[llvm-bugs] [Bug 45854] New: code moved out of omp critical region
via llvm-bugs
llvm-bugs at lists.llvm.org
Fri May 8 16:21:34 PDT 2020
https://bugs.llvm.org/show_bug.cgi?id=45854
Bug ID: 45854
Summary: code moved out of omp critical region
Product: OpenMP
Version: unspecified
Hardware: PC
OS: Linux
Status: NEW
Severity: normal
Priority: P
Component: Clang Compiler Support
Assignee: unassignedclangbugs at nondot.org
Reporter: michael.p.rice at intel.com
CC: llvm-bugs at lists.llvm.org
After the recent changes to capture fewer globals this test started failing
possibly due to an optimization firing that wasn't before:
extern void omp_set_num_threads (int);
extern void omp_set_dynamic (int);
extern int omp_get_max_threads (void);
int errors = 0;
int thds;
int shrd;
int main ()
{
int i;
omp_set_num_threads(2);
thds = omp_get_max_threads();
omp_set_dynamic(0);
shrd = 0;
#pragma omp parallel for default(shared) schedule(static)
for (i=0; i<thds; i++) {
#pragma omp critical
shrd += 1;
}
if (shrd != thds) errors += 1;
return errors;
}
It appears an optimization pass has moved some of the increment of 'shrd'
outside the critical region.
# %bb.2: # %omp.inner.for.body.preheader
cmpl %eax, %ebp
cmovll %eax, %ebp
movl .Lshrd$local(%rip), %ebx
addl $1, %ebx
subl %eax, %ebp
addl $1, %ebp
.p2align 4, 0x90
.LBB1_3: # %omp.inner.for.body
# =>This Inner Loop Header: Depth=1
movl $.L__unnamed_1, %edi
movl $.gomp_critical_user_.var, %edx
movl %r14d, %esi
callq __kmpc_critical
movl %ebx, .Lshrd$local(%rip)
movl $.L__unnamed_1, %edi
movl $.gomp_critical_user_.var, %edx
movl %r14d, %esi
callq __kmpc_end_critical
addl $1, %ebx
addl $-1, %ebp
jne .LBB1_3
.LBB1_4: # %omp.loop.exit
Also here:
https://godbolt.org/z/2YPJ4P
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20200508/163e6232/attachment.html>
More information about the llvm-bugs
mailing list