This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Redundant malloc in structure optimization? (testsuite/gcc.dg/struct/wo_prof_global_var.c)
On Thu, May 1, 2008 at 10:42 PM, Olga Golovanevsky <OLGA@il.ibm.com> wrote:
> Richard Guenther <richard.guenther@gmail.com> wrote on 01/05/2008 16:00:44:
>
>
>
> > On Thu, May 1, 2008 at 2:19 PM, Olga Golovanevsky <OLGA@il.ibm.com>
> wrote:
> > >
> > >
> > > gcc-owner@gcc.gnu.org wrote on 28/04/2008 12:36:44:
> > >
> > >
> > > > Hello,
> > > > I am looking at a testsuite failure (wo_prof_global_var.c) in my
> > > > porting. Somehow, I found GCC 4.3.0 seems to generate unnecessary
> malloc
> > > > during structure optimization. In the code, the structure is split
> into
> > > > two individual fields (D.2240 and D.2242) and they are allocated
> > > > separately. But the original structure (D.2215) is still allocated,
> and
> > > > not used afterward. The following RTL-level optimization cannot
> > > > eliminate it.
> > >
> > > I think that p is global, and in my understanding right now we have no
> > > whole
> > > program dead code elimination optimization in gcc, but may be I am
> wrong.
> >
> > You cannot remove p, but if you could split up the allocation for *p then
> you
> > can for sure remove the original allocation for *p. Otherwise you
> generated
> > wrong code in the first place, which you certainly did, if p was global.
>
> Again, struct-reorg is a whole program optimization. It explicitly checks
> the>
> "whole-program" flag at its gate. It checks the safety of transformation
> through
> ipa-type-escape analysis and other escape analysis inside struct-reorg. If
> it
> decides it is safe to transform, it replace all variables of specific
> structure
> type along with their allocations and accesses. It has obviously enough
> information
> to eliminate both a malloc statement and p, even if p is global. But it's
> question
> whether it's a right place to do it inside struct-reorg. For example,
> function
> level optimizers rely on dce for this matter.
Sure, a malloc call has side-effects, so a DCE pass cannot just remove it.
Only struct-reorg knows that it has replaced all side-effects with others.
Richard.