Closed Bug 890336 Opened 8 years ago Closed 8 years ago
How much of a startup win is delayed accelerated layer manager creation?
Bug 595277 measured a 30–50ms Ts regression due to D3D9 layer manager creation. We should remeasure this, because switching layer managers based on time is crazy and has caused hard-to-diagnose bugs (eg bug 773097). If it's still a huge win, ok fine, but if not, we should reevaluate whether we should do it any more.
Taras should sign off on any attempt to remove this :-)
Vladan can you work with Joe to measure if this makes any difference in cold or warm startup?
(In reply to Taras Glek (:taras) from comment #2) > Vladan can you work with Joe to measure if this makes any difference in cold > or warm startup? Sure
Warm startup times were 47% longer, cold startup times were 66% longer in my tests: https://docs.google.com/spreadsheet/ccc?key=0Aq4ElA0owpEbdHE0S1BiUlBTVWZqeTYtajFfanJzX0E&usp=sharing
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → FIXED
Thanks, Vladan! Given that we definitely won't remove D3D9's lazy initialization.
Just for the record, the machine was Jeff Muizelaar's test laptop "aquaman": AMD C-50 processor dual-core @ 1GHz, 2GB RAM shared with Radeon HD 6250M video card, 64-bit Windows 7.
You need to log in before you can comment on or make changes to this bug.