New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 162514 link

Starred by 5 users

Issue metadata

Status: Archived
Email to this user bounced
Closed: Jan 2015
EstimatedDays: ----
NextAction: ----
OS: All
Pri: 2
Type: Bug

Blocked on:
issue 180885

Sign in to add a comment

Horrible scrolling webpage test case

Project Member Reported by, Nov 24 2012

Issue description


Here's a webpage with particularly bad scrolling. Like multiple seconds for each scroll change to appear on hidpi.
Paint times are dominated by image resize.  Maybe something that the new image cache and impl-side decoder will resolve? 

Comment 2 by, Nov 26 2012

Yes I think we should use this as a test case for the new image cache.

The problem with these pages is the use of large bitmap image as background, which requires upsampling with hidpi screens. Existing code doesn't handle this well by performing upstreaming over and over again, because caching for this case is broken and doesn't keep the entire resized bitmap.

The new image cache will help in a way that it resizes only once. But the behavior with hidpi screen is unknown, because upsampled bitmap takes up large amount of memory and can easily be evicted.

I think the right approach for upsampling is to not cache but let Skia do the job. Either by software bilinear filtering or bicubic via accelerated painting.

This is getting more and more problematic and we need input from Skia folks.

senorblanco, reed? What do you guys think?

Comment 3 by, Nov 26 2012

Status: Assigned
Assign it to me at the moment.
Project Member

Comment 4 by, Mar 10 2013

Labels: -Area-UI -Feature-GPU-Compositing Cr-Internals-GPU-Compositing Cr-UI
Project Member

Comment 5 by, Apr 5 2013

Labels: -Cr-Internals-GPU-Compositing Cr-Internals-Compositing
Blockedon: chromium:180885
In addition to making sure we don't do the resize over and over again due to the hi DPI image upsampling and the image caching being blown, shouldn't we make sure we're not repainting it over and over again as a result of the background-position:fixed too? If Blink is invalidating every frame then no amount of fanciness in the image cache will resolve this, right?

Making blocked on promoting those background images to their own layer, but correct me if I'm wrong here.
Labels: GPU-Assign-Over-One-Year-1-28-15

Comment 8 by, Jan 29 2015

Status: Archived
Bug too old. Probably invalid now.

Sign in to add a comment