This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: PowerPC Code generation (probably altivec related)
- From: Daniel Egger <degger at fhm dot edu>
- To: Aldy Hernandez <aldyh at redhat dot com>
- Cc: GCC Developer Mailinglist <gcc at gcc dot gnu dot org>
- Date: 17 Mar 2002 14:35:58 +0100
- Subject: Re: PowerPC Code generation (probably altivec related)
- References: <1E518CB4-3957-11D6-B8DB-000393750C1E@redhat.com>
Am Son, 2002-03-17 um 04.28 schrieb Aldy Hernandez:
> the code generated is correct. r4 and r3 both contain 0, so it's
> fair to use either one as a base register.
It is correct in this case though unnecessary. Why load a value
twice?
> the assembly looks correct. what is the problem?
My problem is that in a bigger application second used value
is incorrect and thus the application loads wrong data which
it processes and thus writes incorrect data back to the correct
place.
I merely tried to point out a potential problem and suboptimal
generated code. Fixing the latter will automatically fix the
problem, at least I believe so. For now I can fix it by sprinkling
casts and/or statifying the variables which will be less
efficient though.
--
Servus,
Daniel