Canvas demo (https://www.fxhash.xyz/generative/25872 ) is more than twice as fast with skia-canvas compared to gpu-canvas (and appears to improve with WPF-gpu)
Categories
(Core :: Graphics: Canvas2D, defect)
Tracking
()
People
(Reporter: mayankleoboy1, Unassigned)
References
(Blocks 1 open bug)
Details
Attachments
(1 file)
39.63 KB,
text/plain
|
Details |
Enbale gpu-canvas
Go to https://www.fxhash.xyz/generative/25872 (Exact hash used is https://gateway.fxhash2.xyz/ipfs/QmVXBEmp8HZpjxVSfGBESm9tpjT3sPUe2vPBMVUkk6AN7j/?fxhash=ood39PTMC4JUHa5eYvaratFMEry66xuaqBzRrTaEZfVXZmXjwiW)
skia-canvas: https://share.firefox.dev/3FAbYLh (5 second)
gpu-canvas: https://share.firefox.dev/3lrz4Ni (12 second)
Reporter | ||
Comment 1•2 years ago
|
||
Reporter | ||
Comment 2•2 years ago
•
|
||
Profile where i enabled the WPF rasterizer : https://share.firefox.dev/3JQqRf6 (5 second)
gfx.canvas.accelerated.aa-stroke.enabled = false
gfx.canvas.accelerated.stroke-to-fill-path = true
This is as fast as the skia-canvas version. So, AA-gpu-canvas thingy appears to be the slow thing here?
Reporter | ||
Updated•2 years ago
|
Comment 3•2 years ago
|
||
Jeff, this looks like because we don't support AA-Stroke when an alpha < 1 is used...
Updated•2 years ago
|
Reporter | ||
Comment 4•2 years ago
|
||
Profile from latest Nightly:
Gpu-canvas: https://share.firefox.dev/3N0xfAK (5.5s)
skia-canvas: https://share.firefox.dev/3XbDvKV (4.6s)
So this has improved a lot. There is some new PwebGL:Message_GetFrontBuffer at the end that takes 2 seconds.
Reporter | ||
Comment 5•1 year ago
•
|
||
edited. The results were with d2d-canvas. facepalm.
Reporter | ||
Updated•1 year ago
|
Reporter | ||
Comment 6•2 months ago
|
||
This is what I get today: https://share.firefox.dev/4gqkpZC (5s)
Calling it fixed.
Description
•