Official firefox-pgo thread

pyther · 2009-07-18 10:56:13

@blasse

What does -rpath,/usr/lib/firefox-${pkgver} do and why is it in the pkgbuild?

blasse · 2009-07-18 11:00:07

It is taken from official arch firefox pkgbuilds and it points dynamic linker to link library's to this exact folder.

t3ddy · 2009-07-18 16:54:50

blasse wrote:

Disable gnome-vfs support in mozconfig. That should help.

I've added the line:

ac_add_options --disable-gnomevfs

in mozconfig, the result is the same
I really can't understand what's the problem

pyther · 2009-07-18 18:11:43

There is some type of segfault/error:

[firefox-bin[27225] general protection ip:7f1dc3c58431 sp:7fff0dcde988 error:0 in libc-2.10.1.so[7f1dc3bdf000+149000]

- from dmesg

blasse · 2009-07-18 21:00:57

That's something new. I will investigate it in my spare time

savagenator · 2009-07-18 22:14:18

pyther wrote:

There is some type of segfault/error:

[firefox-bin[27225] general protection ip:7f1dc3c58431 sp:7fff0dcde988 error:0 in libc-2.10.1.so[7f1dc3bdf000+149000]

- from dmesg

I'm getting the same thing from swftweasel.

firefox-bin[11043] general protection ip:2ac819037431 sp:7fff23473d18 error:0 in libc-2.10.1.so[2ac818fbe000+149000]

The pid is in the [], and the ip: and sp: values change.

Also keep in mind I cannot compile firefox either. At the moment i'm compiling the firefox-pgo package from AUR in a 32-bit chroot.

EDIT: ok, wth is going on....I'm getting a similar error compiling using 32-bit!

OBJDIR=/tmp/yaourt-tmp-root/aur-firefox-pgo/firefox-pgo/src/mozilla-1.9.1/ff-pgo python /tmp/yaourt-tmp-root/aur-firefox-pgo/firefox-pgo/src/mozilla-1.9.1/ff-pgo/_profile/pgo/profileserver.py
INFO | (automation.py) | Application pid: 13367
Error: no display specified
TEST-UNEXPECTED-FAIL | (automation.py) | Exited with code 1 during test run
INFO | (automation.py) | Application ran for: 0:00:00.346735
make: *** [profiledbuild] Error 1

Last edited by savagenator (2009-07-18 22:24:59)

oddsocks · 2009-07-19 02:13:47

I just ended up compiling from minefield again. Writing this from the minefield build. I really don't get what the problem is with pgo and firefox

blasse · 2009-07-19 07:50:12

savagenator wrote:

EDIT: ok, wth is going on....I'm getting a similar error compiling using 32-bit!

OBJDIR=/tmp/yaourt-tmp-root/aur-firefox-pgo/firefox-pgo/src/mozilla-1.9.1/ff-pgo python /tmp/yaourt-tmp-root/aur-firefox-pgo/firefox-pgo/src/mozilla-1.9.1/ff-pgo/_profile/pgo/profileserver.py
INFO | (automation.py) | Application pid: 13367
Error: no display specified
TEST-UNEXPECTED-FAIL | (automation.py) | Exited with code 1 during test run
INFO | (automation.py) | Application ran for: 0:00:00.346735
make: *** [profiledbuild] Error 1

Not very similar - no display specified means you ave specified wrong export to other display. Try compiling it the way it's written in pkgbuild.

savagenator · 2009-07-19 18:06:04

blasse wrote:

savagenator wrote:
EDIT: ok, wth is going on....I'm getting a similar error compiling using 32-bit!
OBJDIR=/tmp/yaourt-tmp-root/aur-firefox-pgo/firefox-pgo/src/mozilla-1.9.1/ff-pgo python /tmp/yaourt-tmp-root/aur-firefox-pgo/firefox-pgo/src/mozilla-1.9.1/ff-pgo/_profile/pgo/profileserver.py
INFO | (automation.py) | Application pid: 13367
Error: no display specified
TEST-UNEXPECTED-FAIL | (automation.py) | Exited with code 1 during test run
INFO | (automation.py) | Application ran for: 0:00:00.346735
make: *** [profiledbuild] Error 1
Not very similar - no display specified means you ave specified wrong export to other display. Try compiling it the way it's written in pkgbuild.

I disagree. I used the firefox-pgo PKGBUILD to build it, and my DISPLAY variable is set correctly. Keep in mind this is a chroot though, although it shoudln't matter.

blasse · 2009-07-19 21:51:55

It is something different - error we are all dealing on x86-64 has -11 code. This means firefox cannot be started. Your's somehow different Check profile.py file, maybe you'll find what's causing it.

Evanlec · 2009-07-24 21:27:24

Just tried to build latest 3.5.1 last night and got this:

/usr/bin/ld: /usr/lib/libjpeg.a(jcmainct.o): relocation R_X86_64_32S against `a local symbol' can not be used when making a shared object; recompile with -fPIC
/usr/lib/libjpeg.a: could not read symbols: Bad value

blasse · 2009-07-24 23:53:06

Try with libjpeg6 from aur.

Evanlec · 2009-08-01 06:05:45

blasse wrote:

Try with libjpeg6 from aur.

Okay will try that,

Is it okay to install aur/libjpeg6 alongside libjpeg (ver 7) ? or should I remove libjpeg 7 first

blasse · 2009-08-01 07:31:05

It should work just fine alongside libjpeg7. Main libjpeg.so will still be the 7.

Evanlec · 2009-08-01 22:28:36

Tried to build last night after installing libjpeg6 and got the same error.
Perhaps I have to point it specifically to libjpeg6 ?

I dont know...

blasse · 2009-08-02 17:50:46

You may try pointing it in mozconfig or build it without system libjpeg

b4283 · 2009-08-08 10:49:17

Got great news for us:
the sunspider test on Namoroka 3.6alpha1 seems that tracemonkey is enabled!

RESULTS (means and 95% confidence intervals)
--------------------------------------------
Total: 1434.4ms +/- 3.0%
--------------------------------------------

3d: 210.0ms +/- 6.0%
cube: 61.6ms +/- 6.0%
morph: 50.2ms +/- 8.5%
raytrace: 98.2ms +/- 6.7%

access: 216.4ms +/- 7.2%
binary-trees: 58.2ms +/- 8.2%
fannkuch: 93.8ms +/- 7.1%
nbody: 47.2ms +/- 11.8%
nsieve: 17.2ms +/- 11.9%

bitops: 51.6ms +/- 9.7%
3bit-bits-in-byte: 2.0ms +/- 44.0%
bits-in-byte: 11.0ms +/- 13.8%
bitwise-and: 2.2ms +/- 25.3%
nsieve-bits: 36.4ms +/- 10.1%

controlflow: 57.0ms +/- 1.5%
recursive: 57.0ms +/- 1.5%

crypto: 73.4ms +/- 3.7%
aes: 42.6ms +/- 6.1%
md5: 19.4ms +/- 9.7%
sha1: 11.4ms +/- 6.0%

date: 203.4ms +/- 2.8%
format-tofte: 107.2ms +/- 4.5%
format-xparb: 96.2ms +/- 3.1%

math: 54.8ms +/- 4.4%
cordic: 21.0ms +/- 13.2%
partial-sums: 23.2ms +/- 2.4%
spectral-norm: 10.6ms +/- 6.4%

regexp: 105.2ms +/- 12.2%
dna: 105.2ms +/- 12.2%

string: 462.6ms +/- 3.3%
base64: 24.8ms +/- 5.5%
fasta: 110.0ms +/- 3.9%
tagcloud: 133.2ms +/- 3.7%
unpack-code: 144.4ms +/- 9.8%
validate-input: 50.2ms +/- 11.4%

blasse · 2009-08-08 10:56:01

3.6a1 is ready to download on AUR

blasse · 2009-08-08 11:10:12

For me it's faster, but still not so fast as i686.
@b4283
What flags have you used while compiling? Mine is with march=native and -Os

EDIT

https://bugzilla.mozilla.org/show_bug.cgi?id=457879
TM is not enable. 3.6a1 is just faster

============================================
RESULTS (means and 95% confidence intervals)
--------------------------------------------
Total: 2568.2ms +/- 1.3%
--------------------------------------------

3d: 316.8ms +/- 1.2%
cube: 107.0ms +/- 2.7%
morph: 109.2ms +/- 1.2%
raytrace: 100.6ms +/- 1.1%

access: 387.4ms +/- 2.1%
binary-trees: 45.4ms +/- 1.5%
fannkuch: 184.4ms +/- 4.2%
nbody: 110.0ms +/- 1.1%
nsieve: 47.6ms +/- 2.3%

bitops: 259.6ms +/- 1.0%
3bit-bits-in-byte: 44.4ms +/- 5.8%
bits-in-byte: 67.6ms +/- 1.0%
bitwise-and: 70.2ms +/- 1.5%
nsieve-bits: 77.4ms +/- 0.9%

controlflow: 40.0ms +/- 0.0%
recursive: 40.0ms +/- 0.0%

crypto: 150.2ms +/- 1.7%
aes: 60.8ms +/- 0.9%
md5: 43.2ms +/- 4.7%
sha1: 46.2ms +/- 2.9%

date: 220.8ms +/- 0.3%
format-tofte: 101.8ms +/- 1.3%
format-xparb: 119.0ms +/- 1.0%

math: 272.6ms +/- 2.7%
cordic: 119.0ms +/- 1.0%
partial-sums: 103.6ms +/- 7.1%
spectral-norm: 50.0ms +/- 3.5%

regexp: 287.4ms +/- 14.3%
dna: 287.4ms +/- 14.3%

string: 633.4ms +/- 3.7%
base64: 53.4ms +/- 1.3%
fasta: 127.6ms +/- 1.1%
tagcloud: 149.0ms +/- 3.3%
unpack-code: 220.0ms +/- 7.3%
validate-input: 83.4ms +/- 2.7%

Last edited by blasse (2009-08-08 11:31:49)

b4283 · 2009-08-08 12:38:37

blasse wrote:

For me it's faster, but still not so fast as i686.
@b4283
What flags have you used while compiling? Mine is with march=native and -Os

boomer, I've made a mistake
the figures provided above was achieved with official binary version (which is x86 only)

the firefox-pgo-beta package gets me a similar figure about the same as before in the Sunspider test.

Procyon · 2009-08-08 13:57:05

Wow mine built! I haven't been able to since 3.0.7 (and have been using [extra]'s)

And it's fast! It even stays usable while loading huge pages.

Last edited by Procyon (2009-08-08 14:00:32)

blasse · 2009-08-08 17:16:20

I've used tricky workaround for x86-64. I've noticed, that building non-pgo, and than pgo using the same source produces working pgo build. So now makepkg is building firefox for x86-64 twice

Gen2ly · 2009-08-09 01:33:38

lol, blasse, have no idea how you figured that one out. first time I've been able to build and run firefox-3.5+ on x64_86. 3.6 alpha 1 looks to be running good. nice job.

savagenator · 2009-08-09 21:32:00

Thank you everyone! Looks like firefox compiles with pgo support now! My benchmarks jumped up by about double now, and I get about 4489.0ms on the sunspider test (thats still slow )

btw, chrome on linux gets 852.6ms. Jeez.

b4283 · 2009-08-10 04:23:56

Well, there's nothing wrong to run Firefox under 32-bit libs, well, except for additional libraries that will have to be installed. Under multi-lib, Firefox is actually as fast as i686.

The real problem is that for non-native english users like myself, is going to need a Input Method (eg. SCIM or ibus) to input non-english characters, which doesn't support a lib32 environment. So our urgent need is destined to run Firefox under a native 64-bit.

Arch Linux

#251 2009-07-18 10:56:13

Re: Official firefox-pgo thread

#252 2009-07-18 11:00:07

Re: Official firefox-pgo thread

#253 2009-07-18 16:54:50

Re: Official firefox-pgo thread

#254 2009-07-18 18:11:43

Re: Official firefox-pgo thread

#255 2009-07-18 21:00:57

Re: Official firefox-pgo thread

#256 2009-07-18 22:14:18

Re: Official firefox-pgo thread

#257 2009-07-19 02:13:47

Re: Official firefox-pgo thread

#258 2009-07-19 07:50:12

Re: Official firefox-pgo thread

#259 2009-07-19 18:06:04

Re: Official firefox-pgo thread

#260 2009-07-19 21:51:55

Re: Official firefox-pgo thread

#261 2009-07-24 21:27:24

Re: Official firefox-pgo thread

#262 2009-07-24 23:53:06

Re: Official firefox-pgo thread

#263 2009-08-01 06:05:45

Re: Official firefox-pgo thread

#264 2009-08-01 07:31:05

Re: Official firefox-pgo thread

#265 2009-08-01 22:28:36

Re: Official firefox-pgo thread

#266 2009-08-02 17:50:46

Re: Official firefox-pgo thread

#267 2009-08-08 10:49:17

Re: Official firefox-pgo thread

#268 2009-08-08 10:56:01

Re: Official firefox-pgo thread

#269 2009-08-08 11:10:12

Re: Official firefox-pgo thread

#270 2009-08-08 12:38:37

Re: Official firefox-pgo thread

#271 2009-08-08 13:57:05

Re: Official firefox-pgo thread

#272 2009-08-08 17:16:20

Re: Official firefox-pgo thread

#273 2009-08-09 01:33:38

Re: Official firefox-pgo thread

#274 2009-08-09 21:32:00

Re: Official firefox-pgo thread

#275 2009-08-10 04:23:56

Re: Official firefox-pgo thread

Board footer