MozillaZine

pcxfirefox build update to -1

Discussion of third-party/unofficial Firefox/Thunderbird/SeaMonkey builds.
Dun
 
Posts: 24
Joined: September 29th, 2015, 12:26 am
Location: Poland

Post Posted January 31st, 2016, 1:39 pm

xunxun1982 wrote:Thanks.
I remembered someone mentioned it in the past, but I forgot that later.

I will consider it, but I need test it at first, so may not contain the fix until the next edition.

Thanks for reply. If I remember well I used e-mail contact via your webpage using fake e-mail :D
Sure, take your time. https://bugzilla.mozilla.org/show_bug.c ... 208415#c17
Maybe just using max int or 100% would do the trick without breaking other stuff ;)

Of course the best would be if someone would add an about:config setting to completely disable video decoding fallback in native Firefox version... :-" I found it to be too much hassle for me and my old netbook :?

xunxun1982
 
Posts: 308
Joined: June 20th, 2011, 10:37 am

Post Posted January 31st, 2016, 7:36 pm

Dun wrote:
xunxun1982 wrote:Thanks.
I remembered someone mentioned it in the past, but I forgot that later.

I will consider it, but I need test it at first, so may not contain the fix until the next edition.

Thanks for reply. If I remember well I used e-mail contact via your webpage using fake e-mail :D
Sure, take your time. https://bugzilla.mozilla.org/show_bug.c ... 208415#c17
Maybe just using max int or 100% would do the trick without breaking other stuff ;)

Of course the best would be if someone would add an about:config setting to completely disable video decoding fallback in native Firefox version... :-" I found it to be too much hassle for me and my old netbook :?


I build a debug version:
http://sourceforge.net/projects/pcxfire ... z/download


I set three about:config
Code: Select all
mcorruptframes.init
mcorruptframes.presentedframes
mcorruptframes.mean


the related pseudo-code change is
Code: Select all
mCorruptFrames(60) -> mCorruptFrames(mcorruptframes.init)
frameStats.GetPresentedFrames() > 60 -> frameStats.GetPresentedFrames() > mcorruptframes.presentedframes
mCorruptFrames.mean() >= 2 -> mCorruptFrames.mean() >= mcorruptframes.mean


But unfortunately, because of some reasons ( maybe my GPU is so fast? ), though I tried many ways, I can't trigger the function all the time, so you need test it yourself using below method and report the result to me.
1) keep the default value, dxva enabled or disabled?
2) set the three about:config to 0 or 1, dxva enabled or disabled?
3) set the three about:config to a larger value, dxva enabled or disabled?
Last edited by xunxun1982 on January 31st, 2016, 11:42 pm, edited 1 time in total.

Dun
 
Posts: 24
Joined: September 29th, 2015, 12:26 am
Location: Poland

Post Posted January 31st, 2016, 11:14 pm

https://bugzilla.mozilla.org/attachment ... ction=diff
Here is the change which is already implemented in stable Firefox. It makes the video decoding fallback a little bit less sensitive... but still very easy to trigger (at least for me).
In my case 4k@60fps always trigger the fallback (currently only for one problematic video instead of whole Firefox session), so I shouldn't have much problems testing your changes :)

xunxun1982
 
Posts: 308
Joined: June 20th, 2011, 10:37 am

Post Posted January 31st, 2016, 11:45 pm

Dun wrote:https://bugzilla.mozilla.org/attachment.cgi?id=8656218&action=diff
Here is the change which is already implemented in stable Firefox. It makes the video decoding fallback a little bit less sensitive... but still very easy to trigger (at least for me).
In my case 4k@60fps always trigger the fallback (currently only for one problematic video instead of whole Firefox session), so I shouldn't have much problems testing your changes :)


Please try http://sourceforge.net/projects/pcxfire ... z/download

I set three about:config
Code: Select all
mcorruptframes.init
mcorruptframes.presentedframes
mcorruptframes.mean

Dun
 
Posts: 24
Joined: September 29th, 2015, 12:26 am
Location: Poland

Post Posted February 1st, 2016, 6:55 am

Thank you very much for the test build :)
Dunno how detailed feedback you want so let it be '-verbose'. I use Windows 7 x64 with disabled DWM. The E-450 APU has dual core CPU @1,6 GHz, graphics is HD6320 with 512MB (384MB by default, I use my custom unlocked bios) video memory taken from RAM. First I closed all apps possible except ESET, then I have changed values one by one and restarted after every single change. Values which didn't work - extreme shuttering
Code: Select all
mcorruptframes.init -1
mcorruptframes.presentedframes -1
mcorruptframes.mean 11

So I decided to try those
Code: Select all
mcorruptframes.init 2147483647
mcorruptframes.presentedframes 2147483647
mcorruptframes.mean 10

to be absolutely sure nothing will trigger the fallback. Those worked, I have not noticed the fallback. 300 instead of max int worked as well. Even at 4k@60fps there was less than 100% dropped frames (of course 4k@60fps is way too much to handle by my E-450).
So this is a good thing... and here comes the bad thing, not directly related to the test build.

In Firefox builds >37.0.2 the statistics functions which counts decoded/displayed/corrupted/dropped frames as well as those checking functions if x >= y if to trigger the fallback works all the time and may increase CPU usage. Could this be the reason why video playback is not as smooth as using Firefox 37.0.2 x86?

Firefox 37.0.2 x86 does not have the video decoding fallback and related statistics functions implemented (except dropped frames counter for stats for nerds). At the beginning of the video when I switch to full screen mode I have like 20-30 dropped frames total for first few seconds. Then the playback is butter smooth (0-5 dropped frames for next 10 seconds). When I move playback forwards/backwards I have some (~10-15) dropped frames one second later, then the playback is butter smooth again.

Using your pcxFirefox44 x86 test build I had more dropped frames during normal (not moving around) playback, 1-5 frames every 2-5 seconds. That was bit weird. I have noticed the same thing using Firefox Nightly builds (but those differ way too much).

As long as video decoding fallback function does not trigger, which is very easy for me, the native Firefox 44 x64 seems to perform sightly (a little bit lower CPU usage/less CPU usage spikes) than x86 version. However I will probably never get rid of 100% dropped frames thanks to bad AMD drivers...

//edit@:
xunxun1982 wrote:1) keep the default value, dxva enabled or disabled?
2) set the three about:config to 0 or 1, dxva enabled or disabled?
3) set the three about:config to a larger value, dxva enabled or disabled?

I'll test a bit more with those values.

********************************************
Using Firefox Nightly builds there is a (hidden?)
Code: Select all
gfx.blocklist.all
preference, which if is set up to '-1', it overrides all blacklists. Just saying. ;)

Dun
 
Posts: 24
Joined: September 29th, 2015, 12:26 am
Location: Poland

Post Posted February 1st, 2016, 2:21 pm

Thank you for the patch https://bugzilla.mozilla.org/show_bug.c ... 208415#c18 Let's hope someone review it and accept it.

Default values, all three 0 and all three 1 still triggers the fallback. I think I'll keep those
Code: Select all
mcorruptframes.init   300
mcorruptframes.mean   10
mcorruptframes.presentedframes   2147483647

xunxun1982
 
Posts: 308
Joined: June 20th, 2011, 10:37 am

Post Posted February 1st, 2016, 6:55 pm

Dun wrote:Thank you for the patch https://bugzilla.mozilla.org/show_bug.c ... 208415#c18 Let's hope someone review it and accept it.

Default values, all three 0 and all three 1 still triggers the fallback. I think I'll keep those
Code: Select all
mcorruptframes.init   300
mcorruptframes.mean   10
mcorruptframes.presentedframes   2147483647

I think the patch is not so good, because of some reasons, if making the corrupt frame detector adjustable, the dropped frames is very high, the performance is so bad.

Dun
 
Posts: 24
Joined: September 29th, 2015, 12:26 am
Location: Poland

Post Posted February 2nd, 2016, 3:16 am

Good to know it wasn't just me :)
Dun wrote:Using your pcxFirefox44 x86 test build I had more dropped frames during normal (not moving around) playback, 1-5 frames every 2-5 seconds. That was bit weird. I have noticed the same thing using Firefox Nightly builds (but those differ way too much)

BTW: I've send you and e-mail.

avada
 
Posts: 1447
Joined: February 10th, 2008, 6:30 am
Location: Hungary

Post Posted March 13th, 2016, 4:20 am

Hi!
Will you release your FF45 build soon?

DAC324
 
Posts: 171
Joined: October 19th, 2007, 1:54 am

Post Posted March 14th, 2016, 2:34 am

avada wrote:Hi!
Will you release your FF45 build soon?

I guess XunXun got sick of releasing a new version of his Firefox every week. There is now 45.0 but I'd expect a 45.0.1 or 45.0.2 to arrive very soon. There is already a 46.0 beta available on the Mozilla FTP server. These continuous updates can be very annoying.

Kind regards,
DAC324

avada
 
Posts: 1447
Joined: February 10th, 2008, 6:30 am
Location: Hungary

Post Posted March 14th, 2016, 2:37 am

DAC324 wrote:
avada wrote:Hi!
Will you release your FF45 build soon?

There is already a 46.0 beta available on the Mozilla FTP server. These continuous updates can be very annoying.

Kind regards,
DAC324

That's nothing new. They always release the first beta right after a stable release.

DAC324
 
Posts: 171
Joined: October 19th, 2007, 1:54 am

Post Posted March 23rd, 2016, 3:03 am

Thank you very much xunxun1982 for the new 45.0.1 release of PCXFirefox.

However, there seems to be a bug:
This version does not seem to be able to install any new add-ons.

Example:
If I try to install a dictionary, it gets stuck in Add-On manager saying "downloaded" but it will never install.
This also happens with any other add-on I try to install.

However, it seems to work in Safe Mode.

Kind regards,

DAC324

avada
 
Posts: 1447
Joined: February 10th, 2008, 6:30 am
Location: Hungary

Post Posted May 7th, 2016, 2:11 pm

Hi xunxun1982!

I'm wondering wether you're going to keep around the ability to disable signature checking. I'd very much like that because I have addons that are not signed.


They recently removed it in beta and a bunch of my addons got disabled. (Which couldn't even be re-enabled on other channels, so I had to reinstall them)
(They also kept the about:config setting lying around which now does nothing on beta.)

DAC324
 
Posts: 171
Joined: October 19th, 2007, 1:54 am

Post Posted May 21st, 2016, 2:34 pm

Crash in XUL.DLL:
Code: Select all
Name der fehlerhaften Anwendung: firefox.exe, Version: 46.0.0.5962, Zeitstempel: 0x572330d1
Name des fehlerhaften Moduls: xul.dll, Version: 46.0.0.5962, Zeitstempel: 0x57234b1f
Ausnahmecode: 0xc0000005
Fehleroffset: 0x00000000004b4408
ID des fehlerhaften Prozesses: 0x17dc
Startzeit der fehlerhaften Anwendung: 0x01d1b3a1c898c6a3
Pfad der fehlerhaften Anwendung: C:\Program Files\firefox\firefox.exe
Pfad des fehlerhaften Moduls: C:\Program Files\firefox\xul.dll
Berichtskennung: 0ccc9231-1f9a-11e6-82e0-18cf5ee65873
Vollständiger Name des fehlerhaften Pakets:
Anwendungs-ID, die relativ zum fehlerhaften Paket ist:

maur0
 
Posts: 34
Joined: October 9th, 2014, 3:46 am

Post Posted May 23rd, 2016, 6:13 pm

pcx im testin firefox 48 alpha and result is direct2d and page movment with using GPU is mutch more faster then preview version in future build your you using new render firefox 48 alpha, in your builds firefox 47 for example, result is totaly more faster and perfect for this feature.

Return to Third Party/Unofficial Builds


Who is online

Users browsing this forum: No registered users and 1 guest

cron