User-Agent: Mozilla/5.0 (Windows NT 6.1; rv:6.0a1) Gecko/20110412 Firefox/6.0a1 Build Identifier: Comparing to IE10 Platform Preview, Firefox Nightly (6.0a1) is slower. (17fps IE10 - 13fps Firefox 6.0a1 on 250 fish) Reproducible: Always

Nightly - 9 FPS on 250 fish / IE9 x86 - 60 FPS on 250 fish Graphics Adapter Description NVIDIA GeForce GTX 460 Vendor ID 10de Device ID 0e22 Adapter RAM 1024 Adapter Drivers nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um Driver Version 8.17.12.6658 Driver Date 1-7-2011 Direct2D Enabled true DirectWrite Enabled true (6.1.7601.17563, font cache n/a) WebGL Renderer Google Inc. -- ANGLE -- OpenGL ES 2.0 (ANGLE 0.0.0.611) GPU Accelerated Windows 1/1 Direct3D 10 Mozilla/5.0 (Windows NT 6.1; WOW64; rv:6.0a1) Gecko/20110413 Firefox/6.0a1

While pursuing score on this test is not a target by itself, could be useful to investigate if we can be faster and where we are slower, thus I'm confirming the bug since I get constant 6fps on this test (D3D9 though, but I get good fps on fishIETank). I think however that something is broken in this test, since increasing number of fishes doesn't seem to reduce my 6fps either :\

FWIW, this demo uses a mp4 video for the water effect (and this does not work in FF). Not sure if this benefits FF performance or not though. Also, it seems that the "reflect" layer (or whatever it's called, I don't remember) inflicts a major performance decrease (by a factor of 2 on a Intel GMA*) * About:Support Graphics Section : Adapter Description: Intel(R) G41 Express Chipset Vendor ID: 8086 Device ID: 2e32 Adapter RAM: Unknown Adapter Driver: sigdumdx32 igd10umd32 Driver Version: 8.15.10.2226 Driver Date: 10-15-2010 Direct2D Enabled: true DirectWrite Enabled: true (6.1.7600.20905, font cache 62,12 MB) WebGL Renderer: Google Inc. -- ANGLE -- OpenGL ES 2.0 (ANGLE 0.0.0.611) GPU Accelerated Windows: 1/1 Direct3D 10

There is a huge improvement in last trunk (probably roc's work also benefiting speed reading and fish tank). However, the "shine" layer (wrongly called "reflect" layer in previous comment) still triggers a massive drop in performance (fwiw, it seems it is the only graphical element on top of the fishes). Also, the "shadow" and "mask" layers trigger a much more moderate drop (~5% each).

With 3D39 I don't see any improvement, still marking 6fps, IE9 can do 60fps.

(In reply to comment #5) > With 3D39 I don't see any improvement, still marking 6fps, IE9 can do 60fps. D3D9 does not benefit this demo at all. D3D9 acceleration at this point only accelerates page composition and not drawing to a canvas.

(In reply to comment #2) > While pursuing score on this test is not a target by itself, could be useful to > investigate if we can be faster and where we are slower, thus I'm confirming > the bug since I get constant 6fps on this test (D3D9 though, but I get good fps > on fishIETank). > > I think however that something is broken in this test, since increasing number > of fishes doesn't seem to reduce my 6fps either :\ The reason we are slow on this is because of custom composition operators. Since Azure fixes these performance problems and actually makes us faster than IE9 on this test, we've decided not to invest in crafting optimizations for the Canvas Direct2D backend but rather let this bug be fixed by Azure.

I can confirm with Azure (NVIDIA GT 230) Firefox 7 is faster than IE9 in this demo. However Firefox uses more CPU.

I think we can resolve this bug as WFM. I get >60 fps on Windows with Firefox 8.

After discussing on irc #gfx, we determined it might be better to ask the original reporter (or someone else who can reproduce the bug) to retest. Reopening; Oskar, can you confirm this is resolved?

Just FYI, here you are my test results on IE Test Drive demos. My PC specs ==> http://pastebin.com/6j8as3XE FishBowl (10 fish, all features ON) Mozilla Firefox 9.0.1 ==> 3-4 FPS Mozilla Firefox latest nightly + "layers.acceleration.force-enabled" = TRUE ==> 3-4 FPS Google Chrome Canary 18.0.1017.2 + "Override software rendering list" = ENABLES ==> 8 FPS Opera Browser 11.61 ==> 5 FPS Personal note: Firefox performance can be improved in this area (see Chrome results) but I don't think is worth consuming resources on old O.S. (Windows XP) and old graphics cards (Mobile Intel(R) 965 Express Chipset Family).

(In reply to RNicoletto from comment #11) > Mozilla Firefox 9.0.1 ==> 3-4 FPS > Mozilla Firefox latest nightly + "layers.acceleration.force-enabled" = TRUE > ==> 3-4 FPS > Google Chrome Canary 18.0.1017.2 + "Override software rendering list" = > ENABLES ==> 8 FPS > Opera Browser 11.61 ==> 5 FPS > > Personal note: Firefox performance can be improved in this area (see Chrome > results) but I don't think is worth consuming resources on old O.S. (Windows > XP) and old graphics cards (Mobile Intel(R) 965 Express Chipset Family). What are the results for nightly versions of Firefox and Chrome, without forcing hardware acceleration?

(In reply to Marco Castelluccio from comment #12) > > What are the results for nightly versions of Firefox and Chrome, without > forcing hardware acceleration? Mozilla Firefox latest nightly ==> 3 FPS Google Chrome Canary 18.0.1017.2 ==> 6 FPS

AFAIK, Intel 965 chipsets are blacklisted because of insufficient max texture size (see https://bugzilla.mozilla.org/show_bug.cgi?id=637089). Hence, forcing HA does nothing at all.

(In reply to William Lachance (:wlach) from comment #10) > After discussing on irc #gfx, we determined it might be better to ask the > original reporter (or someone else who can reproduce the bug) to retest. > Reopening; Oskar, can you confirm this is resolved? WFM Firefox 15 on Windows 7. I get 60 fps with 250 fishes. Original reporter has not answered and the tester in comment 11 is using a blacklisted graphic card. Should this bug be marked as WFM?

On my Mid-2012 Retina MacBook Pro: * Firefox 23 = 25 fps @ 1 fish * Firefox 26 = 25 fps @ 1 fish * Chrome 31 = 60 fps @ 1000 fish * Chrome 31 = 40 fps @ 2000 fish

On my Lenovo W540 with Intel HD Graphics 4600: * Firefox 38 = 8-10 fps @ 2000 fish * Chrome 43 = 40-60 fps @ 2000 fish * IE 11 = 28-31 @ 2000 fish If I force the programs to use the high-performance graphics card NVIDIA Quadro K1100M: * Firefox 38 = 48-56 fps @ 2000 fish (with terrible flash) * Chrome 43 = 60 fps @ 2000 fish * IE 11 = 43-45 fps @ 2000 fish (with wrong background color)