On this site we can see normal animation which looks very bad on Mozilla due to Mozilla's DHTML performance. CPU Athlon 750 RAM 392 Win 2k Trunk 2002070304 Riva TNT2 On Opera6,NS4,IE6 site works OK.

a profile would help here to see if this something we already know about or is something new.

Dunna know. What i see are moving layers with opacity (this is mozilla killer!). Afair layer move perfomance was reported many times. I'm unshure if moving layers with opacity performance was reported.

probably related to bug 139986

Probably. But it's a separate problem. Moving Layer over blank page takes some time. If this layer has opacity it takes 4-10 times more!

There is already a bug in compositor-land that has that exact profile...

Reassigning to dcone. Don, looks like its spending the bulk of its time in nsBlender::Do24Blend(float, int, int, unsigned char *, unsigned char *, unsigned char *, int, int, nsBlendQuality)

Don, might this get fixed with any short upcoming checkins?

I put in a fix to bug 148598 that fixes the tiling problems.. should speed pages up. Not sure if it will fix this.. if this is having a blender problem.

What can be done about this blender problem?

It's still very, very slow. Any chance to focus on this until Moz1.2?

After fixin bug 163528 i see quite nice impreovement, but it's still not smooth. I'm not sure if this bug should be resolved or not

I noticed over a 10x improvement from 7.0 on this page. I would say this is resoloved. There may be other speed inprovements that can be made but we have other bugs covering some of these issues, like GIF blitting speeds, DHTML card problems, etc. Marking WFE.. I am assuming kevin's fixes solved this problem.

I'm sorry but have to reopen this one. Using trunk build 2002102008 on win-xp pro,1.1ghz,512ram this is still choppy and not comparible at all to the speed of MSIE6. Might this be related to bug 170272 ?

It is related - tried with hardware acceleration one step down and things were considerably faster. Marking as DUPE. *** This bug has been marked as a duplicate of 170272 ***