This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Feedback optimization errors running SPEC with GCC 3.4 on Darwin 6.4 (Fwd)
- From: "Mostafa Hagog" <MUSTAFA at il dot ibm dot com>
- To: Jan Hubicka <jh at suse dot cz>
- Cc: "Ayal Zaks" <ZAKS at il dot ibm dot com>, David Edelsohn <dje at watson dot ibm dot com>, gcc-patches at gcc dot gnu dot org, Jan Hubicka <jh at suse dot cz>
- Date: Tue, 30 Sep 2003 12:21:39 +0300
- Subject: Re: Feedback optimization errors running SPEC with GCC 3.4 on Darwin 6.4 (Fwd)
Are iam.gcda and iam.gcno (on darwin6.4), could be of any help to you?
Jan Hubicka
<jh@suse.cz> To: Jan Hubicka <jh@suse.cz>
cc: David Edelsohn <dje@watson.ibm.com>, gcc-patches@gcc.gnu.org, Mostafa
30/09/03 12:05 Hagog/Haifa/IBM@IBMIL, Ayal Zaks/Haifa/IBM@IBMIL
Subject: Re: Feedback optimization errors running SPEC with GCC 3.4 on Darwin 6.4 (Fwd)
> > Unfortunately, I still get corrupted profile info on PowerPC.
> > Also, the file that fails has changed! This could be a wild write or
> > uninitialized variable :-(.
>
> I guess it is another bug in Nathanel's i/o rewrite. The new error
> message is the new sanity check I added. It means that we are reading
> complete nonsence. The i/o code seems to be very hard optimized making
> it very dificult to figure out what it is doing.
> Can I have dump/da files for the new file?
Note that this time I need only 16.bp dumps, but I need also the .gc*
files. The problem is obvously somewhere low level (either file
handling on summary production), so no wonder that the profiling didn't
used to help much for PPC...
I am trying to reproduce similar problem on mips now. It can be related
to endianity.
Honza
>
> Thanks!
> Honza
> >
> > analyze-linkage.c: In function `advance_DIS':
> >
> > analyze-linkage.c:248: error: corrupted profile info: edge from -1 to 0
> > exceeds maximal count
> > analyze-linkage.c:248: error: corrupted profile info: edge from 3 to 6
> > exceeds maximal count
> > analyze-linkage.c:248: error: corrupted profile info: edge from 4 to 5
> > exceeds maximal count
> > analyze-linkage.c:248: error: corrupted profile info: edge from 5 to 8
> > exceeds maximal count
> > analyze-linkage.c:248: error: corrupted profile info: edge from 7 to 8
> > exceeds maximal count
> > specmake: *** [analyze-linkage.o] Error 1