New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 923964 link

Starred by 2 users

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 1
Type: Bug-Regression



Sign in to add a comment

Video decode performance regression due to libva-intel-driver 2.3.0 uprev

Project Member Reported by dstaessens@chromium.org, Yesterday (30 hours ago)

Issue description

Comment 2 by dstaessens@chromium.org, Yesterday (30 hours ago)

Small sample of regressions:
cros-eve	video_VDAPerf	crowd2160_vp8.delivery_time.percentile_0.25	451.8%	5.603 ms
cros-snappy	video_VDAPerf	crowd2160_vp8.delivery_time.percentile_0.25	215.6%	9.741 ms
cros-nocturne	video_VDAPerf	crowd2160_vp8.delivery_time.percentile_0.25	357.4%	6.709 ms
cros-auron_yuna	video_PlaybackPerf	hw_video_dropped_frames_vp9_4k_60fps	40.6%	168.000 frames
cros-rikku	video_PlaybackPerf	hw_video_cpu_usage_h264_1080p_60fps	12.0%	142.8%

Comment 3 by hiroh@chromium.org, Yesterday (28 hours ago)

Summary: Video decode performance regression due to libva 2.3.0 uprev (was: Video decode performance regression)

Comment 4 by posciak@chromium.org, Today (22 hours ago)

Summary: Video decode performance regression due to libva-intel-driver 2.3.0 uprev (was: Video decode performance regression due to libva 2.3.0 uprev)
Local bisect of libva-intel-driver 2.1.0..2.3.0 points to https://github.com/intel/intel-vaapi-driver/commit/5bdb6754ae91c6d54290c541fc2aeaf20f6550e5 as the cause.

Comment 5 by posciak@chromium.org, Today (21 hours ago)

Cc: marc...@chromium.org

Comment 6 by hiroh@chromium.org, Today (5 hours ago)

NOTE: 3 CLs to be reverted.
libva 2.3.0 uprev: crrev.com/c/1339059
libva-intel-driver 2.3.0 uprev: crrev.com/c/1372294
libva-utils 2.3.0: crrev.com/c/1374992

Still discussing we should revert all of above them in  
https://chromium-review.googlesource.com/c/chromiumos/overlays/chromiumos-overlay/+/1426277

Comment 7 by acourbot@chromium.org, Today (4 hours ago)

Cc: haihao.x...@intel.com

Comment 8 by hiroh@chromium.org, Today (19 minutes ago)

The conclusion is we're going to revert the above three CLs tomorrow.
crrev.com/c/1344634 will be submitted soon for one-day experiment.
We're going to revert crrev.com/c/1344634 tomorrow too.

Sign in to add a comment