143 s 0.000308191 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 / cuda / Primal 0.000304896 s 0.000306526 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 / cuda / Forward 0.000583166 s 0.000583421 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 / cuda / Forward 0.000566142 s 0.000567996 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 / cuda / Forward 0.000582558 s 0.000583261 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 / cuda / Forward 0.0005831349999999 s 0.000584317 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 / cuda / Forward 0.0005832939999999 s 0.000583612 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 / cuda / Forward 0.000583807 s 0.000582877 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 / cuda / Forward 0.0005828459999999 s 0.000583357 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 / cuda / PreRev 0.0010578529999999 s 0.001059611 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 / cuda / PostRev 0.001012989 s 0.001014075 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 / cuda / BothRev 0.001052318 s 0.001054907 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 / cuda / BothRev 0.001006877 s 0.0010080579999999 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 / cuda / PreRev 0.001036317 s 0.0010398019999999 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 / cuda / PostRev 0.001062973 s 0.001065435 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 / cuda / BothRev 0.001037758 s 0.001042203 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 / cuda / PreRev 0.001052029 s 0.0010528269999999 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 / cuda / PostRev 0.001000253 s 0.001002811 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 / cuda / BothRev 0.001052093 s 0.001053563 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 / cuda / PreRev 0.001053725 s 0.001056218 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 / cuda / PostRev 0.0010005729999999 s 0.001001754 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 / cuda / BothRev 0.001054012 s 0.001055162 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 / cuda / PreRev 0.001051549 s 0.001056987 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 / cuda / PostRev 0.000986109 s 0.000991611 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 / cuda / BothRev 0.001052861 s 0.0010569219999999 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 / cuda / PreRev 0.001051997 s 0.0010560259999999 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 / cuda / PostRev 0.001053789 s 0.00105913 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 / cuda / BothRev 0.001054078 s 0.001058234 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 / Primal 0.000124571 s 0.000131029 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 / Jax / tpu / Primal 0.0001269539999999 s 0.00012435075 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 / tpu / Primal 0.000152598 s 0.000160308 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 / PartOpt / tpu / Primal 0.00013434825 s 0.0001308517499999 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 / tpu / Primal 0.00013122275 s 0.000138425 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 / DefOpt / tpu / Primal 0.00014772825 s 0.000145365 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 / tpu / Primal 0.00015114425 s 0.000158025 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 / JaXPipe / tpu / Forward 0.0002117357499999 s 0.00021332725 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 / Jax / tpu / Forward 0.00026111375 s 0.000262598 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 / Forward 0.00021204625 s 0.00022060475 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 / PartOpt / tpu / Forward 0.00021834825 s 0.00021477225 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 / tpu / Forward 0.0002117345 s 0.000215888 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 / DefOpt / tpu / Forward 0.00021829925 s 0.0002181132499999 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 0.0002119749999999 s 0.0002159965 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 / tpu / PreRev 0.00035460125 s 0.0003561094999999 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 0.00025655575 s 0.0002560205 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 / BothRev 0.000354654 s 0.00035627425 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 0.00025699525 s 0.00025682575 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 / PreRev 0.00035466125 s 0.000355867 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 0.00029118825 s 0.00029123525 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 / BothRev 0.0003546645 s 0.00035601675 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 0.000355458 s 0.0003562984999999 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 / PostRev 0.0002713919999999 s 0.0002720207499999 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 0.00035549025 s 0.0003566655 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 / PreRev 0.0003546425 s 0.000355987 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 / PostRev 0.00027209375 s 0.00027238875 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 / BothRev 0.00035473575 s 0.000356185 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 / PreRev 0.00035783125 s 0.0003581459999999 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 / PostRev 0.00028334675 s 0.00028436275 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 0.000357901 s 0.00035823625 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 / PreRev 0.0003567242499999 s 0.00035828 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 0.00030093925 s 0.00030107525 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 0.00035688425 s 0.00035868325 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 / cpu / Primal 0.002113162 s 0.0008776109996688 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 / Jax / cpu / Primal 0.00183245 s 0.0008988255998701 s 2.04
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / HLOOpt / cpu / Primal 0.002195878 s 0.0009498715997324 s 2.31
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 0.002086306 s 0.0008953810000093 s 2.33
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 0.0017782009999999 s 0.0009070822001376 s 1.96
llama_dim_288_hidden_dim_768_n_layers_6_n_heads_6_n_kv_heads_6_vocab_size_32000_seq_len_256 / DefOpt / cpu / Primal 0.001677778 s 0.0009808615999645 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 / IDefOpt / cpu / Primal 0.001681002 s 0.000965224200263 s 1.74
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 0.005081218 s 0.0021659400001226 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 / Jax / cpu / Forward 0.005390594 s 0.0022857151994685 s 2.36
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 0.004539823 s 0.002169238800343 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 / PartOpt / cpu / Forward 0.004875755 s 0.002201275599873 s 2.21
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 0.005059647 s 0.0022307005994662 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 / DefOpt / cpu / Forward 0.005465504 s 0.0022027423998224 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 / IDefOpt / cpu / Forward 0.005162198 s 0.0021937772002274 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 / PreRev 0.008817735 s 0.0054776116001448 s 1.61
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 0.008867063 s 0.0051365493993216 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 / JaXPipe / cpu / BothRev 0.009667969 s 0.0044106954002927 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 / Jax / cpu / BothRev 0.008053285 s 0.0037571628003206 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 / HLOOpt / cpu / PreRev 0.009588276 s 0.0032911693997448 s 2.91
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 0.007182244 s 0.0034393395995721 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 / HLOOpt / cpu / BothRev 0.008501049 s 0.0051362879999942 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 / PartOpt / cpu / PreRev 0.007810628 s 0.0041970395999669 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 / PartOpt / cpu / PostRev 0.008251922 s 0.003264855600355 s 2.53
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 0.00788562 s 0.0033160025999677 s 2.38
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 0.007469742 s 0.0032031649996497 s 2.33
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 0.0084722399999999 s 0.0036078554003324 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 / IPartOpt / cpu / BothRev 0.007604206 s 0.0031930009998177 s 2.38
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 0.009127958 s 0.0035410561995377 s 2.58
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 0.006877182 s 0.0030184160001226 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 / DefOpt / cpu / BothRev 0.0093124749999999 s 0.0034935592000692 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 / IDefOpt / cpu / PreRev 0.008663146 s 0.0032312440001987 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 / IDefOpt / cpu / PostRev 0.0082441329999999 s 0.003169360999891 s 2.60
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 0.007943999 s 0.0039165231999504 s 2.03
scatter_sum / JaXPipe / cpu / Primal 0.00000752614005250507 s 0.000008171760018740315 s 0.92
scatter_sum / Jax / cpu / Primal 0.0000074308400235167935 s 0.000007970799961185549 s 0.93
scatter_sum / HLOOpt / cpu / Primal 0.000007615140011694166 s 0.000007748800053377636 s 0.98
scatter_sum / PartOpt / cpu / Primal 0.000007615700023961836 s 0.000007296480107470415 s 1.04
scatter_sum / IPartOpt / cpu / Primal 0.0000075831200501852435 s 0.000007890339948062319 s 0.96
scatter_sum / DefOpt / cpu / Primal 0.000007356060013989917 s 0.000007875180017435923 s 0.93
scatter_sum / IDefOpt / cpu / Primal 0.000007581100026072818 s 0.00000782291990617523 s 0.97
scatter_sum / JaXPipe / cpu / Forward 0.00001171202001387428 s 0.00001187668000056874 s 0.99
scatter_sum / Jax / cpu / Forward 0.000011680520001391417 s 0.000012111020150769037 s 0.96
scatter_sum / HLOOpt / cpu / Forward 0.000012281220042495989 s 0.000012101219908799976 s 1.01
scatter_sum / PartOpt / cpu / Forward 0.00001184676000775653 s 0.000012010039936285466 s 0.99
scatter_sum / IPartOpt / cpu / Forward 0.00001199494002321444 s 0.00001217569995787926 s 0.99
scatter_sum / DefOpt / cpu / Forward 0.000011832940008389417 s 0.000011737019958673043 s 1.01
scatter_sum / IDefOpt / cpu / Forward 0.000011979159990005427 s 0.000011661860007734503 s 1.03
scatter_sum / JaXPipe / cpu / PreRev 0.000011260920009590336 s 0.000011499279898998794 s 0.98
scatter_sum / JaXPipe / cpu / PostRev 0.000011303319997750804 s 0.00001180788007332012 s 0.96
scatter_sum / JaXPipe / cpu / BothRev 0.00001190994002172374 s 0.000012415540113579482 s 0.96
scatter_sum / Jax / cpu / BothRev 0.000011323220014674007 s 0.000011430880113039164 s 0.99
scatter_sum / HLOOpt / cpu / PreRev 0.000011752840036933775 s 0.000012566820078063755 s 0.94
scatter_sum / HLOOpt / cpu / PostRev 0.000013000700009797583 s 0.00001453337990824366 s 0.89
scatter_sum / HLOOpt / cpu / BothRev 0.000011535159983395716 s 0.000012527280123322273 s 0.92
scatter_sum / PartOpt / cpu / PreRev 0.000011647460014501122 s 0.000011573640003916807 s 1.01
scatter_sum / PartOpt / cpu / PostRev 0.000011373460065442488 s 0.000012084200061508456 s 0.94
scatter_sum / PartOpt / cpu / BothRev 0.000012076640023224171 s 0.000012801519987988286 s 0.94
scatter_sum / IPartOpt / cpu / PreRev 0.00001152347999777703 s 0.000011545180132088718 s 1.00
scatter_sum / IPartOpt / cpu / PostRev 0.000011456920019554672 s 0.000012068579999322537 s 0.95
scatter_sum / IPartOpt / cpu / BothRev 0.000011455280018708436 s 0.000012279979855520651 s 0.93
scatter_sum / DefOpt / cpu / PreRev 0.000011356320001141283 s 0.000011816779915534424 s 0.96
scatter_sum / DefOpt / cpu / PostRev 0.000011563680009203382 s 0.000012119859893573449 s 0.95
scatter_sum / DefOpt / cpu / BothRev 0.000011308919956718455 s 0.00001199918013298884 s 0.94
scatter_sum / IDefOpt / cpu / PreRev 0.00001153621997218579 s 0.000011675239911710375 s 0.99
scatter_sum / IDefOpt / cpu / PostRev 0.000011533860006238684 s 0.00001227513999765506 s 0.94
scatter_sum / IDefOpt / cpu / BothRev 0.000011358179999660934 s 0.000012509239968494513 s 0.91
scatter_sum / JaXPipe / cuda / Primal 0.000010847 s 0.000010784 s 1.01
scatter_sum / Jax / cuda / Primal 0.000011936 s 0.000010528 s 1.13
scatter_sum / HLOOpt / cuda / Primal 0.00001248 s 0.000010912 s 1.14
scatter_sum / PartOpt / cuda / Primal 0.000010752 s 0.000010528 s 1.02
scatter_sum / IPartOpt / cuda / Primal 0.000010785 s 0.000010976 s 0.98
scatter_sum / DefOpt / cuda / Primal 0.000010911 s 0.000012 s 0.91
scatter_sum / IDefOpt / cuda / Primal 0.000010976 s 0.000011008 s 1.00
scatter_sum / JaXPipe / cuda / Forward 0.000018016 s 0.000018176 s 0.99
scatter_sum / Jax / cuda / Forward 0.000017664 s 0.000017824 s 0.99
scatter_sum / HLOOpt / cuda / Forward 0.000018016 s 0.00001744 s 1.03
scatter_sum / PartOpt / cuda / Forward 0.000017760000000000003 s 0.000019712 s 0.90
scatter_sum / IPartOpt / cuda / Forward 0.000017023 s 0.000017823 s 0.96
scatter_sum / DefOpt / cuda / Forward 0.000018112 s 0.000017472 s 1.04
scatter_sum / IDefOpt / cuda / Forward 0.000017760000000000003 s 0.000019808 s 0.90
scatter_sum / JaXPipe / cuda / PreRev 0.000017536 s 0.000017375999999999998 s 1.01
scatter_sum / JaXPipe / cuda / PostRev 0.0000176 s 0.000017406999999999998 s 1.01
scatter_sum / JaXPipe / cuda / BothRev 0.00001728 s 0.000017664 s 0.98
scatter_sum / Jax / cuda / BothRev 0.000017664 s 0.000017472 s 1.01
scatter_sum / HLOOpt / cuda / PreRev 0.000017664 s 0.000017567 s 1.01
scatter_sum / HLOOpt / cuda / PostRev 0.000017408 s 0.000017408 s 1
scatter_sum / HLOOpt / cuda / BothRev 0.000016896000000000002 s 0.000017375999999999998 s 0.97
scatter_sum / PartOpt / cuda / PreRev 0.000018112 s 0.000017793 s 1.02
scatter_sum / PartOpt / cuda / PostRev 0.00001712 s 0.000017184 s 1.00
scatter_sum / PartOpt / cuda / BothRev 0.000017760000000000003 s 0.000017568000000000002 s 1.01
scatter_sum / IPartOpt / cuda / PreRev 0.000018208 s 0.000017952 s 1.01
scatter_sum / IPartOpt / cuda / PostRev 0.000017824 s 0.000016352 s 1.09
scatter_sum / IPartOpt / cuda / BothRev 0.000017568000000000002 s 0.000018048 s 0.97
scatter_sum / DefOpt / cuda / PreRev 0.00001808 s 0.000017409000000000002 s 1.04
scatter_sum / DefOpt / cuda / PostRev 0.000017536 s 0.000017632 s 0.99
scatter_sum / DefOpt / cuda / BothRev 0.000017663 s 0.000017345 s 1.02
scatter_sum / IDefOpt / cuda / PreRev 0.0000176 s 0.000017919999999999998 s 0.98
scatter_sum / IDefOpt / cuda / PostRev 0.000017696 s 0.000016864 s 1.05
scatter_sum / IDefOpt / cuda / BothRev 0.000017344 s 0.000017471 s 0.99
scatter_sum / JaXPipe / tpu / Primal 0.000001350275 s 0.000001344225 s 1.00
scatter_sum / Jax / tpu / Primal 0.000001342825 s 0.000001358175 s 0.99
scatter_sum / HLOOpt / tpu / Primal 0.00000135045 s 0.000001344275 s 1.00
scatter_sum / PartOpt / tpu / Primal 0.000001343625 s 0.0000013579 s 0.99
scatter_sum / IPartOpt / tpu / Primal 0.000001350275 s 0.0000013447 s 1.00
scatter_sum / DefOpt / tpu / Primal 0.000001343575 s 0.000001357375 s 0.99
scatter_sum / IDefOpt / tpu / Primal 0.00000135085 s 0.000001344475 s 1.00
scatter_sum / JaXPipe / tpu / Forward 0.000002691225 s 0.0000027374500000000004 s 0.98
scatter_sum / Jax / tpu / Forward 0.000002728 s 0.000002752225 s 0.99
scatter_sum / HLOOpt / tpu / Forward 0.0000026902 s 0.000002741175 s 0.98
scatter_sum / PartOpt / tpu / Forward 0.000002693975 s 0.000002711925 s 0.99
scatter_sum / IPartOpt / tpu / Forward 0.00000268195 s 0.000002735925 s 0.98
scatter_sum / DefOpt / tpu / Forward 0.00000269585 s 0.00000271585 s 0.99
scatter_sum / IDefOpt / tpu / Forward 0.000002681825 s 0.00000273715 s 0.98
scatter_sum / JaXPipe / tpu / PreRev 0.000002692575 s 0.00000271315 s 0.99
scatter_sum / JaXPipe / tpu / PostRev 0.000002687325 s 0.00000273235 s 0.98
scatter_sum / JaXPipe / tpu / BothRev 0.000002710525 s 0.0000027247 s 0.99
scatter_sum / Jax / tpu / BothRev 0.000002743 s 0.000002794075 s 0.98
scatter_sum / HLOOpt / tpu / PreRev 0.00000270295 s 0.0000027256750000000004 s 0.99
scatter_sum / HLOOpt / tpu / PostRev 0.0000027362999999999995 s 0.0000028006 s 0.98
scatter_sum / HLOOpt / tpu / BothRev 0.0000027107750000000003 s 0.00000272225 s 1.00
scatter_sum / PartOpt / tpu / PreRev 0.00000274275 s 0.000002789825 s 0.98
scatter_sum / PartOpt / tpu / PostRev 0.0000027067000000000003 s 0.0000027312 s 0.99
scatter_sum / PartOpt / tpu / BothRev 0.0000027333750000000003 s 0.000002793825 s 0.98
scatter_sum / IPartOpt / tpu / PreRev 0.000002703975 s 0.00000272725 s 0.99
scatter_sum / IPartOpt / tpu / PostRev 0.000002735 s 0.00000279215 s 0.98
scatter_sum / IPartOpt / tpu / BothRev 0.000002713425 s 0.000002726525 s 1.00
scatter_sum / DefOpt / tpu / PreRev 0.0000027333750000000003 s 0.0000027877 s 0.98
scatter_sum / DefOpt / tpu / PostRev 0.000002706975 s 0.000002721 s 0.99
scatter_sum / DefOpt / tpu / BothRev 0.0000027349000000000003 s 0.00000279375 s 0.98
scatter_sum / IDefOpt / tpu / PreRev 0.00000270485 s 0.000002733225 s 0.99
scatter_sum / IDefOpt / tpu / PostRev 0.000002736425 s 0.0000027860750000000003 s 0.98
scatter_sum / IDefOpt / tpu / BothRev 0.0000027026 s 0.0000027353250000000004 s 0.99
scatter_sum / JaXPipe / cpu / Primal 0.000015504 s 0.000008171760018740315 s 1.90
scatter_sum / Jax / cpu / Primal 0.000025121 s 0.000007970799961185549 s 3.15
scatter_sum / HLOOpt / cpu / Primal 0.00001567 s 0.000007748800053377636 s 2.02
scatter_sum / PartOpt / cpu / Primal 0.000015734000000000002 s 0.000007296480107470415 s 2.16
scatter_sum / IPartOpt / cpu / Primal 0.000015835 s 0.000007890339948062319 s 2.01
scatter_sum / DefOpt / cpu / Primal 0.000015913 s 0.000007875180017435923 s 2.02
scatter_sum / IDefOpt / cpu / Primal 0.00001551 s 0.00000782291990617523 s 1.98
scatter_sum / JaXPipe / cpu / Forward 0.000023416 s 0.00001187668000056874 s 1.97
scatter_sum / Jax / cpu / Forward 0.000023243 s 0.000012111020150769037 s 1.92
scatter_sum / HLOOpt / cpu / Forward 0.000022345 s 0.000012101219908799976 s 1.85
scatter_sum / PartOpt / cpu / Forward 0.000023351 s 0.000012010039936285466 s 1.94
scatter_sum / IPartOpt / cpu / Forward 0.00002389 s 0.00001217569995787926 s 1.96
scatter_sum / DefOpt / cpu / Forward 0.000024197 s 0.000011737019958673043 s 2.06
scatter_sum / IDefOpt / cpu / Forward 0.000022677 s 0.000011661860007734503 s 1.94
scatter_sum / JaXPipe / cpu / PreRev 0.000023244 s 0.000011499279898998794 s 2.02
scatter_sum / JaXPipe / cpu / PostRev 0.000022621 s 0.00001180788007332012 s 1.92
scatter_sum / JaXPipe / cpu / BothRev 0.000023003 s 0.000012415540113579482 s 1.85
scatter_sum / Jax / cpu / BothRev 0.000022681 s 0.000011430880113039164 s 1.98
scatter_sum / HLOOpt / cpu / PreRev 0.000022888 s 0.000012566820078063755 s 1.82
scatter_sum / HLOOpt / cpu / PostRev 0.000022797 s 0.00001453337990824366 s 1.57
scatter_sum / HLOOpt / cpu / BothRev 0.000022891 s 0.000012527280123322273 s 1.83
scatter_sum / PartOpt / cpu / PreRev 0.000022542 s 0.000011573640003916807 s 1.95
scatter_sum / PartOpt / cpu / PostRev 0.000022521 s 0.000012084200061508456 s 1.86
scatter_sum / PartOpt / cpu / BothRev 0.00002303 s 0.000012801519987988286 s 1.80
scatter_sum / IPartOpt / cpu / PreRev 0.000023163 s 0.000011545180132088718 s 2.01
scatter_sum / IPartOpt / cpu / PostRev 0.000023029 s 0.000012068579999322537 s 1.91
scatter_sum / IPartOpt / cpu / BothRev 0.000022954 s 0.000012279979855520651 s 1.87
scatter_sum / DefOpt / cpu / PreRev 0.00002246 s 0.000011816779915534424 s 1.90
scatter_sum / DefOpt / cpu / PostRev 0.000022644 s 0.000012119859893573449 s 1.87
scatter_sum / DefOpt / cpu / BothRev 0.000022933 s 0.00001199918013298884 s 1.91
scatter_sum / IDefOpt / cpu / PreRev 0.000023248 s 0.000011675239911710375 s 1.99
scatter_sum / IDefOpt / cpu / PostRev 0.000022711 s 0.00001227513999765506 s 1.85
scatter_sum / IDefOpt / cpu / BothRev 0.000023079 s 0.000012509239968494513 s 1.84
slicing / JaXPipe / cpu / Primal 0.000006327620030788239 s 0.0000061953599652042615 s 1.02
slicing / Jax / cpu / Primal 0.000006110100021032849 s 0.000006929080154804979 s 0.88
slicing / HLOOpt / cpu / Primal 0.00000673963999361149 s 0.0000068508998811012135 s 0.98
slicing / PartOpt / cpu / Primal 0.000006227499998203712 s 0.000006121540027379524 s 1.02
slicing / IPartOpt / cpu / Primal 0.000006236400022316957 s 0.000006725820057909005 s 0.93
slicing / DefOpt / cpu / Primal 0.00000610848000178521 s 0.000007069700077408925 s 0.86
slicing / IDefOpt / cpu / Primal 0.000006536679984492367 s 0.000006266299933486152 s 1.04
slicing / JaXPipe / cpu / Forward 0.000009629480000512558 s 0.0000099524600955192 s 0.97
slicing / Jax / cpu / Forward 0.00000934474000132468 s 0.000009137279994320124 s 1.02
slicing / HLOOpt / cpu / Forward 0.000010158279983443207 s 0.000009985880024032668 s 1.02
slicing / PartOpt / cpu / Forward 0.000008899780004867352 s 0.000009016019794216845 s 0.99
slicing / IPartOpt / cpu / Forward 0.000010137659965039349 s 0.000009696300039649942 s 1.05
slicing / DefOpt / cpu / Forward 0.000009306539977842476 s 0.0000097391199960839 s 0.96
slicing / IDefOpt / cpu / Forward 0.00000947069996072969 s 0.000009139280082308688 s 1.04
slicing / JaXPipe / cpu / PreRev 0.0000100505200043699 s 0.000010525399957259652 s 0.95
slicing / JaXPipe / cpu / PostRev 0.00001006576003419468 s 0.00000999433996184962 s 1.01
slicing / JaXPipe / cpu / BothRev 0.000010081459977300256 s 0.000010574719926808027 s 0.95
slicing / Jax / cpu / BothRev 0.000010107100042660022 s 0.000010336499981349334 s 0.98
slicing / HLOOpt / cpu / PreRev 0.000010265639966746676 s 0.000010184479833696967 s 1.01
slicing / HLOOpt / cpu / PostRev 0.000011905739975190952 s 0.000012298240035306662 s 0.97
slicing / HLOOpt / cpu / BothRev 0.00001016011997307942 s 0.00001021929976559477 s 0.99
slicing / PartOpt / cpu / PreRev 0.00000985252000646142 s 0.000010392360090918371 s 0.95
slicing / PartOpt / cpu / PostRev 0.000010124879991053605 s 0.000010358800100220831 s 0.98
slicing / PartOpt / cpu / BothRev 0.000009897120044115582 s 0.000010531820043979678 s 0.94
slicing / IPartOpt / cpu / PreRev 0.000009764179985722876 s 0.000009982420051528606 s 0.98
slicing / IPartOpt / cpu / PostRev 0.000009950640014722012 s 0.000010332759920856917 s 0.96
slicing / IPartOpt / cpu / BothRev 0.000009960500019587924 s 0.000010125199878530111 s 0.98
slicing / DefOpt / cpu / PreRev 0.00001003487997877528 s 0.00001015016001474578 s 0.99
slicing / DefOpt / cpu / PostRev 0.00000988970001344569 s 0.000010507639999559614 s 0.94
slicing / DefOpt / cpu / BothRev 0.000010041919977084034 s 0.000009575999865774065 s 1.05
slicing / IDefOpt / cpu / PreRev 0.000009768779982550768 s 0.000010578259971225634 s 0.92
slicing / IDefOpt / cpu / PostRev 0.00000974606001364009 s 0.000009717359898786524 s 1.00
slicing / IDefOpt / cpu / BothRev 0.000010449660012454842 s 0.000010080320062115788 s 1.04
slicing / JaXPipe / cuda / Primal 0.000002272 s 0.000002304 s 0.99
slicing / Jax / cuda / Primal 0.000002272 s 0.000002303 s 0.99
slicing / HLOOpt / cuda / Primal 0.000002272 s 0.000002303 s 0.99
slicing / PartOpt / cuda / Primal 0.000002271 s 0.000002303 s 0.99
slicing / IPartOpt / cuda / Primal 0.000002271 s 0.000002303 s 0.99
slicing / DefOpt / cuda / Primal 0.000002272 s 0.000002303 s 0.99
slicing / IDefOpt / cuda / Primal 0.000002272 s 0.000002303 s 0.99
slicing / JaXPipe / cuda / Forward 0.000011008 s 0.000010528 s 1.05
slicing / Jax / cuda / Forward 0.000012256 s 0.000010656 s 1.15
slicing / HLOOpt / cuda / Forward 0.000010816 s 0.000011616 s 0.93
slicing / PartOpt / cuda / Forward 0.000010784 s 0.000010464 s 1.03
slicing / IPartOpt / cuda / Forward 0.000010624 s 0.000011455999999999998 s 0.93
slicing / DefOpt / cuda / Forward 0.000010432 s 0.0000104 s 1.00
slicing / IDefOpt / cuda / Forward 0.000010848 s 0.000010656 s 1.02
slicing / JaXPipe / cuda / PreRev 0.000010688 s 0.00001024 s 1.04
slicing / JaXPipe / cuda / PostRev 0.000011776 s 0.000010688 s 1.10
slicing / JaXPipe / cuda / BothRev 0.000011648 s 0.000010944 s 1.06
slicing / Jax / cuda / BothRev 0.000010816 s 0.000010304 s 1.05
slicing / HLOOpt / cuda / PreRev 0.000010944 s 0.000010879 s 1.01
slicing / HLOOpt / cuda / PostRev 0.000010303 s 0.000010624 s 0.97
slicing / HLOOpt / cuda / BothRev 0.00001024 s 0.000010208 s 1.00
slicing / PartOpt / cuda / PreRev 0.000010976 s 0.0000104 s 1.06
slicing / PartOpt / cuda / PostRev 0.000010751 s 0.0000104 s 1.03
slicing / PartOpt / cuda / BothRev 0.00001072 s 0.000010752 s 1.00
slicing / IPartOpt / cuda / PreRev 0.000010336 s 0.00001088 s 0.95
slicing / IPartOpt / cuda / PostRev 0.000010847 s 0.000010208 s 1.06
slicing / IPartOpt / cuda / BothRev 0.000010528 s 0.000010496 s 1.00
slicing / DefOpt / cuda / PreRev 0.000011071 s 0.000010369 s 1.07
slicing / DefOpt / cuda / PostRev 0.000010496 s 0.000010848 s 0.97
slicing / DefOpt / cuda / BothRev 0.000010368 s 0.000010272 s 1.01
slicing / IDefOpt / cuda / PreRev 0.000010687 s 0.00001056 s 1.01
slicing / IDefOpt / cuda / PostRev 0.0000104 s 0.000010688 s 0.97
slicing / IDefOpt / cuda / BothRev 0.000010496 s 0.000010816 s 0.97
slicing / JaXPipe / tpu / Primal 9.6305e-7 s 9.56525e-7 s 1.01
slicing / Jax / tpu / Primal 9.635249999999998e-7 s 9.6695e-7 s 1.00
slicing / HLOOpt / tpu / Primal 9.674e-7 s 9.62375e-7 s 1.01
slicing / PartOpt / tpu / Primal 9.66325e-7 s 9.6405e-7 s 1.00
slicing / IPartOpt / tpu / Primal 9.67375e-7 s 9.49075e-7 s 1.02
slicing / DefOpt / tpu / Primal 9.61175e-7 s 9.61675e-7 s 1.00
slicing / IDefOpt / tpu / Primal 9.649e-7 s 9.4925e-7 s 1.02
slicing / JaXPipe / tpu / Forward 0.0000014130000000000005 s 0.000001397875 s 1.01
slicing / Jax / tpu / Forward 0.000001422775 s 0.0000014018 s 1.01
slicing / HLOOpt / tpu / Forward 0.000001524225 s 0.000001504925 s 1.01
slicing / PartOpt / tpu / Forward 0.0000014355499999999995 s 0.0000014259250000000002 s 1.01
slicing / IPartOpt / tpu / Forward 0.0000015263 s 0.000001502825 s 1.02
slicing / DefOpt / tpu / Forward 0.0000014351249999999997 s 0.000001418525 s 1.01
slicing / IDefOpt / tpu / Forward 0.0000015184 s 0.000001504 s 1.01
slicing / JaXPipe / tpu / PreRev 0.00000233075 s 0.000002326175 s 1.00
slicing / JaXPipe / tpu / PostRev 0.0000024985 s 0.000002511375 s 0.99
slicing / JaXPipe / tpu / BothRev 0.000002351175 s 0.000002344625 s 1.00
slicing / Jax / tpu / BothRev 0.0000025301 s 0.000002525925 s 1.00
slicing / HLOOpt / tpu / PreRev 0.000002356325 s 0.00000234655 s 1.00
slicing / HLOOpt / tpu / PostRev 0.000002519675 s 0.0000025209 s 1.00
slicing / HLOOpt / tpu / BothRev 0.000002353675 s 0.000002349525 s 1.00
slicing / PartOpt / tpu / PreRev 0.00000252335 s 0.000002525675 s 1.00
slicing / PartOpt / tpu / PostRev 0.0000023431 s 0.0000023427 s 1.00
slicing / PartOpt / tpu / BothRev 0.0000025238500000000004 s 0.0000025337 s 1.00
slicing / IPartOpt / tpu / PreRev 0.0000023512 s 0.00000235335 s 1.00
slicing / IPartOpt / tpu / PostRev 0.00000252095 s 0.000002518825 s 1.00
slicing / IPartOpt / tpu / BothRev 0.0000023483 s 0.000002355 s 1.00
slicing / DefOpt / tpu / PreRev 0.00000252625 s 0.0000025295750000000003 s 1.00
slicing / DefOpt / tpu / PostRev 0.00000234735 s 0.00000234935 s 1.00
slicing / DefOpt / tpu / BothRev 0.0000025253250000000003 s 0.0000025281 s 1.00
slicing / IDefOpt / tpu / PreRev 0.000002352 s 0.000002341675 s 1.00
slicing / IDefOpt / tpu / PostRev 0.000002530325 s 0.00000253535 s 1.00
slicing / IDefOpt / tpu / BothRev 0.0000023613750000000003 s 0.00000234925 s 1.01
slicing / JaXPipe / cpu / Primal 0.00001286 s 0.0000061953599652042615 s 2.08
slicing / Jax / cpu / Primal 0.000012651 s 0.000006929080154804979 s 1.83
slicing / HLOOpt / cpu / Primal 0.000012756 s 0.0000068508998811012135 s 1.86
slicing / PartOpt / cpu / Primal 0.000012528 s 0.000006121540027379524 s 2.05
slicing / IPartOpt / cpu / Primal 0.000012618 s 0.000006725820057909005 s 1.88
slicing / DefOpt / cpu / Primal 0.000012664 s 0.000007069700077408925 s 1.79
slicing / IDefOpt / cpu / Primal 0.000012729 s 0.000006266299933486152 s 2.03
slicing / JaXPipe / cpu / Forward 0.000017395999999999997 s 0.0000099524600955192 s 1.75
slicing / Jax / cpu / Forward 0.000016831 s 0.000009137279994320124 s 1.84
slicing / HLOOpt / cpu / Forward 0.000017214 s 0.000009985880024032668 s 1.72
slicing / PartOpt / cpu / Forward 0.000016893 s 0.000009016019794216845 s 1.87
slicing / IPartOpt / cpu / Forward 0.000017074999999999998 s 0.000009696300039649942 s 1.76
slicing / DefOpt / cpu / Forward 0.000016945 s 0.0000097391199960839 s 1.74
slicing / IDefOpt / cpu / Forward 0.000017001 s 0.000009139280082308688 s 1.86
slicing / JaXPipe / cpu / PreRev 0.000017879 s 0.000010525399957259652 s 1.70
slicing / JaXPipe / cpu / PostRev 0.000017512 s 0.00000999433996184962 s 1.75
slicing / JaXPipe / cpu / BothRev 0.000017447 s 0.000010574719926808027 s 1.65
slicing / Jax / cpu / BothRev 0.000017292 s 0.000010336499981349334 s 1.67
slicing / HLOOpt / cpu / PreRev 0.000017497 s 0.000010184479833696967 s 1.72
slicing / HLOOpt / cpu / PostRev 0.000017397 s 0.000012298240035306662 s 1.41
slicing / HLOOpt / cpu / BothRev 0.00001749 s 0.00001021929976559477 s 1.71
slicing / PartOpt / cpu / PreRev 0.000017364000000000002 s 0.000010392360090918371 s 1.67
slicing / PartOpt / cpu / PostRev 0.00001749 s 0.000010358800100220831 s 1.69
slicing / PartOpt / cpu / BothRev 0.000017513 s 0.000010531820043979678 s 1.66
slicing / IPartOpt / cpu / PreRev 0.000017696 s 0.000009982420051528606 s 1.77
slicing / IPartOpt / cpu / PostRev 0.000017459 s 0.000010332759920856917 s 1.69
slicing / IPartOpt / cpu / BothRev 0.000017384 s 0.000010125199878530111 s 1.72
slicing / DefOpt / cpu / PreRev 0.000017818 s 0.00001015016001474578 s 1.76
slicing / DefOpt / cpu / PostRev 0.000017578 s 0.000010507639999559614 s 1.67
slicing / DefOpt / cpu / BothRev 0.000017409000000000002 s 0.000009575999865774065 s 1.82
slicing / IDefOpt / cpu / PreRev 0.000017547 s 0.000010578259971225634 s 1.66
slicing / IDefOpt / cpu / PostRev 0.000017590000000000003 s 0.000009717359898786524 s 1.81
slicing / IDefOpt / cpu / BothRev 0.000017479 s 0.000010080320062115788 s 1.73
sum / JaXPipe / cpu / Primal 0.000007554379990324378 s 0.000007981139933690429 s 0.95
sum / Jax / cpu / Primal 0.000007960660022945376 s 0.000007507980044465512 s 1.06
sum / HLOOpt / cpu / Primal 0.000007627059949300019 s 0.000008308500073326285 s 0.92
sum / PartOpt / cpu / Primal 0.000007592980009576422 s 0.000007578700096928514 s 1.00
sum / IPartOpt / cpu / Primal 0.000008399579983233706 s 0.00000764742013416253 s 1.10
sum / DefOpt / cpu / Primal 0.000007867059975978918 s 0.000007487280381610617 s 1.05
sum / IDefOpt / cpu / Primal 0.000007898500034571044 s 0.000007432820129906759 s 1.06
sum / JaXPipe / cpu / Forward 0.000011029500001313863 s 0.000011446299795352389 s 0.96
sum / Jax / cpu / Forward 0.000011038420034310547 s 0.000011345720013196116 s 0.97
sum / HLOOpt / cpu / Forward 0.000011531520012795228 s 0.000011818420098279605 s 0.98
sum / PartOpt / cpu / Forward 0.000011068879985032254 s 0.000010789639927679672 s 1.03
sum / IPartOpt / cpu / Forward 0.00001204667993079056 s 0.000011666479986160994 s 1.03
sum / DefOpt / cpu / Forward 0.000011732200018741425 s 0.00001131830027588876 s 1.04
sum / IDefOpt / cpu / Forward 0.000011114459948657896 s 0.000011210779812245164 s 0.99
sum / JaXPipe / cpu / PreRev 0.000010913880032603627 s 0.000011084600300819147 s 0.98
sum / JaXPipe / cpu / PostRev 0.000011153299992656684 s 0.000010880019872274717 s 1.03
sum / JaXPipe / cpu / BothRev 0.00001089274001060403 s 0.000011212040044483727 s 0.97
sum / Jax / cpu / BothRev 0.00001085040001271409 s 0.000010437400087539572 s 1.04
sum / HLOOpt / cpu / PreRev 0.000011326420017212513 s 0.000011505000002216549 s 0.98
sum / HLOOpt / cpu / PostRev 0.000012447360013538855 s 0.000012359059837763198 s 1.01
sum / HLOOpt / cpu / BothRev 0.000010493820009287446 s 0.000010557379791862331 s 0.99
sum / PartOpt / cpu / PreRev 0.000010942959997919388 s 0.000010827380028786138 s 1.01
sum / PartOpt / cpu / PostRev 0.000011010499993062694 s 0.000011040499921364243 s 1.00
sum / PartOpt / cpu / BothRev 0.000011881060045197955 s 0.000011394000066502486 s 1.04
sum / IPartOpt / cpu / PreRev 0.00001094047999686154 s 0.000010900399975071197 s 1.00
sum / IPartOpt / cpu / PostRev 0.000010565060010776506 s 0.000011152820188726765 s 0.95
sum / IPartOpt / cpu / BothRev 0.000010662920058166492 s 0.000010597239997878205 s 1.01
sum / DefOpt / cpu / PreRev 0.000010670539977581938 s 0.00001111023982957704 s 0.96
sum / DefOpt / cpu / PostRev 0.000010902719986916054 s 0.000010991519957315177 s 0.99
sum / DefOpt / cpu / BothRev 0.000010851880006157444 s 0.00001119037999160355 s 0.97
sum / IDefOpt / cpu / PreRev 0.000010496120039533708 s 0.000010858259993256067 s 0.97
sum / IDefOpt / cpu / PostRev 0.000010656240028765751 s 0.000010737400079960937 s 0.99
sum / IDefOpt / cpu / BothRev 0.000011423879996073084 s 0.000010744440078269693 s 1.06
sum / JaXPipe / cuda / Primal 0.000002463 s 0.000002464 s 1.00
sum / Jax / cuda / Primal 0.000002464 s 0.000002464 s 1
sum / HLOOpt / cuda / Primal 0.000002463 s 0.000002464 s 1.00
sum / PartOpt / cuda / Primal 0.000002463 s 0.000002463 s 1
sum / IPartOpt / cuda / Primal 0.000002464 s 0.000002464 s 1
sum / DefOpt / cuda / Primal 0.000002463 s 0.000002463 s 1
sum / IDefOpt / cuda / Primal 0.000002463 s 0.000002463 s 1
sum / JaXPipe / cuda / Forward 0.000011072 s 0.000010656 s 1.04
sum / Jax / cuda / Forward 0.000010656 s 0.0000104 s 1.02
sum / HLOOpt / cuda / Forward 0.000011296 s 0.00001104 s 1.02
sum / PartOpt / cuda / Forward 0.00001088 s 0.00001104 s 0.99
sum / IPartOpt / cuda / Forward 0.00001072 s 0.000011136 s 0.96
sum / DefOpt / cuda / Forward 0.000010848 s 0.000010624 s 1.02
sum / IDefOpt / cuda / Forward 0.00001088 s 0.000010783 s 1.01
sum / JaXPipe / cuda / PreRev 0.000010304 s 0.00000992 s 1.04
sum / JaXPipe / cuda / PostRev 0.000010592 s 0.000010176 s 1.04
sum / JaXPipe / cuda / BothRev 0.00001024 s 0.000010208 s 1.00
sum / Jax / cuda / BothRev 0.000010848 s 0.000010496 s 1.03
sum / HLOOpt / cuda / PreRev 0.000010816 s 0.000010688 s 1.01
sum / HLOOpt / cuda / PostRev 0.000010432 s 0.000010432 s 1
sum / HLOOpt / cuda / BothRev 0.000010624 s 0.000009376 s 1.13
sum / PartOpt / cuda / PreRev 0.000010432 s 0.000010752 s 0.97
sum / PartOpt / cuda / PostRev 0.000010368 s 0.000010304 s 1.01
sum / PartOpt / cuda / BothRev 0.000010176 s 0.00001024 s 0.99
sum / IPartOpt / cuda / PreRev 0.000010688 s 0.000010399 s 1.03
sum / IPartOpt / cuda / PostRev 0.000009984 s 0.000010368 s 0.96
sum / IPartOpt / cuda / BothRev 0.000010304 s 0.000010432 s 0.99
sum / DefOpt / cuda / PreRev 0.000010656 s 0.00001056 s 1.01
sum / DefOpt / cuda / PostRev 0.000010272 s 0.00001056 s 0.97
sum / DefOpt / cuda / BothRev 0.000010624 s 0.00001024 s 1.04
sum / IDefOpt / cuda / PreRev 0.000010816 s 0.000010592 s 1.02
sum / IDefOpt / cuda / PostRev 0.000010656 s 0.000010912 s 0.98
sum / IDefOpt / cuda / BothRev 0.000010528 s 0.000010497 s 1.00
sum / JaXPipe / tpu / Primal 5.034750000000001e-7 s 5.168749999999999e-7 s 0.97
sum / Jax / tpu / Primal 5.471999999999999e-7 s 5.471999999999999e-7 s 1
sum / HLOOpt / tpu / Primal 5.03425e-7 s 5.170249999999999e-7 s 0.97
sum / PartOpt / tpu / Primal 5.471999999999999e-7 s 5.4715e-7 s 1.00
sum / IPartOpt / tpu / Primal 5.0325e-7 s 5.17125e-7 s 0.97
sum / DefOpt / tpu / Primal 5.47175e-7 s 5.470250000000001e-7 s 1.00
sum / IDefOpt / tpu / Primal 5.02925e-7 s 5.169000000000001e-7 s 0.97
sum / JaXPipe / tpu / Forward 0.0000015574 s 0.000001552425 s 1.00
sum / Jax / tpu / Forward 0.000001490275 s 0.0000015041 s 0.99
sum / HLOOpt / tpu / Forward 0.000001536175 s 0.000001532025 s 1.00
sum / PartOpt / tpu / Forward 0.000001488025 s 0.0000015017 s 0.99
sum / IPartOpt / tpu / Forward 0.000001533275 s 0.000001537425 s 1.00
sum / DefOpt / tpu / Forward 0.0000014934 s 0.000001507025 s 0.99
sum / IDefOpt / tpu / Forward 0.0000015309 s 0.000001533225 s 1.00
sum / JaXPipe / tpu / PreRev 9.9215e-7 s 0.000001 s 0.99
sum / JaXPipe / tpu / PostRev 0.0000010353 s 0.00000103075 s 1.00
sum / JaXPipe / tpu / BothRev 0.000001008625 s 0.00000100355 s 1.01
sum / Jax / tpu / BothRev 0.000001036775 s 0.00000103535 s 1.00
sum / HLOOpt / tpu / PreRev 0.000001007275 s 0.000001011625 s 1.00
sum / HLOOpt / tpu / PostRev 0.000001040675 s 0.00000104265 s 1.00
sum / HLOOpt / tpu / BothRev 9.97275e-7 s 0.000001003775 s 0.99
sum / PartOpt / tpu / PreRev 0.00000104875 s 0.000001035525 s 1.01
sum / PartOpt / tpu / PostRev 9.91025e-7 s 0.0000010048749999999998 s 0.99
sum / PartOpt / tpu / BothRev 0.000001039875 s 0.000001032275 s 1.01
sum / IPartOpt / tpu / PreRev 9.9225e-7 s 0.000001011075 s 0.98
sum / IPartOpt / tpu / PostRev 0.0000010369250000000002 s 0.0000010369250000000002 s 1
sum / IPartOpt / tpu / BothRev 9.9275e-7 s 0.0000010005 s 0.99
sum / DefOpt / tpu / PreRev 0.000001047175 s 0.000001044675 s 1.00
sum / DefOpt / tpu / PostRev 9.91425e-7 s 0.0000010031249999999998 s 0.99
sum / DefOpt / tpu / BothRev 0.000001033275 s 0.00000103255 s 1.00
sum / IDefOpt / tpu / PreRev 9.9125e-7 s 0.0000010112000000000002 s 0.98
sum / IDefOpt / tpu / PostRev 0.000001042375 s 0.000001041925 s 1.00
sum / IDefOpt / tpu / BothRev 9.95775e-7 s 0.000001004525 s 0.99
sum / JaXPipe / cpu / Primal 0.000014628 s 0.000007981139933690429 s 1.83
sum / Jax / cpu / Primal 0.00001465 s 0.000007507980044465512 s 1.95
sum / HLOOpt / cpu / Primal 0.00001424 s 0.000008308500073326285 s 1.71
sum / PartOpt / cpu / Primal 0.000014231 s 0.000007578700096928514 s 1.88
sum / IPartOpt / cpu / Primal 0.000014729 s 0.00000764742013416253 s 1.93
sum / DefOpt / cpu / Primal 0.00001461 s 0.000007487280381610617 s 1.95
sum / IDefOpt / cpu / Primal 0.000014898 s 0.000007432820129906759 s 2.00
sum / JaXPipe / cpu / Forward 0.000020428 s 0.000011446299795352389 s 1.78
sum / Jax / cpu / Forward 0.000020552 s 0.000011345720013196116 s 1.81
sum / HLOOpt / cpu / Forward 0.000019971 s 0.000011818420098279605 s 1.69
sum / PartOpt / cpu / Forward 0.00002012 s 0.000010789639927679672 s 1.86
sum / IPartOpt / cpu / Forward 0.000019861 s 0.000011666479986160994 s 1.70
sum / DefOpt / cpu / Forward 0.000019925 s 0.00001131830027588876 s 1.76
sum / IDefOpt / cpu / Forward 0.000020319 s 0.000011210779812245164 s 1.81
sum / JaXPipe / cpu / PreRev 0.000019488 s 0.000011084600300819147 s 1.76
sum / JaXPipe / cpu / PostRev 0.000019078000000000003 s 0.000010880019872274717 s 1.75
sum / JaXPipe / cpu / BothRev 0.000019369 s 0.000011212040044483727 s 1.73
sum / Jax / cpu / BothRev 0.000019337 s 0.000010437400087539572 s 1.85
sum / HLOOpt / cpu / PreRev 0.000019543 s 0.000011505000002216549 s 1.70
sum / HLOOpt / cpu / PostRev 0.000019368 s 0.000012359059837763198 s 1.57
sum / HLOOpt / cpu / BothRev 0.000019006 s 0.000010557379791862331 s 1.80
sum / PartOpt / cpu / PreRev 0.000019406 s 0.000010827380028786138 s 1.79
sum / PartOpt / cpu / PostRev 0.000018651 s 0.000011040499921364243 s 1.69
sum / PartOpt / cpu / BothRev 0.000019137 s 0.000011394000066502486 s 1.68
sum / IPartOpt / cpu / PreRev 0.000018736 s 0.000010900399975071197 s 1.72
sum / IPartOpt / cpu / PostRev 0.000018639 s 0.000011152820188726765 s 1.67
sum / IPartOpt / cpu / BothRev 0.000018917 s 0.000010597239997878205 s 1.79
sum / DefOpt / cpu / PreRev 0.000019038 s 0.00001111023982957704 s 1.71
sum / DefOpt / cpu / PostRev 0.000018629 s 0.000010991519957315177 s 1.69
sum / DefOpt / cpu / BothRev 0.000019094 s 0.00001119037999160355 s 1.71
sum / IDefOpt / cpu / PreRev 0.000019054 s 0.000010858259993256067 s 1.75
sum / IDefOpt / cpu / PostRev 0.000018765 s 0.000010737400079960937 s 1.75
sum / IDefOpt / cpu / BothRev 0.000019194 s 0.000010744440078269693 s 1.79
value_and_grad / JaXPipe / cpu / Primal 0.000014259099998525926 s 0.000014093239915382585 s 1.01
value_and_grad / Jax / cpu / Primal 0.00001363746003335109 s 0.000013464080038829709 s 1.01
value_and_grad / HLOOpt / cpu / Primal 0.000013362600011532775 s 0.000013915920135332271 s 0.96
value_and_grad / PartOpt / cpu / Primal 0.000013150839977242868 s 0.000013872219788026997 s 0.95
value_and_grad / IPartOpt / cpu / Primal 0.000013315900005181904 s 0.0000147368999387254 s 0.90
value_and_grad / DefOpt / cpu / Primal 0.000013488180038621066 s 0.000013849619972461368 s 0.97
value_and_grad / IDefOpt / cpu / Primal 0.000013784360025965726 s 0.000013435360015137122 s 1.03
value_and_grad / JaXPipe / cuda / Primal 0.000032606999999999995 s 0.000033472 s 0.97
value_and_grad / Jax / cuda / Primal 0.000033248 s 0.000034527 s 0.96
value_and_grad / HLOOpt / cuda / Primal 0.00003264 s 0.00003456 s 0.94
value_and_grad / PartOpt / cuda / Primal 0.000033279000000000004 s 0.000034527 s 0.96
value_and_grad / IPartOpt / cuda / Primal 0.00003328 s 0.000034144000000000004 s 0.97
value_and_grad / DefOpt / cuda / Primal 0.000033184 s 0.000034048 s 0.97
value_and_grad / IDefOpt / cuda / Primal 0.000033632 s 0.000033888 s 0.99
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 0.000022994 s 0.000014093239915382585 s 1.63
value_and_grad / Jax / cpu / Primal 0.00002281 s 0.000013464080038829709 s 1.69
value_and_grad / HLOOpt / cpu / Primal 0.000023392 s 0.000013915920135332271 s 1.68
value_and_grad / PartOpt / cpu / Primal 0.000023 s 0.000013872219788026997 s 1.66
value_and_grad / IPartOpt / cpu / Primal 0.000023049 s 0.0000147368999387254 s 1.56
value_and_grad / DefOpt / cpu / Primal 0.000023094 s 0.000013849619972461368 s 1.67
value_and_grad / IDefOpt / cpu / Primal 0.000023222 s 0.000013435360015137122 s 1.73
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / JaXPipe / cuda / Primal 1.7019069219999998 s 1.703185884 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / Jax / cuda / Primal 1.7052245339999998 s 1.706024122 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / HLOOpt / cuda / Primal 1.716587634 s 1.71937241 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / PartOpt / cuda / Primal 1.695886676 s 1.696651472 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / IPartOpt / cuda / Primal 1.693926152 s 1.694673498 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / DefOpt / cuda / Primal 1.665075052 s 1.665913635 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / IDefOpt / cuda / Primal 1.911157946 s 1.912403586 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / JaXPipe / tpu / Primal 3.038948373125 s 3.03822168 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / Jax / tpu / Primal 3.03943760875 s 3.03861887625 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / HLOOpt / tpu / Primal 3.121902861875 s 3.1209755175 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / PartOpt / tpu / Primal 3.06036302125 s 3.059448075 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / IPartOpt / tpu / Primal 3.060507205 s 3.0596829975 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / DefOpt / tpu / Primal 2.102436190625 s 2.10219199125 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_24_outer_steps_4 / IDefOpt / tpu / Primal 4.35665004375 s 4.355976231875 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / JaXPipe / cpu / Primal 6.009743265 s 6.146210463 s 0.98
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / Jax / cpu / Primal 6.056264656000001 s 6.310200541 s 0.96
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / HLOOpt / cpu / Primal 6.09426268 s 6.218582628 s 0.98
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / PartOpt / cpu / Primal 6.224692692 s 6.2440171950000005 s 1.00
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / IPartOpt / cpu / Primal 6.224546395 s 6.188685470999999 s 1.01
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / DefOpt / cpu / Primal 2.379932348 s 2.5103855910000004 s 0.95
neuralgcm_v1/deterministic_2_8_deg_inner_steps_2_outer_steps_2 / IDefOpt / cpu / Primal 6.598827396 s 6.835344457 s 0.97

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

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.

3 participants