Last active
May 3, 2022 18:37
-
-
Save davidberard98/b55d9ec36f47c4ad04fc63b1d18c9fae to your computer and use it in GitHub Desktop.
convolution measurements - A100, channels_last
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
N | tdim | wdim | ch | is_channels_last | ms | |
---|---|---|---|---|---|---|
16 | 100 | 3 | 3 | 1 | 18.43367154942825 | |
16 | 100 | 3 | 3 | 0 | 18.246989184990525 | |
16 | 100 | 3 | 7 | 1 | 18.320465798024088 | |
16 | 100 | 3 | 7 | 0 | 18.22661985643208 | |
16 | 100 | 3 | 8 | 1 | 20.038793003186584 | |
16 | 100 | 3 | 8 | 0 | 18.317282549105585 | |
16 | 100 | 5 | 3 | 1 | 18.30171929905191 | |
16 | 100 | 5 | 3 | 0 | 18.09040275402367 | |
16 | 100 | 5 | 7 | 1 | 29.462874494493008 | |
16 | 100 | 5 | 7 | 0 | 25.831757998093963 | |
16 | 100 | 5 | 8 | 1 | 51.52064310386777 | |
16 | 100 | 5 | 8 | 0 | 31.876941304653883 | |
16 | 100 | 7 | 3 | 1 | 18.537260207813233 | |
16 | 100 | 7 | 3 | 0 | 18.110481044277552 | |
16 | 100 | 7 | 7 | 1 | 53.05891130119562 | |
16 | 100 | 7 | 7 | 0 | 47.530143009498715 | |
16 | 100 | 7 | 8 | 1 | 99.12664885632694 | |
16 | 100 | 7 | 8 | 0 | 59.04654811602086 | |
16 | 100 | 9 | 3 | 1 | 18.69109240360558 | |
16 | 100 | 9 | 3 | 0 | 18.230828805826604 | |
16 | 100 | 9 | 7 | 1 | 81.12724008969963 | |
16 | 100 | 9 | 7 | 0 | 67.67505290918052 | |
16 | 100 | 9 | 8 | 1 | 160.0374965928495 | |
16 | 100 | 9 | 8 | 0 | 83.96956697106361 | |
16 | 100 | 11 | 3 | 1 | 24.356922297738492 | |
16 | 100 | 11 | 3 | 0 | 20.960539998486638 | |
16 | 100 | 11 | 7 | 1 | 118.52847051341087 | |
16 | 100 | 11 | 7 | 0 | 101.30427253898233 | |
16 | 100 | 11 | 8 | 1 | 236.2886720802635 | |
16 | 100 | 11 | 8 | 0 | 126.48218043614179 | |
16 | 244 | 3 | 3 | 1 | 21.351740299724042 | |
16 | 244 | 3 | 3 | 0 | 18.690762598998845 | |
16 | 244 | 3 | 7 | 1 | 45.8222656045109 | |
16 | 244 | 3 | 7 | 0 | 78.0768459662795 | |
16 | 244 | 3 | 8 | 1 | 53.245655214414 | |
16 | 244 | 3 | 8 | 0 | 99.66220404021442 | |
16 | 244 | 5 | 3 | 1 | 45.01768241170794 | |
16 | 244 | 5 | 3 | 0 | 27.741778106428683 | |
16 | 244 | 5 | 7 | 1 | 101.61644197069108 | |
16 | 244 | 5 | 7 | 0 | 144.5265645161271 | |
16 | 244 | 5 | 8 | 1 | 120.28961698524654 | |
16 | 244 | 5 | 8 | 0 | 185.34037203062326 | |
16 | 244 | 7 | 3 | 1 | 80.84342093206942 | |
16 | 244 | 7 | 3 | 0 | 49.77555158548057 | |
16 | 244 | 7 | 7 | 1 | 185.42232853360474 | |
16 | 244 | 7 | 7 | 0 | 264.8957180790603 | |
16 | 244 | 7 | 8 | 1 | 220.69369396194816 | |
16 | 244 | 7 | 8 | 0 | 341.45430591888726 | |
16 | 244 | 9 | 3 | 1 | 128.54479893576354 | |
16 | 244 | 9 | 3 | 0 | 70.11849293485284 | |
16 | 244 | 9 | 7 | 1 | 296.58175096847117 | |
16 | 244 | 9 | 7 | 0 | 374.50749101117253 | |
16 | 244 | 9 | 8 | 1 | 353.35517884232104 | |
16 | 244 | 9 | 8 | 0 | 483.67772600613534 | |
16 | 244 | 11 | 3 | 1 | 188.00333852414042 | |
16 | 244 | 11 | 3 | 0 | 104.70633255317807 | |
16 | 244 | 11 | 7 | 1 | 435.82991999574006 | |
16 | 244 | 11 | 7 | 0 | 563.7340219691396 | |
16 | 244 | 11 | 8 | 1 | 520.4557671677321 | |
16 | 244 | 11 | 8 | 0 | 728.8549048826098 | |
16 | 512 | 3 | 3 | 1 | 102.50197246205062 | |
16 | 512 | 3 | 3 | 0 | 78.18600907921791 | |
16 | 512 | 3 | 7 | 1 | 201.3016720302403 | |
16 | 512 | 3 | 7 | 0 | 342.28865802288055 | |
16 | 512 | 3 | 8 | 1 | 231.97088111191988 | |
16 | 512 | 3 | 8 | 0 | 437.2869550716132 | |
16 | 512 | 5 | 3 | 1 | 206.11139107495546 | |
16 | 512 | 5 | 3 | 0 | 130.80083951354027 | |
16 | 512 | 5 | 7 | 1 | 445.2961441129446 | |
16 | 512 | 5 | 7 | 0 | 630.8648940175772 | |
16 | 512 | 5 | 8 | 1 | 525.1300630625337 | |
16 | 512 | 5 | 8 | 0 | 814.3553459085524 | |
16 | 512 | 7 | 3 | 1 | 362.5257189851254 | |
16 | 512 | 7 | 3 | 0 | 227.82374895177782 | |
16 | 512 | 7 | 7 | 1 | 811.7622800637037 | |
16 | 512 | 7 | 7 | 0 | 1157.9886570107192 | |
16 | 512 | 7 | 8 | 1 | 967.0884439256042 | |
16 | 512 | 7 | 8 | 0 | 1505.2870861254632 | |
16 | 512 | 9 | 3 | 1 | 571.1262919940054 | |
16 | 512 | 9 | 3 | 0 | 315.63887116499245 | |
16 | 512 | 9 | 7 | 1 | 1300.1056250650436 | |
16 | 512 | 9 | 7 | 0 | 1636.7696318775415 | |
16 | 512 | 9 | 8 | 1 | 1544.389443937689 | |
16 | 512 | 9 | 8 | 0 | 2127.304224995896 | |
16 | 512 | 11 | 3 | 1 | 832.3811821173877 | |
16 | 512 | 11 | 3 | 0 | 467.4088421743363 | |
16 | 512 | 11 | 7 | 1 | 1910.3905488736928 | |
16 | 512 | 11 | 7 | 0 | 2466.9426139444113 | |
16 | 512 | 11 | 8 | 1 | 2278.6464418750256 | |
16 | 512 | 11 | 8 | 0 | 3209.5337680075318 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
N | tdim | wdim | ch | is_channels_last | ms | |
---|---|---|---|---|---|---|
16 | 100 | 3 | 3 | 1 | 18.617491715122018 | |
16 | 100 | 3 | 3 | 0 | 18.78675235202536 | |
16 | 100 | 3 | 7 | 1 | 19.152427499648184 | |
16 | 100 | 3 | 7 | 0 | 18.516116950195283 | |
16 | 100 | 3 | 8 | 1 | 20.04905662033707 | |
16 | 100 | 3 | 8 | 0 | 18.60648614820093 | |
16 | 100 | 5 | 3 | 1 | 19.169622543267906 | |
16 | 100 | 5 | 3 | 0 | 18.820264446549118 | |
16 | 100 | 5 | 7 | 1 | 29.462046385742724 | |
16 | 100 | 5 | 7 | 0 | 25.832909997552633 | |
16 | 100 | 5 | 8 | 1 | 51.55018719378859 | |
16 | 100 | 5 | 8 | 0 | 31.883728713728487 | |
16 | 100 | 7 | 3 | 1 | 19.15375804528594 | |
16 | 100 | 7 | 3 | 0 | 19.243654911406338 | |
16 | 100 | 7 | 7 | 1 | 53.01895779557526 | |
16 | 100 | 7 | 7 | 0 | 47.49531680718064 | |
16 | 100 | 7 | 8 | 1 | 98.84221595712006 | |
16 | 100 | 7 | 8 | 0 | 58.97610879037529 | |
16 | 100 | 9 | 3 | 1 | 19.0016643027775 | |
16 | 100 | 9 | 3 | 0 | 18.49844044772908 | |
16 | 100 | 9 | 7 | 1 | 81.02516806684434 | |
16 | 100 | 9 | 7 | 0 | 67.36795790493488 | |
16 | 100 | 9 | 8 | 1 | 160.0180856185034 | |
16 | 100 | 9 | 8 | 0 | 83.91591906547546 | |
16 | 100 | 11 | 3 | 1 | 24.340565013699234 | |
16 | 100 | 11 | 3 | 0 | 20.957220904529095 | |
16 | 100 | 11 | 7 | 1 | 118.49813652224839 | |
16 | 100 | 11 | 7 | 0 | 101.25651210546494 | |
16 | 100 | 11 | 8 | 1 | 236.2594180740416 | |
16 | 100 | 11 | 8 | 0 | 126.45251443609595 | |
16 | 244 | 3 | 3 | 1 | 21.335488883778453 | |
16 | 244 | 3 | 3 | 0 | 18.89237290015444 | |
16 | 244 | 3 | 7 | 1 | 45.82061788532883 | |
16 | 244 | 3 | 7 | 0 | 78.01525900140405 | |
16 | 244 | 3 | 8 | 1 | 53.23789780959487 | |
16 | 244 | 3 | 8 | 0 | 99.63596891611814 | |
16 | 244 | 5 | 3 | 1 | 45.00673818401992 | |
16 | 244 | 5 | 3 | 0 | 27.740272181108594 | |
16 | 244 | 5 | 7 | 1 | 101.5442309435457 | |
16 | 244 | 5 | 7 | 0 | 144.46621004026383 | |
16 | 244 | 5 | 8 | 1 | 120.23189302999526 | |
16 | 244 | 5 | 8 | 0 | 185.31600560527295 | |
16 | 244 | 7 | 3 | 1 | 80.83452610298991 | |
16 | 244 | 7 | 3 | 0 | 49.78251978754997 | |
16 | 244 | 7 | 7 | 1 | 185.226792935282 | |
16 | 244 | 7 | 7 | 0 | 264.93781502358615 | |
16 | 244 | 7 | 8 | 1 | 220.7418200559914 | |
16 | 244 | 7 | 8 | 0 | 341.69769706204534 | |
16 | 244 | 9 | 3 | 1 | 128.59959551133215 | |
16 | 244 | 9 | 3 | 0 | 70.15452487394214 | |
16 | 244 | 9 | 7 | 1 | 296.9771730713546 | |
16 | 244 | 9 | 7 | 0 | 374.7915520798415 | |
16 | 244 | 9 | 8 | 1 | 353.53076411411166 | |
16 | 244 | 9 | 8 | 0 | 483.90447115525603 | |
16 | 244 | 11 | 3 | 1 | 188.1598614854738 | |
16 | 244 | 11 | 3 | 0 | 104.75643642712384 | |
16 | 244 | 11 | 7 | 1 | 436.16899801418185 | |
16 | 244 | 11 | 7 | 0 | 564.1035118605942 | |
16 | 244 | 11 | 8 | 1 | 520.6638311501592 | |
16 | 244 | 11 | 8 | 0 | 729.0159750264138 | |
16 | 512 | 3 | 3 | 1 | 102.5122330756858 | |
16 | 512 | 3 | 3 | 0 | 78.164700884372 | |
16 | 512 | 3 | 7 | 1 | 201.3238740619272 | |
16 | 512 | 3 | 7 | 0 | 342.31415297836065 | |
16 | 512 | 3 | 8 | 1 | 231.97667696513236 | |
16 | 512 | 3 | 8 | 0 | 437.3319619335234 | |
16 | 512 | 5 | 3 | 1 | 206.09992602840066 | |
16 | 512 | 5 | 3 | 0 | 130.80425304360688 | |
16 | 512 | 5 | 7 | 1 | 445.32433804124594 | |
16 | 512 | 5 | 7 | 0 | 631.1094518750906 | |
16 | 512 | 5 | 8 | 1 | 525.3238540608436 | |
16 | 512 | 5 | 8 | 0 | 814.3894399981946 | |
16 | 512 | 7 | 3 | 1 | 362.3560150153935 | |
16 | 512 | 7 | 3 | 0 | 227.8254448901862 | |
16 | 512 | 7 | 7 | 1 | 814.5140670239925 | |
16 | 512 | 7 | 7 | 0 | 1158.3332540467381 | |
16 | 512 | 7 | 8 | 1 | 965.2099870145321 | |
16 | 512 | 7 | 8 | 0 | 1503.144463058561 | |
16 | 512 | 9 | 3 | 1 | 571.2393971625715 | |
16 | 512 | 9 | 3 | 0 | 315.47448807395995 | |
16 | 512 | 9 | 7 | 1 | 1300.385845825076 | |
16 | 512 | 9 | 7 | 0 | 1636.8507500737906 | |
16 | 512 | 9 | 8 | 1 | 1545.9112760145217 | |
16 | 512 | 9 | 8 | 0 | 2127.843477996066 | |
16 | 512 | 11 | 3 | 1 | 832.3594650719315 | |
16 | 512 | 11 | 3 | 0 | 467.1126550529152 | |
16 | 512 | 11 | 7 | 1 | 1911.026865011081 | |
16 | 512 | 11 | 7 | 0 | 2464.282466098666 | |
16 | 512 | 11 | 8 | 1 | 2277.7668449562043 | |
16 | 512 | 11 | 8 | 0 | 3211.3462639972568 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
N | tdim | wdim | ch | is_channels_last | ms | |
---|---|---|---|---|---|---|
16 | 100 | 3 | 3 | 1 | 17.98505939077586 | |
16 | 100 | 3 | 3 | 0 | 17.783134942874312 | |
16 | 100 | 3 | 7 | 1 | 18.125824455637485 | |
16 | 100 | 3 | 7 | 0 | 18.00207723863423 | |
16 | 100 | 3 | 8 | 1 | 20.031808386556804 | |
16 | 100 | 3 | 8 | 0 | 17.9782064515166 | |
16 | 100 | 5 | 3 | 1 | 17.987724544946104 | |
16 | 100 | 5 | 3 | 0 | 17.954158294014633 | |
16 | 100 | 5 | 7 | 1 | 29.481642902828753 | |
16 | 100 | 5 | 7 | 0 | 25.864325487054884 | |
16 | 100 | 5 | 8 | 1 | 51.561471400782466 | |
16 | 100 | 5 | 8 | 0 | 31.903302296996117 | |
16 | 100 | 7 | 3 | 1 | 18.324633350130167 | |
16 | 100 | 7 | 3 | 0 | 17.88231561658904 | |
16 | 100 | 7 | 7 | 1 | 53.0493127880618 | |
16 | 100 | 7 | 7 | 0 | 47.548712603747845 | |
16 | 100 | 7 | 8 | 1 | 98.93104806542397 | |
16 | 100 | 7 | 8 | 0 | 59.03653670102358 | |
16 | 100 | 9 | 3 | 1 | 18.25640833703801 | |
16 | 100 | 9 | 3 | 0 | 18.012333416845646 | |
16 | 100 | 9 | 7 | 1 | 81.1137561686337 | |
16 | 100 | 9 | 7 | 0 | 67.46744713746011 | |
16 | 100 | 9 | 8 | 1 | 160.08370905183256 | |
16 | 100 | 9 | 8 | 0 | 83.99839396588504 | |
16 | 100 | 11 | 3 | 1 | 24.344576080329716 | |
16 | 100 | 11 | 3 | 0 | 20.973224588669837 | |
16 | 100 | 11 | 7 | 1 | 118.53211000561714 | |
16 | 100 | 11 | 7 | 0 | 101.33329965174198 | |
16 | 100 | 11 | 8 | 1 | 236.33953696116805 | |
16 | 100 | 11 | 8 | 0 | 126.53844512533396 | |
16 | 244 | 3 | 3 | 1 | 21.352207893505692 | |
16 | 244 | 3 | 3 | 0 | 18.491640803404152 | |
16 | 244 | 3 | 7 | 1 | 45.84740300197154 | |
16 | 244 | 3 | 7 | 0 | 78.09652201831341 | |
16 | 244 | 3 | 8 | 1 | 53.2672727946192 | |
16 | 244 | 3 | 8 | 0 | 99.70052912831306 | |
16 | 244 | 5 | 3 | 1 | 45.04391020163894 | |
16 | 244 | 5 | 3 | 0 | 27.769387210719287 | |
16 | 244 | 5 | 7 | 1 | 101.69024544302374 | |
16 | 244 | 5 | 7 | 0 | 144.58980155177414 | |
16 | 244 | 5 | 8 | 1 | 120.33829209394753 | |
16 | 244 | 5 | 8 | 0 | 185.4973214212805 | |
16 | 244 | 7 | 3 | 1 | 80.91538213193417 | |
16 | 244 | 7 | 3 | 0 | 49.828163301572204 | |
16 | 244 | 7 | 7 | 1 | 185.3654155274853 | |
16 | 244 | 7 | 7 | 0 | 265.0411769282073 | |
16 | 244 | 7 | 8 | 1 | 220.8123339805752 | |
16 | 244 | 7 | 8 | 0 | 341.7605140712112 | |
16 | 244 | 9 | 3 | 1 | 128.62315052188933 | |
16 | 244 | 9 | 3 | 0 | 70.326424902305 | |
16 | 244 | 9 | 7 | 1 | 296.890138881281 | |
16 | 244 | 9 | 7 | 0 | 375.1403170172125 | |
16 | 244 | 9 | 8 | 1 | 353.51713793352246 | |
16 | 244 | 9 | 8 | 0 | 484.00816414505243 | |
16 | 244 | 11 | 3 | 1 | 188.15834308043122 | |
16 | 244 | 11 | 3 | 0 | 104.76912185549736 | |
16 | 244 | 11 | 7 | 1 | 436.18130101822317 | |
16 | 244 | 11 | 7 | 0 | 564.080409007147 | |
16 | 244 | 11 | 8 | 1 | 520.6471499986947 | |
16 | 244 | 11 | 8 | 0 | 729.3332149274647 | |
16 | 512 | 3 | 3 | 1 | 102.52439265605062 | |
16 | 512 | 3 | 3 | 0 | 78.38175096549094 | |
16 | 512 | 3 | 7 | 1 | 201.73748093657196 | |
16 | 512 | 3 | 7 | 0 | 342.7497821394354 | |
16 | 512 | 3 | 8 | 1 | 232.08913300186396 | |
16 | 512 | 3 | 8 | 0 | 437.38882592879236 | |
16 | 512 | 5 | 3 | 1 | 206.16268599405885 | |
16 | 512 | 5 | 3 | 0 | 130.8596400776878 | |
16 | 512 | 5 | 7 | 1 | 445.52418286912143 | |
16 | 512 | 5 | 7 | 0 | 631.3505121506751 | |
16 | 512 | 5 | 8 | 1 | 525.5342728924006 | |
16 | 512 | 5 | 8 | 0 | 814.5755738951266 | |
16 | 512 | 7 | 3 | 1 | 362.6846210099757 | |
16 | 512 | 7 | 3 | 0 | 228.12796500511467 | |
16 | 512 | 7 | 7 | 1 | 812.5514201819897 | |
16 | 512 | 7 | 7 | 0 | 1158.9657519944012 | |
16 | 512 | 7 | 8 | 1 | 965.9182881005108 | |
16 | 512 | 7 | 8 | 0 | 1504.0805558674037 | |
16 | 512 | 9 | 3 | 1 | 571.675136918202 | |
16 | 512 | 9 | 3 | 0 | 316.26641703769565 | |
16 | 512 | 9 | 7 | 1 | 1299.9964321497828 | |
16 | 512 | 9 | 7 | 0 | 1636.111926054582 | |
16 | 512 | 9 | 8 | 1 | 1548.0412761680782 | |
16 | 512 | 9 | 8 | 0 | 2131.553997984156 | |
16 | 512 | 11 | 3 | 1 | 833.4145059343427 | |
16 | 512 | 11 | 3 | 0 | 468.1004530284554 | |
16 | 512 | 11 | 7 | 1 | 1914.6338149439543 | |
16 | 512 | 11 | 7 | 0 | 2467.0993939507753 | |
16 | 512 | 11 | 8 | 1 | 2281.0571540612727 | |
16 | 512 | 11 | 8 | 0 | 3211.363078095019 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import torch | |
from torch.utils.benchmark import Timer | |
def conv_add_relu_x100(t, w, pad: int): | |
for i in range(100): | |
conv_out = torch.nn.functional.conv2d(t, w, bias = None, stride = 1, padding = pad) | |
add_out = conv_out.add_(t) | |
relu_out = torch.relu_(add_out) | |
t = relu_out | |
return t | |
def run_test(t, w, pad, *args, **kwargs): | |
t = t.cuda() | |
w = w.cuda() | |
script = torch.jit.script(conv_add_relu_x100) | |
for i in range(20): | |
res = script(t, w, pad) | |
t = Timer(stmt="fn(*inputs)", globals={"fn": script, "inputs": [t, w, pad]}) | |
times = t.blocked_autorange() | |
return times.median * 1000 | |
def main(): | |
configs = [] | |
print("N,tdim,wdim,ch,is_channels_last,ms") | |
N = 16 | |
for tdim in [100, 244, 512]: | |
for wdim in [3, 5, 7, 9, 11]: | |
for ch in [3, 7, 8]: | |
for mem_format in [1,2]: | |
t = torch.rand(N, ch, tdim, tdim).cuda() | |
w = torch.rand(ch, ch, wdim, wdim).cuda() | |
pad = wdim//2 | |
if mem_format == 1: | |
t = t.to(memory_format=torch.channels_last) | |
w = w.to(memory_format=torch.channels_last) | |
if mem_format == 2: | |
t = t.to(memory_format=torch.contiguous_format) | |
w = w.to(memory_format=torch.contiguous_format) | |
result = run_test(t, w, pad) | |
print(','.join([str(x) for x in [N, tdim, wdim, ch, 1 if mem_format == 1 else 0, result]])) | |
if __name__ == "__main__": | |
main() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment