Issue metadata
Sign in to add a comment
|
Investigate performance problems for 'Animometer score > Paths' benchmark |
||||||||||||||||||||||
Issue descriptionAll benchmarks are under pr.gg/animometer/developer.html - Set "Ramp" mode for benchmarking. - Set "Keep at a fixed complexity" for profiling. Set an appropriate [high] complexity next to each test's name. File any performance bugs found as blocking this issue.
,
Apr 29 2016
Ramblings from investigations. All investigations were done with my MacBook Pro Mid 2014 15" Score is from benchmarking with just "ramp". Safari out of the box: ~1950 Canary out of the box (w/ forced GPU rasterization): ~1250-1300 Chromium build out of the box (w/ mac_strip_release=0) (w/ forced GPU rasterization): ~1250-1300 Initial chromium trace shows that benchmark is main thread bound for us: [see trace_chromium_trace_osx_main_thread_bound.json.gz] [see screenshot_1] Lets CPU profile next!
,
Apr 29 2016
CPU Profiling! TL;DR we spend all of our time in Skia SW Path Drawing: See attached Instruments screenshot: And attached Instruments trace: Next up, lets see if we can force the canvas into displaylist+cwtileworker raster!
,
Apr 29 2016
===== Can ignore this if you like, this is just me playing... ===== Created franken-build of Chromium forcing canvas to be drawn on CompositorTileWorker: See attached screen shots + chromium trace. Ramp Result: still ~1250-1300. Main thread is clean! But obviously still blocked on raster. Lets do a quick CPU trace... [see attached screenshot] Horray! We successfully just moved software raster onto a different thread! #fail. TL;DR - Bottleneck is Skia SW Path rendering?
,
May 2 2016
@junov, does this look correct? Who is best person on Skia to cc/?
,
May 2 2016
Merging with Canvas Arcs, which basically has the same bottleneck |
|||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||
Comment 1 by dk...@chromium.org
, Apr 27 2016Status: Assigned (was: Untriaged)