Closed
Bug 1164601
Opened 9 years ago
Closed 8 years ago
very jerky scrolling due to drawing background image slowly
Categories
(Core :: Graphics, defect)
Core
Graphics
Tracking
()
RESOLVED
WORKSFORME
People
(Reporter: tnikkel, Assigned: mchang)
References
(Blocks 1 open bug, )
Details
(Whiteboard: [gfx-noted])
Load http://redrockbiblecamp.com/cyclathon/
Increase the zoom (cmd-+).
Scroll.
Very jerky scrolling.
A profile says the time is spent drawing the page background image. The page background doesn't appear to be complicated in anyway.
Bisected to http://hg.mozilla.org/mozilla-central/rev/2d97da8feb72 using inbound builds.
Reporter | ||
Updated•9 years ago
|
Flags: needinfo?(mchang)
Assignee | ||
Comment 1•9 years ago
|
||
Is this on a mac?
I just tried adding create sampling restricted drawable back and it's still pretty janky on a retina display. It's not any jankier than without CSRD and deleting CSRD should be making background images a lot faster. It's also really slow on Firefox Dev Edition (Gecko 39). I don't think this is a regression from CSRD versus we just always paint background images slowly.
I also tried scrolling with APZ and it checkerboards for a few seconds. I'll take a look at it, making background images paint faster is still good.
Assignee: nobody → mchang
Blocks: paint-fast
Status: NEW → ASSIGNED
Flags: needinfo?(mchang)
Summary: very jerky scrolling due to drawing background image → very jerky scrolling due to drawing background images slowly
Reporter | ||
Comment 2•9 years ago
|
||
Yes on a mac. It's definitely a regression from http://hg.mozilla.org/mozilla-central/rev/2d97da8feb72 The difference is night and day for me. Like in the bad case we are talking almost unusable (not "it could be smoothly"). I'm not on a retina screen though. Firefox Dev, Beta and release are all fast for me.
Reporter | ||
Comment 3•9 years ago
|
||
(In reply to Timothy Nikkel (:tn) from comment #2)
> "it could be smoothly"
"it could be smoother"
Reporter | ||
Comment 4•9 years ago
|
||
And the profiler showed it was under nsDisplayCanvasBackgroundImage, indicating the root background specifically. The other background images were okay.
Summary: very jerky scrolling due to drawing background images slowly → very jerky scrolling due to drawing background image slowly
Assignee | ||
Comment 5•9 years ago
|
||
(In reply to Timothy Nikkel (:tn) from comment #2)
> Yes on a mac. It's definitely a regression from
> http://hg.mozilla.org/mozilla-central/rev/2d97da8feb72 The difference is
> night and day for me. Like in the bad case we are talking almost unusable
> (not "it could be smoothly"). I'm not on a retina screen though. Firefox
> Dev, Beta and release are all fast for me.
Oh wow... hmm. I just tried on my external 27" monitor and it's usable, but not terrible. Both aren't smooth. What kind of external monitor do you have?
Reporter | ||
Comment 6•9 years ago
|
||
(In reply to Mason Chang [:mchang] from comment #5)
> Oh wow... hmm. I just tried on my external 27" monitor and it's usable, but
> not terrible. Both aren't smooth. What kind of external monitor do you have?
It's the same on my external monitor or on my laptop screen (neither is retina).
Assignee | ||
Comment 7•9 years ago
|
||
@mstange: Can you please try the site http://redrockbiblecamp.com/cyclathon/ on your mac and see if it's significantly jankier on nightly versus dev, beta, or release edition? Thanks!
Flags: needinfo?(mstange)
Assignee | ||
Comment 8•9 years ago
|
||
So before with CSRD, we would have a giant image and paint that into one surface with CreateSamplingRestrictedDrawable, then draw each tile. with a portion of the surface. Now we have the image and create an image with a sampling rect and draw that. The performance of a repeated image will depend a lot on the tile size. I wonder if it's better to just create one giant surface, do the repeated image in one big swoop with a call to the native draw target and bypass tiling all together.
Flags: needinfo?(mstange)
Does this bug still reproduce?
Flags: needinfo?(tnikkel)
Whiteboard: [gfx-noted]
Reporter | ||
Comment 10•8 years ago
|
||
Tested on the original machine that I saw this, page scrolls fine now (with and without e10s).
Status: ASSIGNED → RESOLVED
Closed: 8 years ago
Flags: needinfo?(tnikkel)
Resolution: --- → WORKSFORME
You need to log in
before you can comment on or make changes to this bug.
Description
•