Closed
Bug 87808
Opened 23 years ago
Closed 21 years ago
DHTML performance exponentially degrades with amount/type of repaint necessary (such as presence of background)
Categories
(Core :: DOM: Events, defect, P2)
Core
DOM: Events
Tracking
()
RESOLVED
WORKSFORME
Future
People
(Reporter: megabyte, Assigned: kmcclusk)
References
()
Details
(Keywords: regression, top100, topperf)
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:0.9.1+) Gecko/20010624 Look on the specified page. Mouse over various buttons on the top and side of the page. DHTML is not very responsive and maxes CPU usage. The code is cross-platform and the equivalent functions are fast on NN4 and IE. The farther away the mouseover action is from the object being modified, the worse the performance is. For example, run your mouse quickly over the 4 buttons under Online Resources. The performance is poor, but acceptable. Now run it quickly over the 5 buttons under Corporate Info. The performance will be severly degraded. Now, resize your browser window in such a way that the bottom-right corner is close to the Corporate Info section. The performance will have increased. Performance should be the same at all locations. DOM1 performance should also be at least on par with Netscape 4 DOM0.
Reporter | ||
Updated•23 years ago
|
Keywords: mozilla0.9.3,
perf
Comment 1•23 years ago
|
||
What do you mean by "performance"? Is it how fast links are changing to green background, or how fast the "message div" appears? Because the "message div" shows up instantaniously on my machine, the same way as on IE. The only problem I see is that mouseover responds slower... not significantly but slower. If you move your mouse from bottom up the green background sometimes skips 5 lines or more. On IE its never more then 1.
Reporter | ||
Comment 2•23 years ago
|
||
I'm talking about the message div (but you can see the effects on the GIFs) The rate at which changes are recognized and occur. It is especially apparent at the farthest point from the div, the top menu bar. And the skipping of buttons is part of it. Just watch at the update speed of the div.. its not really the appearing or disappearing itself, its the rate at which it can handle the change. Also, if I don't tell it to go in the bottom right corner (it will default to top left corner) the speed is comparable to IE/NS4. When it is working properly you won't see skipping. Also check your resource usage, it will stay at 100% while doing mouseovers.
try www.efestival.hu the scroll effect is really not so smooth like in IE5
Reporter | ||
Comment 4•23 years ago
|
||
Performance is even worse now... On the mentioned page, moving the mouse over the sidebar triggered about 4 of the onmouseovers when moved relatively fast. Now it doesn't really trigger any unless you completely stop your mouse over them.
Blocks: 21762
Reporter | ||
Comment 5•23 years ago
|
||
In fact its so slow now, its hard to even see this bug in effect because everything is _slow_.
Comment 6•23 years ago
|
||
-> DHTML performance. Waterson if you care, give it to someone appropriate, this is not an event system issue (at least I'm pretty sure it isn't)
Assignee: joki → waterson
Reporter | ||
Comment 7•23 years ago
|
||
An animated GIF was added to the page, now DHTML peformance is exponentially worse.
Reporter | ||
Updated•23 years ago
|
Keywords: mozilla0.9.8
Updated•23 years ago
|
Target Milestone: --- → Future
Updated•23 years ago
|
Status: NEW → ASSIGNED
Reporter | ||
Comment 8•23 years ago
|
||
I don't see this bug anymore. Perhaps the fixes to bug 118933 did the trick. I'd like to see some numbers before I close out this, though.
Reporter | ||
Comment 9•23 years ago
|
||
The problem returns 2002020908
Comment 11•23 years ago
|
||
I don't see any differences between performance on the various buttons on the page. The reaction isn't lightning fast but it's acceptable. (Linux Mozilla 0.9.9).
Reporter | ||
Comment 12•23 years ago
|
||
As stated in comment #5 and comment #7, the performace is so bad that you can't even see the degredation because its too slow to compare. The performance practically makes the page unusable on my system (and there's no excuse on my system which is Dual Athlon XP 1700+, 1GB ECC DDR266 RAM, Windows XP). However, you can still test the degredation I believe, though its not exactly what I had in mind upon original report. Ex: If I run at 640x480, the page runs acceptably (I suppose), but if I run at 1600x1200 (my normal res) the page becomes nearly unusable. If you want to see how it should be, try a Feb 2nd build.
Keywords: mozilla1.0
Comment 13•23 years ago
|
||
Let's wait and see what bug 129115 does for this problem before we spend much time on this.
Reporter | ||
Comment 14•23 years ago
|
||
Okay, I've noticed something EXTREMELY interesting. The page is question works GREAT if I simply remove the background image. Updating summary to reflect. I may make a test case for this soon.
Summary: DHTML performance is bad; increasingly degrades with distance from affected object → DHTML performance exponentially degrades with amount/type of repaint necessary (such as presence of background)
Comment 15•23 years ago
|
||
There is a nice testcase from bug 135626 at http://bugzilla.mozilla.org/showattachment.cgi?attach_id=79797
Reporter | ||
Comment 16•23 years ago
|
||
"But the testcase doesn't use any DHTML at all."
Reporter | ||
Comment 17•23 years ago
|
||
Fix for bug 102321 helps this tremendously. Performance is back to the (originally reported) 20010624 level of performance. If only we could get it to the 20020202 performance level. Does anybody have any ideas on why performance was so much better for those few days in Feb?
Comment 18•23 years ago
|
||
That goes hand in hand with the things seen on http://www.world- direct.com/mozilla/dhtml/index.html [ further info in bug 107746 ] They were also fast as hell in those days of february!! Really hope we can find out what happened there.
Comment 19•23 years ago
|
||
Fixing bug 102321 helped a bit - what do the individual tests say?
Updated•22 years ago
|
Reporter | ||
Comment 20•22 years ago
|
||
updating URL. problem is still present, though perhaps not as bad as it used to be.
Keywords: mozilla0.9.9,
mozilla1.1
Comment 21•22 years ago
|
||
I see slow performance mosing over links at the top of the page: "Citizens Telephone" "Citizens InterNET" etc...
Priority: -- → P3
Comment 22•22 years ago
|
||
Using trunk build 2002122808 there is a considerable delay when mousing over the the menus on the left (with CPU usage 100%). When mousing over the links at the top the mouse cursor even jumps around. Reassigning to kmcclusk as he did some work in that area.
Assignee: waterson → kmcclusk
Status: ASSIGNED → NEW
Target Milestone: Future → ---
Assignee | ||
Comment 23•22 years ago
|
||
nsbeta1-.
![]() |
||
Comment 24•21 years ago
|
||
aaron, how about that testcase you were going to write?
![]() |
||
Comment 25•21 years ago
|
||
OK. No testcase, no way to reproduce, lots of irrelevant comments, bug has mutated OSes and become useless (for example, I see no issue on Linux on that page). Marking worksforme. If there is a site or testcase showing the problem, please file a bug with that testcase attached, and make sure not to cc the usual noise-makers.
Status: NEW → RESOLVED
Closed: 21 years ago
Resolution: --- → WORKSFORME
Reporter | ||
Comment 26•21 years ago
|
||
The page has been changed since... I can get a testcase up next week assuming the problem still exists.
You need to log in
before you can comment on or make changes to this bug.
Description
•