This is the mail archive of the
mailing list for the GCC project.
Re: gcc compile-time performance
- From: law at redhat dot com
- To: "David S. Miller" <davem at redhat dot com>
- Cc: espie at quatramaran dot ens dot fr, gcc at gcc dot gnu dot org
- Date: Sun, 19 May 2002 11:05:52 -0600
- Subject: Re: gcc compile-time performance
- Reply-to: law at redhat dot com
In message <email@example.com>, "David S. Miller" write
> From: Marc Espie <firstname.lastname@example.org>
> Date: Sat, 18 May 2002 13:46:43 +0200
> In case you have missed this, I pitted gcc 2.95.3 -O2 against gcc 3.1 -O1
> to see which one was fastest. On the same source. Something big enough to
> be relevant: a BSD kernel.
> gcc 2.95.3 wins hands down, by about 20%.
> Hey Marc, could you run your tests with this patch applied
> to gcc-3.1? It should kill ~10% or more of the compile time
> overhead when optimizations are enabled.
> 2002-05-17 David S. Miller <email@example.com>
> * cselib.c (cselib_invalidate_regno): Only walk over the
> registers we actually need to look at.
Just for reference, a quick test of this on my PAs is showing 1-2% improvements
(I only tried a few files).