[Fortran] OpenMP 5 – permit more sharing clauses for SIMD (PR94690)
gcc/fortran/
PR fortran/94690
* openmp.c (resolve_omp_do):
gcc/testsuite/
PR fortran/94690
* gfortran.dg/gomp/openmp-simd-4.f90: New test.
gcc/fortran/openmp.c | 17 +++----
gcc/testsuite/gfortran.dg/gomp/openmp-simd-4.f90 | 65 ++++++++++++++++++++++++
2 files changed, 71 insertions(+), 11 deletions(-)
@@ -5847,26 +5847,21 @@ resolve_omp_do (gfc_code *code)
"at %L", name, &do_code->loc);
if (code->ext.omp_clauses)
for (list = 0; list < OMP_LIST_NUM; list++)
- if (!is_simd
+ if (!is_simd || code->ext.omp_clauses->collapse > 1
? (list != OMP_LIST_PRIVATE && list != OMP_LIST_LASTPRIVATE)
- : code->ext.omp_clauses->collapse > 1
- ? (list != OMP_LIST_LASTPRIVATE)
- : (list != OMP_LIST_LINEAR))
+ : (list != OMP_LIST_PRIVATE && list != OMP_LIST_LASTPRIVATE
+ && list != OMP_LIST_LINEAR))
for (n = code->ext.omp_clauses->lists[list]; n; n = n->next)
if (dovar == n->sym)
{
- if (!is_simd)
+ if (!is_simd || code->ext.omp_clauses->collapse > 1)
gfc_error ("%s iteration variable present on clause "
"other than PRIVATE or LASTPRIVATE at %L",
name, &do_code->loc);
- else if (code->ext.omp_clauses->collapse > 1)
- gfc_error ("%s iteration variable present on clause "
- "other than LASTPRIVATE at %L",
- name, &do_code->loc);
else
gfc_error ("%s iteration variable present on clause "
- "other than LINEAR at %L",
- name, &do_code->loc);
+ "other than PRIVATE, LASTPRIVATE or "
+ "LINEAR at %L", name, &do_code->loc);
break;
}
if (i > 1)
new file mode 100644
@@ -0,0 +1,65 @@
+! { dg-do compile }
+
+integer :: i, j, k
+integer :: x(5), y(2,5)
+
+!$omp parallel do private(i)
+do i = 1, 5
+ x(i) = 42
+end do
+
+!$omp parallel do lastprivate(i)
+do i = 1, 5
+ x(i) = 42
+end do
+
+
+!$omp simd private(i)
+do i = 1, 5
+ x(i) = 42
+end do
+
+!$omp simd linear(i)
+do i = 1, 5
+ x(i) = 42
+end do
+
+!$omp simd lastprivate(i)
+do i = 1, 5
+ x(i) = 42
+end do
+
+
+!$omp simd private(i) lastprivate(j) collapse(2)
+do i = 1, 5
+ do j = 1, 2
+ y(j, i) = 52
+ end do
+end do
+
+!$omp simd lastprivate(i) private(j) collapse(2)
+do i = 1, 5
+ do j = 1, 2
+ y(j, i) = 52
+ end do
+end do
+
+!$omp parallel do firstprivate(i)
+do i = 1, 5 ! { dg-error "PARALLEL DO iteration variable present on clause other than PRIVATE or LASTPRIVATE" }
+ x(i) = 42
+end do
+
+!$omp parallel do simd firstprivate(i)
+do i = 1, 5 ! { dg-error "PARALLEL DO SIMD iteration variable present on clause other than PRIVATE, LASTPRIVATE or LINEAR" }
+ x(i) = 42
+end do
+
+!$omp simd linear(i) collapse(2)
+do i = 1, 5 ! { dg-error "SIMD iteration variable present on clause other than PRIVATE or LASTPRIVATE" }
+ do j = 1, 2
+ y(j, i) = 52
+ end do
+end do
+
+
+end
Another issue found when looking into PR94690 and its regression/additional issues. This patch lifts a restriction of 2.15.1.1. Data-sharing Attribute Rules for Variables Referenced in a Construct. OpenMP 4.5 had: * The loop iteration variable(s) in the associated do-loop(s) of a do, parallel do, taskloop, or distribute construct may be listed in a *private* or *lastprivate* clause. * The loop iteration variable in the associated do-loop of a simd construct with just one associated do-loop may be listed in a *linear* clause with a linear-step that is the increment of the associated loop. * The loop iteration variables in the associated do-loops of a simd construct with multiple associated do-loops may be listed in a *lastprivate* clause. OpenMP 5.0 changed this in 2.19.1.1 to * (unchanged) * ... a simd construct with just one associated for-loop may be listed in a private, lastprivate, or linear ... * simd construct with multiple associated for-loops may be listed in a private or lastprivate clause Namely: private + lastprivate is now allowed in all cases. OK for the trunk? Tobias PS: The remaining issue of the PR (PR94690) is about predetermined 'private' for loop variables to which complicated rules apply. Those have been implemented on the ME side which does not recognize Fortran loops ("while(1)" constructs*); OpenMP-ed loops use "for(..." and those are handled by the ME. For that reason, the FE adds a bunch of 'private' but not always to the right OpenMP construct and not always recognizing that the user already had specified one. (Causing ICEs or reject value issues.) teams/distribute + workshare cause the extra complexity. [* The while(1) loops are used be the FE to match the Fortran semantic regarding overflowing handling for "i + step" exceeding INT_MAX (stop iteration instead of overflowing and continuing).] ----------------- Mentor Graphics (Deutschland) GmbH, Arnulfstraße 201, 80634 München / Germany Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Alexander Walter