Skip to content

Commit ab2861d

Browse files
Found the issue. Memory locality problem for intel GPU when trying to separate my for loops. Definately not desired behavior for ACC, but simple enough to work around.
1 parent 9f72081 commit ab2861d

File tree

1 file changed

+11
-1
lines changed

1 file changed

+11
-1
lines changed

src/simulation/m_time_steppers.fpp

Lines changed: 11 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -479,7 +479,17 @@ contains
479479
pb_ts(1)%sf(j, k, l, q, i) = &
480480
pb_ts(1)%sf(j, k, l, q, i) &
481481
+ dt*rhs_pb(j, k, l, q, i)
482-
482+
end do
483+
end do
484+
end do
485+
end do
486+
end do
487+
488+
!$acc parallel loop collapse(5) gang vector default(present)
489+
do i = 1, nb
490+
do l = 0, p
491+
do k = 0, n
492+
do j = 0, m
483493
mv_ts(1)%sf(j, k, l, q, i) = &
484494
mv_ts(1)%sf(j, k, l, q, i) &
485495
+ dt*rhs_mv(j, k, l, q, i)

0 commit comments

Comments
 (0)