Closed
Bug 1481938
Opened 7 years ago
Closed 6 years ago
Some raptor unity-webgl subtest results are bogus when run in CI
Categories
(Testing :: Raptor, defect)
Tracking
(Not tracked)
RESOLVED
DUPLICATE
of bug 1482048
People
(Reporter: ahal, Unassigned)
References
Details
Attachments
(2 files)
I've attached an example PERFHERDER_DATA sample from a unity-webgl task. You'll notice that many of the subtests have identical replicates (and with low numbers like 20 or 50). One of the subtests (Particles) even has a value of 0 which causes the Geometric Mean (which is used as the summary value) to also be 0.
I don't think this is a bug in raptor because when I run it locally, the values more closely reflect those reported by AWFY:
https://arewefastyet.com/#machine=36&view=breakdown&suite=unity-webgl
I suspect the workers we're using for these tasks are missing some sort of hardware capability that many of these benchmarks depend on (and therefore can't progress very far, or at all).
Reporter | ||
Comment 1•7 years ago
|
||
Also worth noting this happens in both Chrome and Firefox.
Reporter | ||
Comment 2•6 years ago
|
||
Looks like it was some sort of hardware acceleration issue because bug 1482048 fixed it! Here's a sample of the data when running on the releng-hardware pool.
Reporter | ||
Updated•6 years ago
|
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → DUPLICATE
You need to log in
before you can comment on or make changes to this bug.
Description
•