Issue metadata
Sign in to add a comment
|
cros/intel: Regression in graphics_WebGLAquarium/meminfo_MemUsed between 9676:9687 |
||||||||||||||||||||
Issue descriptionThe performance dashboard identified a 37.0% regression in graphics_WebGLAquarium/meminfo_MemUsed on cros-sand at revision range 31380000967600001:31380000968700000. Manual inspection shows that the performance regression occurred on many Intel boards: at least Apollo Lake, Brasswell, Broadwell, Haswell, Ivybridge. The regression seems to be more than noise, because since its onset around 9687 (2017-06-27), the performance hasn't returned to normal as of 9756 (today, 2017-07-19). I haven't determined if the bug is in Chrome or in the driver stack. Dashboard, zoomed in: https://chromeperf.appspot.com/report?masters=ChromeOS_Graphics&tests=graphics_WebGLAquarium%2Fmeminfo_MemUsed&checked=meminfo_MemUsed%2Cmeminfo_MemUsed_ref%2Cref&bots=cros-sand,cros-samus,cros-link,cros-peppy,cros-cyan&start_rev=31300000966600000&end_rev=31420000969400000 Dashboard, zoomed out: https://chromeperf.appspot.com/report?masters=ChromeOS_Graphics&tests=graphics_WebGLAquarium%2Fmeminfo_MemUsed&checked=meminfo_MemUsed%2Cmeminfo_MemUsed_ref%2Cref&bots=cros-sand,cros-samus,cros-link,cros-peppy,cros-cyan&start_rev=31300000966600000&end_rev=31600000975800000
,
Jul 19 2017
I would treat these memory usage bugs as p3. The bottom line is that this should be tested at the Chrome level, and currently it isn't. So it is mostly impossible for us to get those fixed and keep them under control.
,
Jul 19 2017
,
Aug 1
|
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by chadversary@chromium.org
, Jul 19 2017