Closed Bug 599732 Opened 14 years ago Closed 7 years ago

Browser Hunt Benchmark slower than IE9

Categories

(Core :: Graphics, defect)

Product:
Core
Component:
Graphics
Platform:
x86
Windows 7
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED INCOMPLETE

People

(Reporter: tdowner, Unassigned)

References

(Blocks 1 open bug,
URL
)

Details

(Whiteboard: ietestdrive) 

Currently Firefox gets about 12-14 FPS in IE 9 Browser hunt Benchmark, while IE9 on the same system gets around 40+ FPS.
2D graphics performance on Windows 7 => possibly Direct2D related => CC Bas
As far as I can see Canvas2D is unrelated. It looks to be SVG, haven't looked into this yet though.
Let's call this a Graphics bug then.
Component: Canvas: 2D → Graphics
QA Contact: canvas.2d → thebes
Have you tried profiling, instead of guessing?  Is this even graphics-bound?
I haven't done anything, I admit. OK, back to Canvas 2D.
Component: Graphics → Canvas: 2D
QA Contact: thebes → canvas.2d
Indeed, this has nothing to do with canvas, though it's a tossup whether it should be in SVG or Graphics.  Here's a profile:

https://spreadsheets.google.com/ccc?key=0Asl0BZnEZ615dFBaUG55WElxNGJVbFhySm9waTh2dEE&hl=en&authkey=CJrpt-IG#gid=0

Looks like we're hitting some slow/fallback paths, including things like unpremultiplying image alpha in a number of cases.
Component: Canvas: 2D → Graphics
QA Contact: canvas.2d → thebes
(In reply to comment #4)
> Have you tried profiling, instead of guessing?  Is this even graphics-bound?

I realize my comment was a bit vague. What I meant by Canvas2D being unrelated was that the benchmark doesn't do anything Canvas, but only SVG. I didn't mean to suggest that's where the performance bottleneck was.
Whiteboard: ietestdrive
I am still seeing about 12 fps with the current Nightly, even in full-screen mode.
Well, getting a little better, I was around 17-19 FPS tonight, but still dreadfully slow.
Here are my results:
* Fx 8 or Nightly 11.0a1/20111203: 7 FPS
* IE 9: 27 FPS

Adapter Description:  Mobile Intel(R) 4 Series Express Chipset Family
Vendor ID: 8086
Device ID: 2a42
Adapter RAM: Unknown
Adapter Drivers: igdumd64 igd10umd64 igdumdx32 igd10umd32
Driver Version: 8.15.10.2555
Driver Date: 10-13-2011
Direct2D Enabled: true
DirectWrite Enabled: true (6.1.7601.17563)
ClearType Parameters: Gamma: 2200 Pixel Structure: RGB ClearType Level: 100 Enhanced Contrast: 100 
WebGL Renderer: Google Inc. -- ANGLE (Mobile Intel(R) 4 Series Express Chipset Family) -- OpenGL ES 2.0 (ANGLE 0.0.0.885)
GPU Accelerated Windows: 1/1 Direct3D 10
AzureBackend: direct2d
Summary: Browser Hunt Benchmark slower then IE9 → Browser Hunt Benchmark slower than IE9
Windows 7 with HWA
For me, Nightly 27 and IE 10 both get around 33 fps while Chrome only hits 29 fps.
But Firefox has no sound, while IE and Chrome have.
I get 40-50 fps with recent Release (25) and with MSIE 11 ~30fps.
Can we close this?
Flags: needinfo?(tdowner)
I just ran this on Nightly on Mac 10.9, and Firefox maxed out at 13FPS. I'll check on windows later on.
Windows Results:
Firefox 25, 45-48 FPS
Firefox nightly 4 FPS (We regressed in nightly)
IE 11 31 FPS

So release is good, but nightly has regressed significantly.
Flags: needinfo?(tdowner)
Tyler, are you up for finding a regression range here?
Flags: needinfo?(tdowner)
So I was going to see if bug 934860 lead to the regression here, but in fact I found the opposite (sort of): In the 2013-11-06 Nightly, this demo is unbearably slow - less than 1fps, nearly hanging the browser. In the 2013-11-07 Nightly, I get about 20-22fps, which matches current Nightly. So something in that range significantly increased the performance, though not to anything spectacularly fast.
Looks like this regressed in the 2013-10-18 Nightly (steady 60fps before, <1fps after), giving a regression range of
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=423b9c30c73d&tochange=4e7d1e2c93a6
and was partially fixed in the 2013-11-07 Nightly (<1fps before, 21fps after).
Regressed by
a417424f9213	Robert O'Callahan — Bug 923193. Make transform-origin on SVG elements use the SVG bbox as the reference rectangle. r=heycam

And backed out it in m-i today.
3fb151446ec5	Robert O'Callahan — Backout changesets 37f9a5424227, c1578a4fc86d and a417424f9213 (for bugs 933354, 929021 and 923193 respectively) while we figure out performance regressions
4a5c39f4cf22	Robert O'Callahan — Backout changesets 238b3c81f42c and rest of a417424f9213 (bug 928798 and rest of bug 923193 respectively)
Blocks: 923193
923193 is backed out. We'll re-check this before relanding.
Flags: needinfo?(roc)
Umm, but even with Bug 923193 backed out there still remains another Regression of >10fps on Release/Beta (as stated above >40fps) vs. Aurora/Trunk (~30fps).
(In reply to XtC4UaLL [:xtc4uall] from comment #21)
> Umm, but even with Bug 923193 backed out there still remains another
> Regression of >10fps on Release/Beta (as stated above >40fps) vs.
> Aurora/Trunk (~30fps).

Are you check with pgo build?
There are about 10% differences between PGO and non PGO build.


Beta26b3(pgo)(45~50fps).
http://hg.mozilla.org/releases/mozilla-beta/rev/20e947c64012
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:26.0) Gecko/20100101 Firefox/26.0 ID:20131107161719

28(pgo)(~10fps)
http://hg.mozilla.org/mozilla-central/rev/bc8c1eb0f2ba
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:28.0) Gecko/20100101 Firefox/28.0 ID:20131111030205

After Bug 923193 backed out
m-i(pgo)(50~55fps)
http://hg.mozilla.org/integration/mozilla-inbound/rev/260fbed9b3ef
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:28.0) Gecko/20100101 Firefox/28.0 ID:20131111150003
m-i(non-pgo)(40~45fps)
http://hg.mozilla.org/integration/mozilla-inbound/rev/260fbed9b3ef
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:28.0) Gecko/20100101 Firefox/28.0 ID:20131111145917
The IE Test Drive Browser Hunt benchmark page no longer exists.

https://testdrive-archive.azurewebsites.net/Performance/BrowserHunt/Default.xhtml
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.