optimizer problem on x64-64 3.2.2?

optimizer problem on x64-64 3.2.2?

[Randy Howard]

Using SuSE SLES8 (AMD64) 2.4.19-SMP, gcc 3.2.2, running on a dual opteron 
system.

Running the unixbench benchmark suite with -O works fine and gives
reasonable results.  However, if I change the Makefile and recompile
with -O2, all the tests except for "execl" complete ok.  With the 
optimizer turned up it fails on the first iteration and aborts.

The same test on SLES8 (IA32) w/gcc 3.2 does not have this problem.

-- 
Randy Howard            _o     
2reply remove FOOBAR    \<,    
______________________()/ ()______________________________________________
SCO Spam-magnet: postmaster@sco.com

Re: optimizer problem on x64-64 3.2.2?

[Andreas Jaeger]

[-- Attachment #1: Type: text/plain, Size: 937 bytes --]

Randy Howard <randy.howard@FOOmegapathdslBAR.net> writes:

> Using SuSE SLES8 (AMD64) 2.4.19-SMP, gcc 3.2.2, running on a dual opteron 
> system.

SLES8 comes with an additional gcc-33 compiler which you might want to
use for benchmarking, it gives superior results.

> Running the unixbench benchmark suite with -O works fine and gives
> reasonable results.  However, if I change the Makefile and recompile
> with -O2, all the tests except for "execl" complete ok.  With the 
> optimizer turned up it fails on the first iteration and aborts.

AFAIK there was a bug in execl, it's not 64-bit clean.  But I don't
remember details anymore :-(

> The same test on SLES8 (IA32) w/gcc 3.2 does not have this problem.

Andreas
-- 
 Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
  SuSE Linux AG, Deutschherrnstr. 15-19, 90429 Nürnberg, Germany
   GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]