This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/66142] Loop is not vectorized because not sufficient support for GOMP_SIMD_LANE
- From: "jakub at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Fri, 22 May 2015 11:12:40 +0000
- Subject: [Bug tree-optimization/66142] Loop is not vectorized because not sufficient support for GOMP_SIMD_LANE
- Auto-submitted: auto-generated
- References: <bug-66142-4 at http dot gcc dot gnu dot org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66142
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Ah, but fre2 is able to optimize those:
D.2881[_12].dir.x = x_13;
...
_32 = MEM[(const struct Ray *)&D.2881][_12].dir.x;
cases to
_32 = x_13;
The problem in FRE2 is only with the MEM_REFs.
If in the testcase
- ray.org = p;
+ ray.org.x = p.x;
+ ray.org.y = p.y;
then FRE2 handles it completely and the problem is the same between -fopenmp
and -fno-openmp - ifcvt not ifconverting the loop.