328388 - Selecting/deselecting images slower in Cairo builds

Status: RESOLVED WORKSFORME

(Core :: Graphics, defect)

Description

[Martijn Wargers (dead)]

See upcoming testcase.
The 2006-02-23 cairo build is considerably slower as the 2006-02-22 non-cairo build in selecting/deselecting.

Comment 1

[Martijn Wargers (dead)]

Attachment: testcase

The testcase currently also suffers from two other cairo bugs, namely bug 324698 and bug 328386.

With the 2006-02-22 non-cairo build I get as result: 7701ms
With the 2006-02-23 cairo build I get as result: 16203ms

Comment 2

[Stuart Parmenter]

I think once 324698 is fixed this will be much less of a performance hit.

Comment 3

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Testing this with Fx2 and Fx3 shows identical time for both; can you retest?  (Yes, the blue/red issue still exists)

Comment 4

[Martijn Wargers (dead)]

Attachment: testcase2, using animated gif

This uses an animated gif as image.
I now work from a much faster computer, and with a faster computer, the difference can't be seen with the first testcase (but I bet it still does with a slower computer).
With that, I get 5141ms with a 2006-11-20 Firefox2 build, with a trunk build I get 13844ms. Also the animated gif looks different, I guess that's the anti-aliasing or something?

Comment 5

[Martijn Wargers (dead)]

Attachment: Testcase3, using a lot of animated gifs

This testcase uses a lot of animated gifs
Reults:
2006-11-20 1.8.1 branch build: 14656ms
2006-11-20 trunk build: 24062ms

Comment 6

[Martijn Wargers (dead)]

Ok, this is now worksforme.
All the testcase are now as fast or faster on current trunk build as on branch builds.
However, I want to recheck if bug 381661 gets fixed.

Comment 7

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

If this was affected by that, we have a problem; none of the testcase images are being scaled, I don't think?

Comment 8

[Martijn Wargers (dead)]

The second testcase uses a scaled image, the rest not.
This is what I get as result:

2007-05-22 build: 
Testcase 1 - 4109ms
Testcase 2 - 6641ms
Testcase 3 - 20939ms

2007-05-23: 
Testcase 1 - 4109ms
Testcase 2 - 4188ms
Testcase 3 - 20808ms

Branch build with the 3rd testcase: 11808ms

Sorry, I was a bit to quick with closing the bug. The 3rd testcase is still quite a bit slower compared to branch builds.
But with the first and second testcase, I don't see a difference anymore.

Comment 9

[Martijn Wargers (dead)]

Ok, testcase2 is slower again in current trunk builds, because of the smoothing of scaled images.
The other testcases are quicker in current trunk builds.

Comment 10

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

I see the same speed for testcases 1 and 2, which makes me believe that this is fixed.