View previous topic :: View next topic |
Author |
Message |
PseudoKrazy Tux's lil' helper
Joined: 21 Nov 2003 Posts: 130 Location: USA/NJ
|
Posted: Mon Jul 22, 2013 2:16 am Post subject: Extremely fast compile times with AMD FX 8320 |
|
|
I left Gentoo and linux in general a long time ago so that I could play computer games without the headaches of wine and such. The recent Steam for Linux stuff that valve is doing has gotten me interested again, and so back to Gentoo I come. It was the 'cool distro' to use back when I was heavily into the linux thing, and pretty much the only one that I used, so I figured why not use that. It broke a lot and took forever to install, but hey, I was the cool kid.
And all I can say is, Holy smokes blazing fast compile times!!! I remember setting up X to compile overnight so that I could configure it the next morning! What once took hours now takes minutes! MINUTES! As a matter of fact, anything and everything that once took hours to compile now is a matter of mere minutes! I have -j9 set in my make.conf and am running an AMD FX 8320. This processor alone has taken the one fatal flaw of Gentoo (in my opinion), compile times, and completely made it a moot point. It seems that the hardware has finally caught up to Gentoo's potential.
Just wanted to share my enthusiasm and excitement . What luck! |
|
Back to top |
|
|
eccerr0r Watchman
Joined: 01 Jul 2004 Posts: 9679 Location: almost Mile High in the USA
|
Posted: Mon Jul 22, 2013 3:33 pm Post subject: |
|
|
I don't think it's a fault per se, but it's a penalty for customization.
Nevertheless yes these quad core machines as well as distcc have made builds very fast. And SSDs/portagetmp-on-tmpfs.
My Core2Quad is not bad at all, but yes my i7 with SSD blows it out of the water. Firefox only takes 10 minutes or, even with PGO.
Though I still remember the time my overclocked dual celeron 300 (2x450MHz) took less than 2 minutes to compile its kernel in its day. The i7 can just about do it with modern kernels however, but it just doesn't seem as impressive... _________________ Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching? |
|
Back to top |
|
|
LoTeK Apprentice
Joined: 26 Jul 2012 Posts: 270
|
Posted: Mon Jul 22, 2013 6:18 pm Post subject: |
|
|
Quote: | Firefox only takes 10 minutes or, even with PGO. |
seriously? on my FX-4170 it takes an hour.
Quote: | I have -j9 set in my make.conf and am running an AMD FX 8320. |
ah, I knew I should have bought the octacore
I thought the 4170 is better for me because of the higher frequency (I had to decide between FX-4170 and the 8100 series), but I was wrong. When I switch to "performance mode" in my UEFI-BIOS (with mouse ) the system crashes after 30 min compiling...
today the "compile pain" has switched to embedded devices, micro controller etc. _________________ "I want to see gamma rays! I want to hear X-rays! Do you see the absurdity of what I am? I can't even express these things properly because I have to conceptualize complex ideas in this stupid limiting spoken language!" |
|
Back to top |
|
|
eccerr0r Watchman
Joined: 01 Jul 2004 Posts: 9679 Location: almost Mile High in the USA
|
Posted: Mon Jul 22, 2013 7:28 pm Post subject: |
|
|
Hmm... I think I was lying about 10 minutes (system: Intel Core i7 2700K o/c 4.1GHz, *distcc* make -j8) :
1374015346: >>> emerge (43 of 43) www-client/firefox-17.0.7 to /
1374015347: === (43 of 43) Compiling/Merging (www-client/firefox-17.0.7::/usr/portage/www-client/firefox/firefox-17.0.7.ebuild)
1374015948: ::: completed emerge (43 of 43) www-client/firefox-17.0.7 to /
USE: +alsa +dbus +gstreamer +jit +libnotify +minimal +startup-notification -system-sqlite -wifi
The other machine that may or may not have been helping out on build (because this thing has 8 threads, it probably didn't send much over to the Core2 Duo machine to build.)
Some of the other times I built firefox it took 2000 seconds to build so nowhere near 10 mins. I need to make a more systematic benchmark... But there were a few that were around the 600 second mark...
But yes, ugh, I hate compiling for AVR (embedded, not a general purpose computer)/other cross compiling... Mostly because of having to copy over before testing. Usually that's a pain... |
|
Back to top |
|
|
PseudoKrazy Tux's lil' helper
Joined: 21 Nov 2003 Posts: 130 Location: USA/NJ
|
Posted: Tue Jul 23, 2013 10:13 pm Post subject: |
|
|
eccerr0r, how are you doing your benchmark as of now? |
|
Back to top |
|
|
eccerr0r Watchman
Joined: 01 Jul 2004 Posts: 9679 Location: almost Mile High in the USA
|
Posted: Wed Jul 24, 2013 12:47 am Post subject: |
|
|
Those numbers are the timestamps in /var/log/emerge.log ... but it's not that as the problem, it's more of being methodological (i.e. set conditions, making sure no background tasks, disable distcc, etc.) than the actual timing measurement... _________________ Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching? |
|
Back to top |
|
|
jonnevers Veteran
Joined: 02 Jan 2003 Posts: 1594 Location: Gentoo64 land
|
Posted: Wed Jul 24, 2013 1:00 am Post subject: |
|
|
I average around 22 minutes to compile firefox with my 1090t 6-core AMD.
It was about 12 minutes until last October... which i believe is when xulrunner got merged into the main firefox build. effectively doubling the compile time.
evolution was around 25 minutes but I've recently dumped gnome in favor of dwm ( which is about 8 seconds ).
kernel compile is typically 1 minute and 30 seconds wall clock time |
|
Back to top |
|
|
eccerr0r Watchman
Joined: 01 Jul 2004 Posts: 9679 Location: almost Mile High in the USA
|
Posted: Wed Jul 24, 2013 2:10 am Post subject: |
|
|
Hmm... My last firefox-17.0.7 build on my core2quad (9550S, slightly overclocked, SW RAID5 HDD, 4GB RAM) took 7785-6548=1237 seconds = about 21 minutes.
I do have to admit that my core2quad and i7 have been stealing each other's cpu cycles due to a mis-configured distcc... so who knows which is correct :D _________________ Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching? |
|
Back to top |
|
|
djdunn l33t
Joined: 26 Dec 2004 Posts: 810
|
Posted: Wed Jul 24, 2013 4:39 am Post subject: |
|
|
compiling the kernel is a terrible benchmark :p _________________ “Music is a moral law. It gives a soul to the Universe, wings to the mind, flight to the imagination, a charm to sadness, gaiety and life to everything. It is the essence of order, and leads to all that is good and just and beautiful.”
― Plato |
|
Back to top |
|
|
eccerr0r Watchman
Joined: 01 Jul 2004 Posts: 9679 Location: almost Mile High in the USA
|
Posted: Wed Jul 24, 2013 12:59 pm Post subject: |
|
|
djdunn wrote: | compiling the kernel is a terrible benchmark :p |
Now compiling Firefox... with its gut wrenching link phase... _________________ Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching? |
|
Back to top |
|
|
nlsa8z6zoz7lyih3ap Guru
Joined: 25 Sep 2007 Posts: 388 Location: Canada
|
Posted: Fri Jul 26, 2013 5:14 pm Post subject: |
|
|
Quote: | I have -j9 set in my make.conf and am running an AMD FX 8320 |
I also have an 8 core amd chip (FX 8350)
I include the following in my /etc/make.conf, and am extremely pleased with the "emerge" times:
Code: | MAKEOPTS="-j9 -l16"
EMERGE_DEFAULT_OPTS="--jobs=8 --load-average=16 --keep-going=y --with-bdeps=y --complete-graph" |
|
|
Back to top |
|
|
vaxbrat l33t
Joined: 05 Oct 2005 Posts: 731 Location: DC Burbs
|
Posted: Fri Jul 26, 2013 6:20 pm Post subject: my last two builds have been FX 8350 piledrivers |
|
|
You guys might want to set -march=native in your flags since gcc now knows about family 15 and the new avx instructions. It breaks my heart that I can't go into the local Microcenter and get ECC memory off of the shelf just because Intel forces people into buying price gouging Xeons and server mobo's if they want that. |
|
Back to top |
|
|
nlsa8z6zoz7lyih3ap Guru
Joined: 25 Sep 2007 Posts: 388 Location: Canada
|
Posted: Fri Jul 26, 2013 6:31 pm Post subject: |
|
|
Quote: | You guys might want to set -march=native in your flags since gcc now knows about family 15 |
For my FX-8350 this requires gcc-4.7.x (Which 'tho in ~amd64) runs flawlessly on my machine.
I also use (with great success) the kernel patch from https://github.com/graysky2/kernel_gcc_patch
which supports these CPU's in the kernel configuration. |
|
Back to top |
|
|
PseudoKrazy Tux's lil' helper
Joined: 21 Nov 2003 Posts: 130 Location: USA/NJ
|
Posted: Fri Jul 26, 2013 11:19 pm Post subject: |
|
|
what does the -l16 do in makeopts? |
|
Back to top |
|
|
nlsa8z6zoz7lyih3ap Guru
Joined: 25 Sep 2007 Posts: 388 Location: Canada
|
Posted: Sat Jul 27, 2013 12:12 am Post subject: |
|
|
Quote: | what does the -l16 do in makeopts |
From http://preney.ca/paul/archives/341 I quote:
Quote: | the -l prevents any new parallel job starting unless the load is below the amount specified. |
|
|
Back to top |
|
|
PseudoKrazy Tux's lil' helper
Joined: 21 Nov 2003 Posts: 130 Location: USA/NJ
|
Posted: Sat Jul 27, 2013 12:38 am Post subject: |
|
|
Neato. I'll give it a try, but I haven't noticed any input lag without -l set. Is it there purely to prevent a diminishing returns scenario, where it becomes worthless to start new jobs without first finishing the ones currently going? |
|
Back to top |
|
|
eccerr0r Watchman
Joined: 01 Jul 2004 Posts: 9679 Location: almost Mile High in the USA
|
Posted: Sat Jul 27, 2013 12:53 am Post subject: |
|
|
You probably don't need -l set unless you're trying to run more stuff on the machine (say, you're running some multithreaded number crunching app or another emerge or something), just to make sure your load average doesn't go through the roof. _________________ Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching? |
|
Back to top |
|
|
|