Official firefox-pgo thread

Octoploid · 2010-07-02 06:58:50

bangkok_manouel wrote:

just in case you guys missed it, the thread is huge but all what we need to know is here:
http://gcc.gnu.org/ml/gcc/2010-06/msg00884.html

According to this it would be wise to use --enable-optimize=-O2 in mozconfig.

bangkok_manouel · 2010-07-02 07:10:55

as we, archers, are using gcc 4.5 then yes, definitely.

bobdob · 2010-07-03 09:45:08

Is anyone else get those meta bookmarks re-added every time firefox is started?
It didn't happen with 3.7a5, but just started with 4.0b1.

anish · 2010-07-07 23:49:02

I managed to get it to compile fine on a core i-5 64 bit machine, but on a core2, it just hangs here forver :
localhost.localdomain - - [06/Jul/2010 14:12:14] "GET /js-input/string-tagcloud.html HTTP/1.1" 200 -
localhost.localdomain - - [06/Jul/2010 14:12:17] "GET /js-input/string-unpack-code.html HTTP/1.1" 200 -
localhost.localdomain - - [06/Jul/2010 14:12:20] "GET /js-input/string-validate-input.html HTTP/1.1" 200 -
[dix] Could not init font path element /usr/share/fonts/OTF/, removing from list!

Its been stuck here for the last 3 days running continuously. I can stop it with ctrl+c giving this :

Traceback (most recent call last):
File "/var/abs/local/firefox-pgo/src/mozilla-1.9.2/ff-pgo/_profile/pgo/profileserver.py", line 37, in <module>
timeout = None)
File "/var/abs/local/firefox-pgo/src/mozilla-1.9.2/ff-pgo/build/pgo/automation.py", line 692, in runApp
status = self.waitForFinish(proc, utilityPath, timeout, maxTime, startTime)
File "/var/abs/local/firefox-pgo/src/mozilla-1.9.2/ff-pgo/build/pgo/automation.py", line 563, in waitForFinish
(line, didTimeout) = self.readWithTimeout(logsource, timeout)
File "/var/abs/local/firefox-pgo/src/mozilla-1.9.2/ff-pgo/build/pgo/automation.py", line 490, in readWithTimeout
(r, w, e) = select.select([f], [], [], timeout)
KeyboardInterrupt
make: *** [profiledbuild] Error 1

==> ERROR: Aborted by user! Exiting...

Anyone else seen this ?

pablox · 2010-07-08 23:18:44

Btw, I still can't use the Firefox Sync extension with this build, but I can use it with the releases directly from mozilla so I guess something is going wrong when we build it.

Last edited by pablox (2010-07-08 23:27:31)

blasse · 2010-07-10 13:43:53

There is a problem with native nss or nspr - check in sync-log, or run firefox from terminal and check there. If this is the case, recompile firefox with --wihout-system-nss or --wihout-system-nspr in your mozconfig.

pablox · 2010-07-10 14:09:29

blasse wrote:

There is a problem with native nss or nspr - check in sync-log, or run firefox from terminal and check there. If this is the case, recompile firefox with --wihout-system-nss or --wihout-system-nspr in your mozconfig.

That would help making Firefox Sync work? Or you were answering something else?

blasse · 2010-07-13 18:15:38

This should make firefox sync work. At least it worked for me

pablox · 2010-07-13 23:46:38

blasse wrote:

This should make firefox sync work. At least it worked for me :)

Thanks, I'll try to see if it works... Do you know if it has some other side-effect?

blasse · 2010-07-14 15:52:35

Besides larger size because of static library compiled? Non that I know

theringmaster · 2010-07-20 03:27:23

i've been trying to compile this all day.. and I've come farther this time. it's is failing with this line

make[1]: Leaving directory `/home/jeffp/Downloads/firefox-pgo/src/mozilla-1.9.2'
/home/jeffp/Downloads/firefox-pgo/PKGBUILD: line 141: kill: (23454) - No such process

Marcel- · 2010-07-20 08:48:30

That's the same problem described in the AUR comments? As I said, that's this problem.

Do not change your makepkg.conf, that's hardly ever necessary and won't solve this problem. Change the CFLAGS and CXXFLAGS in your PKGBUILD to something like (in your case):

export CFLAGS="-march=i686 -O2 -pipe"
export CXXFLAGS="${CFLAGS}"

Another possiblity (you don't have to use them both) is to change .makepkg on-the-fly (at least it worked for me). Add this line to your PKGBUILD, just after "$srcdir"/mozconfig is copied to .mozconfig:

sed -i 's/--enable-optimize/--enable-optimize="-O2"/' .mozconfig || return 1

theringmaster · 2010-07-20 17:36:41

I tried what you said marcel- but I still get this. The pid listed below was NOT the actual pid of that process, I was watching it this time.

/home/jeffp/Downloads/firefox-pgo/firefox-pgo/PKGBUILD: line 142: kill: (18585) - No such process
Aborting...

fabertawe · 2010-07-21 09:50:11

Hmn... I've built this (for the first time) using

export CFLAGS="-march=k8-sse3 -O2 -pipe"
export CXXFLAGS="${CFLAGS}"

and

sed -i 's/--enable-optimize/--enable-optimize="-O2"/' .mozconfig || return 1

in PKGBUILD and it's slower than bog-standard Firefox (Namaroko)!

TEST                   COMPARISON            FROM                 TO             DETAILS

=============================================================================

** TOTAL **:           *1.09x as slow*   2227.8ms +/- 5.8%   2432.8ms +/- 5.9%     significant

=============================================================================

Any ideas as to why this might be?

Cheers.

theringmaster · 2010-07-23 02:13:36

bump

pogeymanz · 2010-07-26 01:13:50

Should I be turning off noscript when building with pgo?

If there is javascript being run, should I instead leave noscript enabled, but just allow the scripts before building? How would I know what to enable?

theringmaster · 2010-07-26 04:36:19

I don't see how the noscript plugin has anything to do with building this package with pgo

Ogion · 2010-07-26 07:41:09

In this build (this PKGBUILD) it does the profile guided optimization on its own, not touching your installed firefox profile at all.
(It is possible to do it manually, from what i read)

Ogion

vae77 · 2010-07-28 14:59:38

Marcel- wrote:

About -march=native causing problems that seem to be fixed using -march=i686: it's mentioned earlier here, it affects at least 32-bit Intel Atoms as well as P4s and Core2s (confirmed by harryNID and me. Unfortunately, mentioned AUR comments got deleted).
This issue is present since FF 3.6.
BTW, what do you exactly mean by
Majki wrote:
Why mozilla don't compile firefox that way?

Thanks

pablox · 2010-07-28 23:09:10

vae77 wrote:

Marcel- wrote:
About -march=native causing problems that seem to be fixed using -march=i686: it's mentioned earlier here, it affects at least 32-bit Intel Atoms as well as P4s and Core2s (confirmed by harryNID and me. Unfortunately, mentioned AUR comments got deleted).
This issue is present since FF 3.6.
BTW, what do you exactly mean by
Majki wrote:
Why mozilla don't compile firefox that way?
Thanks =)

It's supposed that you can modify your mozconfig from this:

## Optimize
ac_add_options --enable-optimize --disable-debug --disable-tests

to this:

## Optimize
ac_add_options --enable-optimize="-O2" --disable-debug --disable-tests

I don't remember if I read it here or in the AUR comments.... at least I had compile it that way and it works.

After doing that change, I also modified the CFLAGS in PKGBUILD:

export CFLAGS="-march=core2 -msse4.1 -O2 -pipe"

Anyway, I have never found any difference in performance so I found it kind of useless. Also I found Firefox 4beta2 to be really buggy, so I stopped using it.

Last edited by pablox (2010-07-28 23:12:13)

hanbin · 2010-08-11 12:49:47

I was doing some work with libjpeg-turbo with my system.

I found a page about a integration of mozilla code and libjpeg-turbo.

libjpeg-turbo is a faster jpeg library.

the Bug page's link is here. https://bugzilla.mozilla.org/show_bug.cgi?id=573948

The patch only works with firefox-4.0 and after. When I patch it on the firefox-3.6, I dosn't works. It gives some patch failed message while patching.

Try it!

This thing is a lot faster then JUST libjpeg.

Ranguvar · 2010-08-15 05:04:36

Thanks hanbin!
I'm going to be away for ~5 days, but I'll play with that patch then(-ish) and if it seems decently reliable I'll incorporate it into firefox-pgo-beta.

Cape · 2010-08-28 13:43:05

Right... so i compiled this package on my core 2 duo using default PKGBUILD settings and the following C/CXXFLAGS:

CFLAGS="-march=native -O2 -fomit-frame-pointer -mfpmath=sse -msse4.1 -pipe -fexcess-precision=fast"

and the result is a relly slow and gigantic executable...
What i've noticed is that during the proper compilation, i had a lot of "*.gcda not found execution counts estimated" warnings.

Did i push my CFLAGS a bit too much?

Last edited by Cape (2010-08-28 13:43:47)

Majki · 2010-09-30 20:00:25

Okay, I'm gonna tell my story.
First of all - I could not compile FF 4.0 beta6 until I installed xorg-server-xvfb package. Previously I removed it from depends, becouse some people complained about in AUR comments. Finally, after two unsucessful compilations I installed this package from repo. Then everything went fine. But FF could not connect to Sync server
Error: couldn't open library
Source File: file:///usr/lib/firefox-4.0b6/components/WeaveCrypto.js
Line: 137
So spoken Firefox bug console.

I searched net and found possibly solution in comments on AUR package:

To get firefox sync working with this build system nss have to be disabled. It can be done by adding:
#Firefox sync workaround
sed -i 's/--with-system-nss/--without-system-nss/' \
.mozconfig || return 1
somewhere in PKGBUILD.

I added that lines to PKGBULID - error. Finally I thought, maybe mozconfig file?
After that compilation started and went fine. But connecting to sync server couses segfault.

I'm runnig Arch, fully updated, i686 version. I changed in PKGBULID (as usual, becouse otherwise I get random seg faults)

FROM
export CFLAGS="-march=native -O2 -pipe"
export CXXFLAGS="-march=native -O2 -pipe"

TO

export CFLAGS="-march=i686 -O2 -pipe"
export CXXFLAGS="-march=i686 -O2 -pipe"

pablox · 2010-09-30 20:54:23

I've the same problem, I couldn't run firefox sync so I had to install 3.6 again D:

Arch Linux

#626 2010-07-02 06:58:50

Re: Official firefox-pgo thread

#627 2010-07-02 07:10:55

Re: Official firefox-pgo thread

#628 2010-07-03 09:45:08

Re: Official firefox-pgo thread

#629 2010-07-07 23:49:02

Re: Official firefox-pgo thread

#630 2010-07-08 23:18:44

Re: Official firefox-pgo thread

#631 2010-07-10 13:43:53

Re: Official firefox-pgo thread

#632 2010-07-10 14:09:29

Re: Official firefox-pgo thread

#633 2010-07-13 18:15:38

Re: Official firefox-pgo thread

#634 2010-07-13 23:46:38

Re: Official firefox-pgo thread

#635 2010-07-14 15:52:35

Re: Official firefox-pgo thread

#636 2010-07-20 03:27:23

Re: Official firefox-pgo thread

#637 2010-07-20 08:48:30

Re: Official firefox-pgo thread

#638 2010-07-20 17:36:41

Re: Official firefox-pgo thread

#639 2010-07-21 09:50:11

Re: Official firefox-pgo thread

#640 2010-07-23 02:13:36

Re: Official firefox-pgo thread

#641 2010-07-26 01:13:50

Re: Official firefox-pgo thread

#642 2010-07-26 04:36:19

Re: Official firefox-pgo thread

#643 2010-07-26 07:41:09

Re: Official firefox-pgo thread

#644 2010-07-28 14:59:38

Re: Official firefox-pgo thread

#645 2010-07-28 23:09:10

Re: Official firefox-pgo thread

#646 2010-08-11 12:49:47

Re: Official firefox-pgo thread

#647 2010-08-15 05:04:36

Re: Official firefox-pgo thread

#648 2010-08-28 13:43:05

Re: Official firefox-pgo thread

#649 2010-09-30 20:00:25

Re: Official firefox-pgo thread

#650 2010-09-30 20:54:23

Re: Official firefox-pgo thread

Board footer