demo is slower in Nightly compared to Chrome
Categories
(Core :: JavaScript Engine, task, P3)
Tracking
()
People
(Reporter: mayankleoboy1, Unassigned)
References
(Blocks 1 open bug)
Details
https://share.firefox.dev/3Yq4QJw
FWIW, the profiler appears to suggest that half of the time is spent in gfx (canvas) code. To remove the effect of that, I enabled the experimental gpu-canvas, and that seemed to improve the demo a bit.
|
Reporter | |
Comment 1•2 years ago
|
||
Profile with gpu-canvas : https://share.firefox.dev/3YpFWta
|
||
Comment 2•2 years ago
|
||
Glancing at the flame graph of the profile in comment 1 suggests bug 1805199 could help here too, because we're calling getters under GetName ICs. We should retest after that lands.
|
|
Reporter | |
Comment 3•2 years ago
|
||
Profile with latest nightly: https://share.firefox.dev/3V0IT0F
|
|
Reporter | |
Comment 4•2 years ago
|
||
To put this in more perspective, here is a profile with gpu-canvas disabled : https://share.firefox.dev/3BJv9QO
This shows that each "block" of JS is ~80-90ms long. Compare this with the profile in comment #0, which shows each block takes 180-190ms.
This indicates that the patch from bug 1805199 reduced the time spent in JS by ~50%.
|
|
||
Comment 5•2 years ago
|
||
(In reply to Mayank Bansal from comment #4)
To put this in more perspective, here is a profile with gpu-canvas disabled : https://share.firefox.dev/3BJv9QO
This shows that each "block" of JS is ~80-90ms long. Compare this with the profile in comment #0, which shows each block takes 180-190ms.
This indicates that the patch from bug 1805199 reduced the time spent in JS by ~50%.
Yeah, the flame graph shows a lot of time in canvas code now. I don't see any obvious JS perf cliffs. Maybe fixing bound function overhead would help some. This bug will also be easier to investigate when we have JIT code in the profiler.
|
|
Reporter | |
Updated•1 year ago
|
Description
•