863679 - Jank on page. Profile highlights js_Date

Status: RESOLVED INVALID

(Web Compatibility :: Site Reports, defect)

Description

[Mayank Bansal]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:23.0) Gecko/20130418 Firefox/23.0
Build ID: 20130418031048

Steps to reproduce:

1. Create fresh profile
2. Go to http://support.amd.com/us/gpudownload/windows/Pages/radeonmob_win7-64.aspx#1

3.On the page, there are two tabs : "AMD catalyst package" and "Optional Downloads" .

4. Click on one tab. Notice the jank.
5. Click on the other tab. Notice the jank.

Profiler highlights js_Date, hence filing under  Core :JS engine.


Actual results:

Small janks when clicking on the tabs.

Profile : http://people.mozilla.com/~bgirard/cleopatra/#report=e9fbbc8231650c9be817cf8cfdb2c8285e6eb8f2


Expected results:

No janks.

FWIW, Chrome 27 and Opera 12.15 also jank.

Comment 1

[Mayank Bansal]

This new profile has me clicking through the tabs in quick succession about 12-15 times.Hopefully this is a better profile than the one above .

http://people.mozilla.com/~bgirard/cleopatra/#report=a1024564afa9e8ffce6030e88ee6bf09882e61c5

Comment 2

[u279076]

Benoit, can you have a look at the profile in comment 1?

Comment 3

[Benoit Girard (:BenWa)]

The profile shows that 's.mr() @ scode_support.js:504' is calling JSDate an obscene number of time. The script is minified and is using new Function to call JSDate which will hit all sorts of JS slow paths. But even given that it doesn't look like a JS optimization problem but rather the script calling get date very often.

The line:
c=s_d(c);if(e>0){a=parseInt(i=v.substring(e+5));if(a>3)a=parseFloat(i)}else if(m>0)a=parseFloat(u.substring(m+10));else a=parseFloat(v);if(a<5||v.indexOf('Opera')>=0||u.indexOf('Opera')>=0)c=s_ft(c);if(!s){s=new Object;if(!w.s_c_in){w.s_c_il=new Array;w.s_c_in=0}s._il=w.s_c_il;s._in=w.s_c_in;s._il[s._in]=s;w.s_c_in++;}s._c='s_c';(new Function("s","un","pg","ss",c))(s,un,pg,ss);return s}

Comment 4

[u279076]

So should this be moved to Tech Evangelism?

Comment 5

[Benoit Girard (:BenWa)]

If it reproduces in other browser then yes. The signs point to a bad script.

For example a script such as this one would give a similar profile:

var waitTime = Date.now()
while (waitTime + 500 > Date.now()) {}

Comment 6

[u279076]

Moving this to Tech Evangelism given this also reproduces in other browsers (see comment 0). As per comment 5, the symptoms here point to a bad script on the website and not a Firefox bug persay.

Comment 8

[Karl Dubost💡 :karlcow]

We are being redirected to http://support.amd.com/en-us/download#1
The page seems to have disappeared. Change of design probably.