Source-based experiment

Barrucadu · 2009-02-18 23:19:02

So, for no apparent reason other than "for the lulz", I've decided to reinstall Arch on my desktop computer. And compile everything from source.

This could be fun…

Ashren · 2009-02-18 23:22:53

Do some benchmarks. Measure how long it takes. Are you using special cflags?

Share your lolz.

Barrucadu · 2009-02-18 23:26:45

I will be using special cflags. 10 minutes 'til the new ISO is downloaded. Even though I'll just be recompiling everything anyway, I just want a look at the new ISO

Last edited by Barrucadu (2009-02-18 23:27:13)

Renan Birck · 2009-02-19 03:36:18

This is what I've been curious to see... speed comparations between standard packages and compiled ones.

Ranguvar · 2009-02-19 03:58:06

1.) Please benchmark for others!
2.) Use LDFLAGS! They help more than C(XX)FLAGS in my experience. Though quite a few packages don't respect them... I know in Gentoo they have patches to many packages to make them respect LDFLAGS.
http://en.gentoo-wiki.com/wiki/Safe_LDFLAGS
http://forums.gentoo.org/viewtopic-t-67777.html
http://forums.gentoo.org/viewtopic-t-226909.html
http://forums.gentoo.org/viewtopic-t-316445.html (note: that list is way deprecated)
http://lwn.net/Articles/192624/
http://www.gentoo-wiki.info/CFLAGS_Matrix

I used "-Wl,-O1 -Wl,--sort-common -Wl,--as-needed -Wl,--hash-style=gnu" to much success (app load times, not in-app performance, is what is helped) in Funtoo/Gentoo. Arch uses "--hash-style=both" by default Allan said once IIRC, which (I may easily be wrong) can provide the same performance as gnu, but increased binary size.

Allan · 2009-02-19 08:48:41

Note that our gcc has -Wl,--hash-style=both set

Arch Linux

#1 2009-02-18 23:19:02

Source-based experiment

#2 2009-02-18 23:22:53

Re: Source-based experiment

#3 2009-02-18 23:26:45

Re: Source-based experiment

#4 2009-02-19 03:36:18

Re: Source-based experiment

#5 2009-02-19 03:58:06

Re: Source-based experiment

#6 2009-02-19 08:48:41

Re: Source-based experiment

Board footer