https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78808
Bug ID: 78808
Summary: target_clones not applying to openmp functions
Product: gcc
Version: 6.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: other
Assignee: unassigned at gcc dot gnu.org
Reporter: steven at uplinklabs dot net
Target Milestone: ---
Simple test case:
---
__attribute__((target_clones("arch=haswell", "arch=sandybridge", "default")))
static void _saxpy(int n, float a, float * restrict x, float * restrict y)
{
#pragma omp parallel for
for (int i = 0; i < n; ++i)
y[i] = a*x[i] + y[i];
}
void saxpy(int n, float a, float * restrict x, float * restrict y)
{
return _saxpy(n, a, x, y);
}
---
Compile with:
gcc-6 -O3 -std=gnu11 -ffast-math -S
If -fopenmp is specified, the OpenMP-generated function for the parallelized
for() loop is generated with the default optimizations (not respecting
target_clones).
See attachments for comparison.
