This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-ssa] PPC bootstraps still broken
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Jeff Law <law at redhat dot com>
- Cc: "gcc at gcc dot gnu dot org" <gcc at gcc dot gnu dot org>
- Date: 08 Jul 2003 10:06:23 -0400
- Subject: Re: [tree-ssa] PPC bootstraps still broken
- Organization: Red Hat Canada
- References: <200307081343.h68DhntC008691@speedy.slc.redhat.com>
On Tue, 2003-07-08 at 09:43, law@redhat.com wrote:
> ?!? If the temporary names are differing, then that's a sign of
> instability somewhere. I can't offhand think of a reason why the
> temporary names should be allowed to differ.
>
Er, correct. I don't know what I was thinking. The tree dumps should
be the same.
Hmm, then the instability is present in other ports. It seems as if the
genericizer creates temporaries in different order in stage1 and stage2.
For instance, on x86 compiling gcc.o with stage1/cc1 and stage2/cc1
produces differences like this for the t03.generic dump (left column is
stage1, right column is stage2).
double T.659; | unsigned int j.659;
unsigned int j.660; | unsigned int T.660;
unsigned int T.661; | struct command * T.661;
However, if you compile with -da and -fdump-tree-all then both RTL and
tree dumps are exactly the same. Lovely.
I can't spend much more time on this right now, so I'll add this to the
list of things to fix.
Diego.