277984 - Bad performance laying out table containing ~10000 links

Status: RESOLVED WORKSFORME

(Core :: Layout: Tables, defect)

Description

[931987]

Firefox can't display this site:
http://www.gallileus.info/search/history/parts/5467

Comment 1

[Doug Wright]

What exactly is wrong? Nothing obviously wrong to me on Fx/1.0

Comment 2

[Mats Palmgren (inactive)]

Reproducable freeze (100% CPU) with Mozilla 2005-01-11-07 trunk Linux.

Comment 3

[Mats Palmgren (inactive)]

Attachment: Testcase (no hang, but very bad performance)

Comment 4

[Mats Palmgren (inactive)]

The site does not freeze completely, I was just being impatient - it renders
after 30 seconds or so. (2005-01-11-07 trunk Linux.)

For the testcase - performance is reasonable if I remove the TABLE container.
Changing the links to plain text words also cures the problem.

-> Tables

Comment 5

[Pavel Penaz]

30 second freeze on WinXP as well

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b) Gecko/20050114 Firefox/1.0+

Comment 6

[Markus Hübner]

Maybe this is also related to bug 54542 ?

Comment 7

[Boris Zbarsky [:bzbarsky]]

Sounds like the typical "loading a large table is O(N^2) due to the incremental
reflows it does" thing.... profile shows nothing obvious but lots of reflow
happening.

Comment 8

[Bernd]

I guess the usual picture is:
1) we get some chunk from the parser, reflow it and need CPU cycles 
2) the parser will deliver a smaller chunk on the next round as the reflow has
goten a lot of CPU cycles, we need to reflow the whole table again due to some
reflow optimization that cant be used so we eat even more CPU cycles on reflow
3) loop over to 2) and get slower with every round.

Comment 9

[Markus Hübner]

What can be done here in the parser - resp. in what component is tweaking most 
usefull/efficient?

Comment 10

[Juha Luoma]

Could we some limit on how often reflows are allowed? Like "if page still
loading do not reflow if time from last reflow < 1 s".

Comment 11

[Ryuichi KUBUKI]

*** Bug 268120 has been marked as a duplicate of this bug. ***

Comment 12

[Ryuichi KUBUKI]

WFM with Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b3)
Gecko/20050714 Firefox/1.0+
The URL loads fine for me

Comment 13

[Markus Hübner]

Reporter, can we mark this WFM?

Comment 14

[931987]

uff... sure yes. oh man, sorry.

Comment 15

[Markus Hübner]

Thanks for the respone. So, marking WFM.

Comment 16

[Mats Palmgren (inactive)]

I see no difference in performance since I last tested this bug.
Using SeaMonkey 2005-10-14-06 on Linux. I see the links up to "189" pretty fast,
then I have to wait about 30 seconds (with near 100% CPU) for the rest of the
page to render. All other SeaMonkey windows, including Mail, does not respond
until the page is fully rendered.

Comment 17

[Wayne Mery (:wsmwk)]

(In reply to comment #16)
> ... I see the links up to "189" pretty fast,

This aspect should be fixed.  Perhaps a timer or threading issue?


> then I have to wait about 30 seconds (with near 100% CPU) for the rest of the
> page to render. All other SeaMonkey windows, including Mail, does not respond
> until the page is fully rendered.

This is true also of IE. In fact IE rendered MUCH slower for me (XP).

Mats, is there a reason not to dup to bug 54542 as suggested by Markus?

Comment 18

[Bernd]

Wayne, the difference is that Mats made a profilable testcase, while bug 54542 is trash bin for our performance groupies.

Comment 19

[Wayne Mery (:wsmwk)]

(In reply to comment #18)
> Wayne, the difference is that Mats made a profilable testcase, while bug 54542
> is trash bin for our performance groupies.

Still, the testcase is clearly not linux only. And simlar bugs - eg bug 148338, bug 277984. I assume you're in tune with them.

Ah, 54542 also pretty dead (now I see the dependencies - most haven't been touched 1-2 years). So if this is kept open, shouldn't this be a blocker to bug 54542?

Comment 20

[Bernd]

the testcase loads in < 2sec so is this bug wfm?

Comment 21

[Wayne Mery (:wsmwk)]

agree, WFM Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2a1pre) Gecko/20090621 Minefield/3.6a1pre (.NET CLR 3.5.30729)