GFX guard 8/1/2017 - 8/5/2017: graphics_WebGLAquarium |
|||||||
Issue descriptionGFX guard 8/1/2017 - 8/5/2017: graphics_WebGLAquarium regressions on certain Intel boards These regressions seem to only affect some Intel boards (Braswell?): Test: graphics_WebGLAquarium/avg_energy_rate_1000_fishes Regressions: 15 - 20% https://chromeperf.appspot.com/report?sid=2fb3eeaa04f15118951517a341aa7bab7bf8912d8c20b8809f697bf08dc0ddc7&rev=31120800959207100 https://chromeperf.appspot.com/report?sid=3b8d54faa96eaaa3e51afb9eba11de8ecb36ba814abf2161ffa8dc6f989b0bf9 ChromeOS Version range: 61.9656.0.0 - 61.9675.0.0 Chrome Version range: 61.0.3132.0 - 61.0.3136.5 CrOS changes: https://crosland.corp.google.com/log/9656.0.0..9675.0.0 CrOS changes: https://chromium.googlesource.com/chromium/src/+log/61.0.3132.0..61.0.3136.0?pretty=fuller&n=10000 Nothing obvious -- need to bisect.
,
Aug 4 2017
,
Aug 5 2017
Hmmm Hiz disable happened at https://chromium-review.googlesource.com/c/490534 ( 9505.0.0), but the regressions happened after 9656.0.0 (for reef some 60 backports seem have caused this). The affected boards seem to be reef and cyan. @marcheu has graciously volunteered to investigate...
,
Aug 5 2017
,
Aug 5 2017
,
Aug 5 2017
,
Aug 5 2017
Hah, actually it's notabug. It coincides with us having more fps: https://chromeperf.appspot.com/report?sid=20df28a7abd41d5fbd15abd3f0fdcd8884c3e193ab88f19413d2cb653f2f01e7&rev=31120800959207100 More fps = more energy, but that's ok. I wonder if we can somehow filter out that case. |
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by gurcheta...@chromium.org
, Aug 4 2017