This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH, PR 43846] Less strict scalarization criteria in SRA
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: GCC Patches <gcc-patches at gcc dot gnu dot org>, Richard Guenther <rguenther at suse dot de>
- Date: Sun, 11 Jul 2010 18:08:45 -0700
- Subject: Re: [PATCH, PR 43846] Less strict scalarization criteria in SRA
- References: <20100423120919.GB21767@virgil.suse.cz>
On Fri, Apr 23, 2010 at 5:09 AM, Martin Jambor <email@example.com> wrote:
> PR 43846 shows that we should really create scalar replacements also
> for bits of aggregates that were written to and read in a RHS of an
> assignment as a part of an aggregate (currently we demand that the
> piece is read individually, among other possibilities).
> I have bootstrapped and tested the following on x86_64-linux without
> any problems. ?OK for trunk and perhaps after some time for the 4.5
> branch too?
> 2010-04-22 ?Martin Jambor ?<firstname.lastname@example.org>
> ? ? ? ?PR tree-optimization/43846
> ? ? ? ?* tree-sra.c (struct access): New flag grp_assignment_read.
> ? ? ? ?(build_accesses_from_assign): Set grp_assignment_read.
> ? ? ? ?(sort_and_splice_var_accesses): Propagate grp_assignment_read.
> ? ? ? ?(enum mark_read_status): New type.
> ? ? ? ?(analyze_access_subtree): Propagate grp_assignment_read, create
> ? ? ? ?accesses also if both direct_read and root->grp_assignment_read.
> ? ? ? ?* testsuite/gcc.dg/tree-ssa/sra-10.c: New test.
This patch caused: