Skip to content

anv/ci: Bump the number of tests per group, fix issues, and re-enable TGL/JSL nightlies

Valentine Burley requested to merge Valentine/mesa:anv-tpg into main

Due to the slow startup time of deqp-vk, the previous default of 500 tests per group caused the jobs to run up to twice as slowly compared to using a higher number of tests per group.

Increasing tests_per_group enables us to run more tests and/or decrease runtime, and as a result of that the nightly jobs are now within reasonable time limits, allowing them to be re-enabled.

I stress tested the jobs to make sure they are under the 10 minute time limit:

* anv-adl 1/4  : (10) 68309438:    8m55s, 68311830:    8m23s, 68312965:    8m17s, 68313444:    8m39s, 68314453:    8m20s, 68314886:    8m59s, 68315323:    8m04s, 68315739:    8m36s, 68315788:    8m10s, 68315817:   10m55s
* anv-adl 2/4  : (10) 68309439:    8m50s, 68311849:    9m24s, 68312969:    8m37s, 68313445:    8m34s, 68314454:    9m15s, 68314906:    8m13s, 68315724:    7m40s, 68315773:    8m49s, 68315789:    8m30s, 68316348:    7m43s
* anv-adl 3/4  : (10) 68309440:    9m50s, 68312115:    7m49s, 68312971:   11m31s, 68313450:    9m49s, 68314456:    8m17s, 68314907:    8m37s, 68315725:    8m31s, 68315774:    8m50s, 68315794:    7m59s, 68316349:    7m18s
* anv-adl 4/4  : (10) 68309441:    9m41s, 68312533:    9m49s, 68312972:   13m14s, 68313451:    8m36s, 68314457:    7m59s, 68314909:    8m34s, 68315726:    8m26s, 68315775:    8m29s, 68315795:    7m47s, 68316350:    8m09s
* anv-jsl 1/5  : (10) 68309433:    8m24s, 68312534:    8m43s, 68312973:   13m54s, 68313452:   15m10s, 68314486:    7m15s, 68314918:    7m35s, 68315729:    7m36s, 68315777:    7m12s, 68315796:    7m04s, 68316351:    7m25s
* anv-jsl 2/5  : (10) 68309434:    7m26s, 68312537:    7m43s, 68312975:   13m39s, 68314308:   12m46s, 68314487:    6m44s, 68315315:    7m23s, 68315730:    8m24s, 68315778:    8m29s, 68315818:    7m40s, 68317263:    6m29s
* anv-jsl 3/5  : (10) 68309435:    7m40s, 68312538:    8m55s, 68313395:   13m52s, 68314309:   11m19s, 68314488:    7m41s, 68315316:    7m58s, 68315734:    7m52s, 68315781:    8m13s, 68316238:    7m16s, 68317266:    8m21s
* anv-jsl 4/5  : (10) 68309436:    8m29s, 68312539:    7m39s, 68313421:   12m29s, 68314310:   11m20s, 68314887:    7m54s, 68315720:    7m26s, 68315761:    7m31s, 68315782:   10m18s, 68316325:    7m09s, 68317267:    7m03s
* anv-jsl 5/5  : (10) 68309437:    8m34s, 68312540:    8m34s, 68313422:   12m16s, 68314311:   21m50s, 68315721:    8m17s, 68315766:    9m05s, 68315790:    9m16s, 68316352:    7m26s, 68317270:    7m28s, 68318119:    7m25s
* anv-tgl 1/5  : (10) 68309442:    8m23s, 68312541:    8m14s, 68313423:    7m29s, 68313454:    8m00s, 68314482:    8m15s, 68314910:    7m54s, 68315727:    7m18s, 68315776:    8m02s, 68315797:    8m26s, 68317264:    7m59s
* anv-tgl 2/5  : (10) 68309443:    7m58s, 68312542:    9m21s, 68313425:    9m29s, 68314337:    8m38s, 68314489:    7m35s, 68315317:    7m21s, 68315731:    7m41s, 68315779:    7m38s, 68315798:    7m47s, 68317265:    7m47s
* anv-tgl 3/5  : (10) 68309444:    8m06s, 68312543:    9m12s, 68313439:    8m60s, 68314376:    8m14s, 68314885:    8m00s, 68315322:    7m50s, 68315735:    8m09s, 68315783:    7m42s, 68316346:    8m07s, 68317268:    7m37s
* anv-tgl 4/5  : (10) 68309445:    8m47s, 68312544:    7m38s, 68313441:    8m53s, 68314405:    8m10s, 68314888:    7m58s, 68315722:    8m17s, 68315772:    8m08s, 68315791:    7m47s, 68316772:    7m55s, 68317666:    6m49s
* anv-tgl 5/5  : (10) 68309446:    8m31s, 68312545:    8m38s, 68313447:    8m23s, 68314483:    7m26s, 68314911:    7m53s, 68315728:    7m50s, 68315780:    7m57s, 68316347:    8m01s, 68317269:    7m53s, 68318116:    7m15s

These jobs usually take 7-10 minutes, and the dashboard shows generally good availability for the devices. Therefore, I chose not to reduce parallelism but instead reduced the fractions.

Merge request reports

Loading