620978 - Dromaeo.com Benchmarks hang FF3 but not Namoroka - Menu and scrolling too slow

Status: RESOLVED WORKSFORME

(Firefox :: General, defect)

Description

[Rob]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2.13) Gecko/20101203 Namoroka/3.6.13pre Firefox.3.6 ( .NET CLR 3.5.30729; .NET4.0E)
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2.13) Gecko/20101203 Namoroka/3.6.13pre Firefox.3.6 ( .NET CLR 3.5.30729; .NET4.0E)

Split from Bug 620197 Comment 13 ...

...
3a. Menu doesn't work very fast when running these Tests.
3b. Scrolling is very poor and jumpy.
3c. Typing a comment (like here) is super extra slow.
...

and, this Test does not run consistently on this version of Firefox. Sometimes it terminates silently and sometimes it causes the popup to occur repeatedly.



Running http://dromaeo.com/?all on Namoroka/3.6.13pre (20101203075014) I sometimes get this popup:

"
A script on this page may be busy, or it may have stopped responding. You can stop the script now, or you can continue to see if the script will complete.
"


It occurs during the "Strings:" Test at the END of running http://dromaeo.com/?all which differs from trying to run just http://dromaeo.com/?dromaeo-object-string , then the JavaScript terminates silently and the Test does not complete at all.


Please note that "Namoroka 4.0b9pre" can complete these Tests.

Thanks,
Rob


PS: I know _this_ version of FF is closed for all but important Bugs. This "Performance Bug" / "Termination Bug" occurs on most of the other version of FF (except for "Namoroka 4.0b9pre" at this moment) as well as this version.


Reproducible: Always

Comment 1

[Rob]

Things may be slightly worse now using [20110204033253]:

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2.15pre) Gecko/20110204 Namoroka/3.6.13pre Firefox.3.6 ( .NET CLR 3.5.30729; .NET4.0E)


I'm getting an 'Orange Line' for: RayTrace: 0.64runs/s ±36.48%
and a dismal (popup 4 times) Strings: slice: 0.08runs/s ±46.49%

Results saved. You can access them at a later time at the following URL:
http://dromaeo.com/?id=129335

We really need to LOWER the performance of Javascript by 10% to allow other processes in Namoroka (FF) to operate properly; scrolling is AWFUL and loading (or even re-loading) another Tab is poor (but not terrible considering the load).

We are WAY ahead of IE in terms of Javascript speed so slowing Javascript a bit and making the rest of the Browser operate correctly (quick enough) won't hurt us in the Benchmarks.

If ONE Tab is allowed to make the rest of the Tabs (or Menus, Scroll Bar, etc.) nearly non-functional then what is to prevent a rouge URL (or hidden link) from giving us a 'DOS' (or denial of functionality) attack or creating a 'browser-specific harassment'.

Comment 2

[Tim (fmdeveloper)]

Reporter -> Are you still experiencing this issue with the latest version of Firefox 5? Does the issue occur with the latest nightly? http://nightly.mozilla.org/

Comment 3

[Rob]

(In reply to comment #2)
> Reporter -> Are you still experiencing this issue with the latest version of
> Firefox 5? Does the issue occur with the latest nightly?
> http://nightly.mozilla.org/

Yes, the complaint is STILL valid. 

Now the Browser is "somewhat below acceptable" instead of "awful" as previously described. The scrolling is 'sticky' instead of 'dismal'. It would still be a DOS if there were one or more crafted Pages featuring JavaScript that was crafted to bring our Browser to it's knees.


A few Tests are slower.

The "Compute Bits in Byte (2)" Test dropped in performance from 332.93runs/s to 219.69runs/s.

The "Partial Sum Calculation" Test also dropped from 1040.00runs/s to 994.60runs/s.

The "Prime Number Computation" Test went from 146.53runs/s to 118.79runs/s, whereas the "Prime Number Computation (2)" went from 148.12runs/s to 595.76runs/s; which greatly favors the second test over the first.


Some Tests like "Bitwise And" (and a few others) experienced a PUNY increase in speed.


Running the OTHER Tests on the same Computer using FF 7.0a1 [20110627030739] produces results that averaged 2-7x faster than when the Tests previously ran using 3.6.13pre - great job !

3.6.13: http://dromaeo.com/?id=129335
7.0a1:  http://dromaeo.com/?id=143539



We now have TONS of room to LOWER the performance of JavaScript and allow the Browser to function smoothly. We have left IE in the 'dust of the dust' for JavaScript performance. We should concentrate on improving the smoothness of scrolling while these Tests run, and avoiding JavaScript WebPage DOS attacks.

Comment 4

[Wayne Mery (:wsmwk)]

I see no hanges with Mozilla/5.0 (Windows NT 6.0; rv:12.0a1) Gecko/20120115 Firefox/12.0a1

do you?

GC related?

CC(T+177517.1) collected: 40 (40 waiting for GC), suspected: 1988, duration: 1360 ms.

Comment 5

[Rob]

(In reply to Wayne Mery (:wsmwk) from comment #4)
> I see no hanges with Mozilla/5.0 (Windows NT 6.0; rv:12.0a1) Gecko/20120115
> Firefox/12.0a1

do you?

GC related?

CC(T+177517.1) collected: 40 (40
> waiting for GC), suspected: 1988, duration: 1360 ms.

This was reported against FF 3.6 (which I no longer have).

The complaint that running these Benchmarks "brings the Browser to it's knees is still valid". 

While it may not actually "hang" (the running of the Benchmark (IE: the JavaScript Library) NOT (ever) the Browser's Code) anymore the JavaScript Engine still gets too big a time slice and inhibits the operation of the rest of the Browser's Code (IE: scrolling and Menus).

See Comment 3.

Comment 6

[Rob]

Updated version of the Browser have evolved past the ORIGONAL Bug reported.

The newer 'problem reports' in this BR should be addressed (if they still exist) by opening a new BR.