This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] haifa-sched.c and optabs.c latent bugs
- From: law at redhat dot com
- To: David Edelsohn <dje at watson dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 28 May 2002 17:19:37 -0600
- Subject: Re: [PATCH] haifa-sched.c and optabs.c latent bugs
- Reply-to: law at redhat dot com
In message <200204282033.QAA28310@makai.watson.ibm.com>, David Edelsohn writes:
> As I debug the scheduler stack usage bug, I have found a few
> latent bugs which should be fixed.
[ .. SCHED_GROUP_ stuff omitted ..]
>
> The optabs bug is a combination of "subtarget" and "expand_binop".
> I performed a code review of optabs.c and I could only find the one case
> addressed by my patch where the various prerequisites can occur. The
> requirements are a pseudo with split lifetimes that is resurrected by a
> bare clobber not associated with a libcall/no_conflict block. This patch
> creates a new pseudo for the multi-word computation, setting the result to
> the previous target instead of computing the result in the target. This
> also should allow a register allocator and schedule extra flexibility when
> arranging these disjoint computations.
>
> This was bootstrapped on powerpc-ibm-aix4.3.3.0.
>
> Thanks, David
>
> * optabs.c (expand_binop): Fix nwords sign warnings.
> Always generate pseudo for add_optab.
Retested (i686-pc-linux-gnu) and installed into the mainline sources.
jeff