Home Dashboard Directory Help
Search

offsetWidth and family are a lot slower in IE9+ than Safari 5.1 by ysitu


Status: 

Closed
 as Won't Fix Help for as Won't Fix


Type: Bug
ID: 680588
Opened: 7/24/2011 11:23:19 AM
Access Restriction: Public
0
Workaround(s)
view
0
User(s) can reproduce this bug

Description

Computing offsetWidth and similar attributes are very slow in IE9+. While the reproduction steps provided below are a rather contrived example, to see this impacting practical applications, register on the English Wikipedia and set the user script to

mathJax={fontDir:'http://cdn.mathjax.org/mathjax/latest/fonts'};
importScript('User:Nageh/mathJax.js');

then navigate to http://en.wikipedia.org/wiki/List_of_trigonometric_identities and compare the experience in IE9+ and Safari 5+.
Details
Sign in to post a comment.
Posted by Microsoft on 12/11/2012 at 2:40 PM
Thank you for your feedback. This is a valid issue but not one we expect to get to for IE10. We still have the bug tracked for the next release. Best regards, The Internet Explorer Team
Posted by ysitu on 8/1/2012 at 7:32 PM
So Windows 8 RTM'd. Did you guys fix this or not? Why is a yes/no response so hard to make?
Posted by ysitu on 7/2/2012 at 8:04 AM
MathJax has since worked around IE's cubic behavior. Now you need to load 'http://cdn.mathjax.org/mathjax/1.1-latest/MathJax.js' in the attached test case to reproduce such behavior.
Posted by ysitu on 3/1/2012 at 4:15 PM
An update on the issue, please?
Posted by Microsoft on 9/22/2011 at 2:55 PM
Thank you for your feedback.

We were able to reproduce the issue and are investigating it.

Best regards,

The Internet Explorer Team
Posted by ysitu on 8/9/2011 at 5:28 PM
I have attached a standalone test case. Extract and open test.html from the zip archive. The embedded script creates ten copies of a TeX block and invokes MathJax to typeset them. Upon completion the page contents will replaced with timing statistics.

Compare the results from IE10 PP2, Firefox 5 and Safari 5.1 (pay attention to the "New Math" lines, which show how long it took to typeset each TeX block):

(IE10 PP2)
Begin Typeset: 0
Begin Process: 17
New Math: 2879
New Math: 1522
New Math: 1688
New Math: 1900
New Math: 2152
New Math: 2428
New Math: 2740
New Math: 3090
New Math: 3548
New Math: 3829
End Process: 0
End Typeset: 0

(Firefox 5)
Begin Typeset: 0
Begin Process: 12
New Math: 1290
New Math: 567
New Math: 521
New Math: 462
New Math: 420
New Math: 420
New Math: 567
New Math: 421
New Math: 419
New Math: 425
End Process: 1
End Typeset: 0

(Safari 5.1)
Begin Typeset: 0
Begin Process: 5
New Math: 974
New Math: 315
New Math: 322
New Math: 384
New Math: 312
New Math: 303
New Math: 314
New Math: 320
New Math: 328
New Math: 331
End Process: 1
End Typeset: 0

It is evident that not only was IE10 slower right from the very beginning, it got slower over the course. In fact, IE10's data, outliers removed (the first and ninth data points), show a statistically-significant quadratic trend <http://www.wolframalpha.com/input/?i=fit+%7B%7B2%2C1522%7D%2C%7B3%2C1688%7D%2C%7B4%2C1900%7D%2C%7B5%2C2152%7D%2C%7B6%2C2428%7D%2C%7B7%2C2740%7D%2C%7B8%2C3090%7D%2C%7B10%2C3829%7D%7D>. This indicates that IE10 has *CUBIC* time complexity. In contrast, Firefox's and Safari's running times grow linearly with the number of TeX blocks.

I would find myself ashamed if I publish an application that runs two orders of growth slower on top of a larger constant factor.
Posted by ysitu on 7/29/2011 at 6:29 PM
Forgot to mention that for MathJax to work on Wikipedia, you need to go to "My preferences->Appearance->Math" and select "Leave it as TeX (for text browsers)".
Sign in to post a workaround.
File Name Submitted By Submitted On File Size  
Untitled.png 7/24/2011 235 KB
test.zip 8/9/2011 7 KB