This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: 3.0.1 performance
- To: gcc at gcc dot gnu dot org
- Subject: Re: 3.0.1 performance
- From: Kurt Garloff <kurt at garloff dot de>
- Date: Fri, 17 Aug 2001 13:16:58 +0200
- Organization: TU/e(NL), SuSE(DE)
- References: <20010817111644.D10580@gum01m.etpnet.phys.tue.nl>
On Fri, Aug 17, 2001 at 11:16:44AM +0200, Kurt Garloff wrote:
> I did not yet play with parameters, like using plain -O3 (is it useable in
> 3.0.1?) or -finline-limit yet. Should I?
I did. It helped. So, IMHO, the default is too low.
inline Compile Executable Performance
limit time (u) (stripped) (TBCI SPEC)
2.95.3 3:35.44 88364 0.869
3.0.0 3:48.11 89488 0.855
3.0.1 2:49.84 95620 0.320
3.0.1 200 1:50.29 82212 0.297
3.0.1 500 2:39.00 93868 0.318
3.0.1 1000 3:05.27 92524 0.805
3.0.1 2000 3:32.33 91572 0.826
3.0.1 5000 3:46.81 90052 0.834
3.0.1 10000 3:53.47 89180 0.861
3.0.1 50000 3:50.31 89180 0.860
Docu says default for inline-limit is 10000, but based on the above
numbers, I tend to not believe this. It's more like 600.
And for my benchmark, it should be more like 2000.
Regards,
--
Kurt Garloff <kurt@garloff.de> [Eindhoven, NL]
Physics: Plasma simulations <K.Garloff@Phys.TUE.NL> [TU Eindhoven, NL]
Linux: SCSI, Security <garloff@suse.de> [SuSE Nuernberg, DE]
(See mail header or public key servers for PGP2 and GPG public keys.)
PGP signature