This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug middle-end/66702] New: #pragma omp declare simd uniform and linear issues
- 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: Tue, 30 Jun 2015 09:43:00 +0000
- Subject: [Bug middle-end/66702] New: #pragma omp declare simd uniform and linear issues
- Auto-submitted: auto-generated
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66702
Bug ID: 66702
Summary: #pragma omp declare simd uniform and linear issues
Product: gcc
Version: 5.1.1
Status: UNCONFIRMED
Keywords: wrong-code
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: jakub at gcc dot gnu.org
Target Milestone: ---
void
bar (int &a, int &b, int *&c, int &d)
{
volatile int x;
int *volatile y;
x = a; a = x;
x = b; b = x;
y = c; c = y;
x = d; d = x;
}
void (*volatile barp) (int &, int &, int *&, int &) = bar;
#pragma omp declare simd uniform(b, c) linear(d:2) aligned(c:32) notinbranch
int
foo (int a, int b, int *c, int d)
{
a++;
b++;
c += 8;
d += 2;
barp (a, b, c, d);
return a + b + *c + d;
}
volatile int e = 5;
int c[64] __attribute__((aligned (32)));
int
main ()
{
int d = 7, r = 0;
int b = e;
for (int i = 0; i < 64; i++)
c[i] = i;
#pragma omp simd reduction(+:r) linear(d:2)
for (int i = 0; i < 64; i++)
{
r += foo (i, b, c, d);
d += 2;
}
if (r != 7584)
__builtin_abort ();
}
is miscompiled with -O2 -fopenmp, because uniform and linear clauses were
properly handled only for non-addressable variables of gimple type.
For addressable vars (both linear and uniform) or for non-gimple types or C++
args passed by invisible reference (uniform only), if the arguments are
modified in the function, we need to make a copy of the initial var and restore
it for each VL.