We're currently using settings that should be about equivalent to the
"light" reference workload:
https://github.com/mozilla-b2g/gaia/blob/master/test_media/reference-workload/makeReferenceWorkload.sh#L31
Although this is mostly defined adhoc inside the test harness:
https://github.com/mozilla/eideticker/blob/master/src/eideticker/eideticker/b2gtestmixins.py
This brings up the point that we should probably come up with some way
of making sure that we're always using the same workload between our
three perf testing frameworks (b2gperf, make test-perf, and eideticker),
or at least being able to audit that's the case. Anyway, at the least
these results should be able to track relative improvements with some
reliability. :)
Will
On 14-07-17 01:51 PM, Milan Sreckovic wrote:
This is great.
What’s the “size” of the user data on the phone? One of the
reference-workload-medium or one of the other ones?
--
- Milan
On Jul 17, 2014, at 12:07 , William Lachance <[email protected]> wrote:
Just wanted to announce that we now have Eideticker startup time results coming
in for these two branches for the Flame (512M configuration), with videos:
2.0:
http://eideticker.mozilla.org/b2g/#/flame-512/2.0/b2g-camera-startup/timetostableframe
b2g-inbound:
http://eideticker.mozilla.org/b2g/#/flame-512/b2g-inbound/b2g-camera-startup/timetostableframe
These tests use a visual comparison of live video to determine when things have
loaded, which is similar to the methodology used by some of our partners.
Will
_______________________________________________
dev-b2g mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-b2g
_______________________________________________
dev-b2g mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-b2g
