[PR middle-end/71373] Handle more OMP_CLAUSE_* in nested function decomposition

	gcc/
	* gimplify.c (gimplify_adjust_omp_clauses): Discard
	OMP_CLAUSE_TILE.
	* omp-low.c (scan_sharing_clauses): Don't expect OMP_CLAUSE_TILE.
	gcc/testsuite/
	* c-c++-common/goacc/combined-directives.c: XFAIL tree scanning
	for OpenACC tile clauses.
	* gfortran.dg/goacc/combined-directives.f90: Likewise.

	gcc/
	PR middle-end/71373
	* tree-nested.c (convert_nonlocal_omp_clauses)
	(convert_local_omp_clauses): Handle OMP_CLAUSE_ASYNC,
	OMP_CLAUSE_WAIT, OMP_CLAUSE_INDEPENDENT, OMP_CLAUSE_AUTO,
	OMP_CLAUSE__CACHE_, OMP_CLAUSE_TILE.
	gcc/testsuite/
	PR middle-end/71373
	* gcc.dg/goacc/nested-function-1.c: New file.
	* gcc.dg/goacc/nested-function-2.c: Likewise.
	* gcc.dg/goacc/pr71373.c: Likewise.
	* gfortran.dg/goacc/cray-2.f95: Likewise.
	* gfortran.dg/goacc/loop-1-2.f95: Likewise.
	* gfortran.dg/goacc/loop-3-2.f95: Likewise.
	* gfortran.dg/goacc/cray.f95: Update.
	* gfortran.dg/goacc/loop-1.f95: Likewise.
	* gfortran.dg/goacc/loop-3.f95: Likewise.
	* gfortran.dg/goacc/subroutines.f90: Update, and rename to...
	* gfortran.dg/goacc/nested-function-1.f90: ... this new file.
	libgomp/testsuite/
	PR middle-end/71373
	* libgomp.oacc-c/nested-function-1.c: New file.
	* libgomp.oacc-c/nested-function-2.c: Likewise.
	* libgomp.oacc-fortran/nested-function-1.f90: Likewise.
	* libgomp.oacc-fortran/nested-function-2.f90: Likewise.
	* libgomp.oacc-fortran/nested-function-3.f90: Likewise.

Co-Authored-By: Cesar Philippidis <cesar@codesourcery.com>

From-SVN: r237291

5 files changed, 694 insertions, 0 deletions

diff --git a/libgomp/testsuite/libgomp.oacc-c/nested-function-1.c b/libgomp/testsuite/libgomp.oacc-c/nested-function-1.c
new file mode 100644
index 0000000..fb2a3ac
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c/nested-function-1.c
@@ -0,0 +1,52 @@
+/* Exercise nested function decomposition, gcc/tree-nested.c.  */
+
+int
+main (void)
+{
+  void test1 ()
+  {
+    int i, j, k;
+    int a[4][7][8];
+
+    __builtin_memset (a, 0, sizeof (a));
+
+#pragma acc parallel
+#pragma acc loop collapse(4 - 1)
+    for (i = 1; i <= 3; i++)
+      for (j = 4; j <= 6; j++)
+	for (k = 5; k <= 7; k++)
+	  a[i][j][k] = i + j + k;
+
+    for (i = 1; i <= 3; i++)
+      for (j = 4; j <= 6; j++)
+	for (k = 5; k <= 7; k++)
+	  if (a[i][j][k] != i + j + k)
+	    __builtin_abort();
+  }
+
+  void test2 ()
+  {
+    int i, j, k;
+    int a[4][4][4];
+
+    __builtin_memset (a, 0, sizeof (a));
+
+#pragma acc parallel
+#pragma acc loop collapse(3)
+    for (i = 1; i <= 3; i++)
+      for (j = 1; j <= 3; j++)
+	for (k = 1; k <= 3; k++)
+	  a[i][j][k] = 1;
+
+    for (i = 1; i <= 3; i++)
+      for (j = 1; j <= 3; j++)
+	for (k = 1; k <= 3; k++)
+	  if (a[i][j][k] != 1)
+	    __builtin_abort ();
+  }
+
+  test1 ();
+  test2 ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-c/nested-function-2.c b/libgomp/testsuite/libgomp.oacc-c/nested-function-2.c
new file mode 100644
index 0000000..2c3f3fe
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-c/nested-function-2.c
@@ -0,0 +1,155 @@
+/* Exercise nested function decomposition, gcc/tree-nested.c.  */
+
+int
+main (void)
+{
+  int p1 = 2, p2 = 6, p3 = 0, p4 = 4, p5 = 13, p6 = 18, p7 = 1, p8 = 1, p9 = 1;
+
+  void test1 ()
+  {
+    int i, j, k;
+    int a[4][4][4];
+
+    __builtin_memset (a, '\0', sizeof (a));
+
+#pragma acc parallel
+#pragma acc loop collapse(3)
+    for (i = 1; i <= 3; i++)
+      for (j = 1; j <= 3; j++)
+	for (k = 2; k <= 3; k++)
+	  a[i][j][k] = 1;
+
+    for (i = 1; i <= 3; i++)
+      for (j = 1; j <= 3; j++)
+	for (k = 2; k <= 3; k++)
+	  if (a[i][j][k] != 1)
+	    __builtin_abort();
+  }
+
+  void test2 (int v1, int v2, int v3, int v4, int v5, int v6)
+  {
+    int i, j, k, l = 0, r = 0;
+    int a[7][5][19];
+    int b[7][5][19];
+
+    __builtin_memset (a, '\0', sizeof (a));
+    __builtin_memset (b, '\0', sizeof (b));
+
+#pragma acc parallel reduction (||:l)
+#pragma acc loop reduction (||:l) collapse(3)
+    for (i = v1; i <= v2; i++)
+      for (j = v3; j <= v4; j++)
+	for (k = v5; k <= v6; k++)
+	  {
+	    l = l || i < 2 || i > 6 || j < 0 || j > 4 || k < 13 || k > 18;
+	    if (!l)
+	      a[i][j][k] += 1;
+	  }
+
+    for (i = v1; i <= v2; i++)
+      for (j = v3; j <= v4; j++)
+	for (k = v5; k <= v6; k++)
+	  {
+	    r = r || i < 2 || i > 6 || j < 0 || j > 4 || k < 13 || k > 18;
+	    if (!r)
+	      b[i][j][k] += 1;
+	  }
+
+    if (l != r)
+      __builtin_abort ();
+
+    for (i = v1; i <= v2; i++)
+      for (j = v3; j <= v4; j++)
+	for (k = v5; k <= v6; k++)
+	  if (b[i][j][k] != a[i][j][k])
+	    __builtin_abort ();
+  }
+
+  void test3 (int v1, int v2, int v3, int v4, int v5, int v6, int v7, int v8,
+      int v9)
+  {
+    int i, j, k, l = 0, r = 0;
+    int a[7][5][19];
+    int b[7][5][19];
+
+    __builtin_memset (a, '\0', sizeof (a));
+    __builtin_memset (b, '\0', sizeof (b));
+
+#pragma acc parallel reduction (||:l)
+#pragma acc loop reduction (||:l) collapse(3)
+    for (i = v1; i <= v2; i += v7)
+      for (j = v3; j <= v4; j += v8)
+	for (k = v5; k <= v6; k += v9)
+	  {
+	    l = l || i < 2 || i > 6 || j < 0 || j > 4 || k < 13 || k > 18;
+	    if (!l)
+	      a[i][j][k] += 1;
+	  }
+
+    for (i = v1; i <= v2; i += v7)
+      for (j = v3; j <= v4; j += v8)
+	for (k = v5; k <= v6; k += v9)
+	  {
+	    r = r || i < 2 || i > 6 || j < 0 || j > 4 || k < 13 || k > 18;
+	    if (!r)
+	      b[i][j][k] += 1;
+	  }
+
+    if (l != r)
+      __builtin_abort ();
+
+    for (i = v1; i <= v2; i++)
+      for (j = v3; j <= v4; j++)
+	for (k = v5; k <= v6; k++)
+	  if (b[i][j][k] != a[i][j][k])
+	    __builtin_abort ();
+  }
+
+  void test4 ()
+  {
+    int i, j, k, l = 0, r = 0;
+    int a[7][5][19];
+    int b[7][5][19];
+    int v1 = p1, v2 = p2, v3 = p3, v4 = p4, v5 = p5, v6 = p6, v7 = p7, v8 = p8,
+      v9 = p9;
+
+    __builtin_memset (a, '\0', sizeof (a));
+    __builtin_memset (b, '\0', sizeof (b));
+
+#pragma acc parallel reduction (||:l)
+#pragma acc loop reduction (||:l) collapse(3)
+    for (i = v1; i <= v2; i += v7)
+      for (j = v3; j <= v4; j += v8)
+	for (k = v5; k <= v6; k += v9)
+	  {
+	    l = l || i < 2 || i > 6 || j < 0 || j > 4 || k < 13 || k > 18;
+	    if (!l)
+	      a[i][j][k] += 1;
+	  }
+
+    for (i = v1; i <= v2; i += v7)
+      for (j = v3; j <= v4; j += v8)
+	for (k = v5; k <= v6; k += v9)
+	  {
+	    r = r || i < 2 || i > 6 || j < 0 || j > 4 || k < 13 || k > 18;
+	    if (!r)
+	      b[i][j][k] += 1;
+	  }
+
+    if (l != r)
+      __builtin_abort ();
+
+    for (i = v1; i <= v2; i++)
+      for (j = v3; j <= v4; j++)
+	for (k = v5; k <= v6; k++)
+	  if (b[i][j][k] != a[i][j][k])
+	    __builtin_abort ();
+  }
+
+  test1 ();
+  test2 (p1, p2, p3, p4, p5, p6);
+  test3 (p1, p2, p3, p4, p5, p6, p7, p8, p9);
+  test4 ();
+
+  return 0;
+}
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/nested-function-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/nested-function-1.f90
new file mode 100644
index 0000000..fdbca44
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/nested-function-1.f90
@@ -0,0 +1,70 @@
+! Exercise nested function decomposition, gcc/tree-nested.c.
+
+! { dg-do run }
+
+program collapse2
+  call test1
+  call test2
+contains
+  subroutine test1
+    integer :: i, j, k, a(1:3, 4:6, 5:7)
+    logical :: l
+    l = .false.
+    a(:, :, :) = 0
+    !$acc parallel reduction (.or.:l)
+    !$acc loop worker vector collapse(4 - 1)
+      do 164 i = 1, 3
+        do 164 j = 4, 6
+          do 164 k = 5, 7
+            a(i, j, k) = i + j + k
+164      end do
+    !$acc loop worker vector reduction(.or.:l) collapse(2)
+firstdo: do i = 1, 3
+        do j = 4, 6
+          do k = 5, 7
+            if (a(i, j, k) .ne. (i + j + k)) l = .true.
+          end do
+        end do
+      end do firstdo
+    !$acc end parallel
+    if (l) call abort
+  end subroutine test1
+
+  subroutine test2
+    integer :: a(3,3,3), k, kk, kkk, l, ll, lll
+    a = 0
+    !$acc parallel
+    ! Use "gang(static:1)" here and below to effectively turn gang-redundant
+    ! execution mode into something like gang-single.
+    !$acc loop gang(static:1) collapse(1)
+      do 115 k=1,3
+         !$acc loop collapse(2)
+  dokk: do kk=1,3
+          do kkk=1,3
+            a(k,kk,kkk) = 1
+          enddo
+        enddo dokk
+115   continue
+    !$acc loop gang(static:1) collapse(1)
+      do k=1,3
+         if (any(a(k,1:3,1:3).ne.1)) call abort
+      enddo
+    ! Use "gang(static:1)" here and below to effectively turn gang-redundant
+    ! execution mode into something like gang-single.
+    !$acc loop gang(static:1) collapse(1)
+ dol: do 120 l=1,3
+    !$acc loop collapse(2)
+  doll: do ll=1,3
+          do lll=1,3
+            a(l,ll,lll) = 2
+          enddo
+        enddo doll
+120   end do dol
+    !$acc loop gang(static:1) collapse(1)
+     do l=1,3
+        if (any(a(l,1:3,1:3).ne.2)) call abort
+     enddo
+    !$acc end parallel
+  end subroutine test2
+
+end program collapse2
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/nested-function-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/nested-function-2.f90
new file mode 100644
index 0000000..4e28196
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/nested-function-2.f90
@@ -0,0 +1,173 @@
+! Exercise nested function decomposition, gcc/tree-nested.c.
+
+! { dg-do run }
+
+program collapse3
+  integer :: p1, p2, p3, p4, p5, p6, p7, p8, p9
+  p1 = 2
+  p2 = 6
+  p3 = -2
+  p4 = 4
+  p5 = 13
+  p6 = 18
+  p7 = 1
+  p8 = 1
+  p9 = 1
+  call test1
+  call test2 (p1, p2, p3, p4, p5, p6)
+  call test3 (p1, p2, p3, p4, p5, p6, p7, p8, p9)
+  call test4
+contains
+  subroutine test1
+    integer :: a(3,3,3), k, kk, kkk, l, ll, lll
+    !$acc parallel
+    !$acc loop collapse(3)
+      do 115 k=1,3
+dokk:   do kk=1,3
+          do kkk=1,3
+            a(k,kk,kkk) = 1
+          enddo
+        enddo dokk
+115   continue
+    !$acc end parallel
+    if (any(a(1:3,1:3,1:3).ne.1)) call abort
+    !$acc parallel
+    !$acc loop collapse(3)
+dol:  do 120 l=1,3
+doll:   do ll=1,3
+          do lll=1,3
+            a(l,ll,lll) = 2
+          enddo
+        enddo doll
+120   end do dol
+    !$acc end parallel
+    if (any(a(1:3,1:3,1:3).ne.2)) call abort
+    end subroutine test1
+
+  subroutine test2(v1, v2, v3, v4, v5, v6)
+    integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+    integer :: v1, v2, v3, v4, v5, v6
+    logical :: l, r
+    l = .false.
+    r = .false.
+    a(:, :, :) = 0
+    b(:, :, :) = 0
+    !$acc parallel reduction (.or.:l)
+    !$acc loop reduction (.or.:l) collapse (3)
+      do i = v1, v2
+        do j = v3, v4
+          do k = v5, v6
+            l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+            l = l.or.k.lt.13.or.k.gt.18
+            if (.not.l) a(i, j, k) = a(i, j, k) + 1
+          end do
+        end do
+      end do
+    !$acc end parallel
+    do i = v1, v2
+      do j = v3, v4
+        do k = v5, v6
+          r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          r = r.or.k.lt.13.or.k.gt.18
+          if (.not.l) b(i, j, k) = b(i, j, k) + 1
+        end do
+      end do
+    end do
+    if (l .neqv. r) call abort
+    do i = v1, v2
+      do j = v3, v4
+        do k = v5, v6
+           if (a(i, j, k) .ne. b(i, j, k)) call abort
+        end do
+      end do
+    end do
+  end subroutine test2
+
+  subroutine test3(v1, v2, v3, v4, v5, v6, v7, v8, v9)
+    integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+    integer :: v1, v2, v3, v4, v5, v6, v7, v8, v9
+    logical :: l, r
+    l = .false.
+    r = .false.
+    a(:, :, :) = 0
+    b(:, :, :) = 0
+    !$acc parallel reduction (.or.:l)
+    !$acc loop reduction (.or.:l) collapse (3)
+      do i = v1, v2, v7
+        do j = v3, v4, v8
+          do k = v5, v6, v9
+            l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+            l = l.or.k.lt.13.or.k.gt.18
+            if (.not.l) a(i, j, k) = a(i, j, k) + 1
+          end do
+        end do
+      end do
+    !$acc end parallel
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+          r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          r = r.or.k.lt.13.or.k.gt.18
+          if (.not.l) b(i, j, k) = b(i, j, k) + 1
+        end do
+      end do
+    end do
+    if (l .neqv. r) call abort
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+           if (a(i, j, k) .ne. b(i, j, k)) call abort
+        end do
+      end do
+    end do
+  end subroutine test3
+
+  subroutine test4
+    integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+    integer :: v1, v2, v3, v4, v5, v6, v7, v8, v9
+    logical :: l, r
+    l = .false.
+    r = .false.
+    a(:, :, :) = 0
+    b(:, :, :) = 0
+    v1 = p1
+    v2 = p2
+    v3 = p3
+    v4 = p4
+    v5 = p5
+    v6 = p6
+    v7 = p7
+    v8 = p8
+    v9 = p9
+    !$acc parallel reduction (.or.:l)
+    !$acc loop reduction (.or.:l) collapse (3)
+      do i = v1, v2, v7
+        do j = v3, v4, v8
+          do k = v5, v6, v9
+            l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+            l = l.or.k.lt.13.or.k.gt.18
+            if (.not.l) a(i, j, k) = a(i, j, k) + 1
+          end do
+        end do
+      end do
+    !$acc end parallel
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+          r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          r = r.or.k.lt.13.or.k.gt.18
+          if (.not.r) b(i, j, k) = b(i, j, k) + 1
+        end do
+      end do
+    end do
+    if (l .neqv. r) call abort
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+         do k = v5, v6, v9
+           if (a(i, j, k) .ne. b(i, j, k)) call abort
+         end do
+      end do
+    end do
+  end subroutine test4
+
+end program collapse3
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/nested-function-3.f90 b/libgomp/testsuite/libgomp.oacc-fortran/nested-function-3.f90
new file mode 100644
index 0000000..2f6485e
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/nested-function-3.f90
@@ -0,0 +1,244 @@
+! Exercise nested function decomposition, gcc/tree-nested.c.
+
+! { dg-do run }
+
+program sub_collapse_3
+  call test1
+  call test2 (2, 6, -2, 4, 13, 18)
+  call test3 (2, 6, -2, 4, 13, 18, 1, 1, 1)
+  call test4
+  call test5 (2, 6, -2, 4, 13, 18)
+  call test6 (2, 6, -2, 4, 13, 18, 1, 1, 1)
+contains
+  subroutine test1
+    integer :: a(3,3,3), k, kk, kkk, l, ll, lll
+    !$acc parallel
+    !$acc loop collapse(3)
+      do 115 k=1,3
+dokk:   do kk=1,3
+          do kkk=1,3
+            a(k,kk,kkk) = 1
+          enddo
+        enddo dokk
+115   continue
+    !$acc end parallel
+    if (any(a(1:3,1:3,1:3).ne.1)) call abort
+    !$acc parallel
+    !$acc loop collapse(3)
+dol:  do 120 l=1,3
+doll:   do ll=1,3
+          do lll=1,3
+            a(l,ll,lll) = 2
+          enddo
+        enddo doll
+120   end do dol
+    !$acc end parallel
+    if (any(a(1:3,1:3,1:3).ne.2)) call abort
+  end subroutine test1
+
+  subroutine test2(v1, v2, v3, v4, v5, v6)
+    integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+    integer :: v1, v2, v3, v4, v5, v6
+    logical :: l, r
+    l = .false.
+    r = .false.
+    a(:, :, :) = 0
+    b(:, :, :) = 0
+    !$acc parallel pcopyin (v1, v2, v3, v4, v5, v6) reduction (.or.:l)
+    !$acc loop reduction (.or.:l) collapse (3)
+      do i = v1, v2
+        do j = v3, v4
+          do k = v5, v6
+            l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+            l = l.or.k.lt.13.or.k.gt.18
+            if (.not.l) a(i, j, k) = a(i, j, k) + 1
+          end do
+        end do
+      end do
+    !$acc end parallel
+    do i = v1, v2
+      do j = v3, v4
+        do k = v5, v6
+          r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          r = r.or.k.lt.13.or.k.gt.18
+          if (.not.l) b(i, j, k) = b(i, j, k) + 1
+        end do
+      end do
+    end do
+    if (l .neqv. r) call abort
+    do i = v1, v2
+      do j = v3, v4
+        do k = v5, v6
+           if (a(i, j, k) .ne. b(i, j, k)) call abort
+        end do
+      end do
+    end do
+  end subroutine test2
+
+  subroutine test3(v1, v2, v3, v4, v5, v6, v7, v8, v9)
+    integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+    integer :: v1, v2, v3, v4, v5, v6, v7, v8, v9
+    logical :: l, r
+    l = .false.
+    r = .false.
+    a(:, :, :) = 0
+    b(:, :, :) = 0
+    !$acc parallel pcopyin (v1, v2, v3, v4, v5, v6, v7, v8, v9) reduction (.or.:l)
+    !$acc loop reduction (.or.:l) collapse (3)
+      do i = v1, v2, v7
+        do j = v3, v4, v8
+          do k = v5, v6, v9
+            l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+            l = l.or.k.lt.13.or.k.gt.18
+            if (.not.l) a(i, j, k) = a(i, j, k) + 1
+          end do
+        end do
+      end do
+    !$acc end parallel
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+          r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          r = r.or.k.lt.13.or.k.gt.18
+          if (.not.l) b(i, j, k) = b(i, j, k) + 1
+        end do
+      end do
+    end do
+    if (l .neqv. r) call abort
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+           if (a(i, j, k) .ne. b(i, j, k)) call abort
+        end do
+      end do
+    end do
+  end subroutine test3
+
+  subroutine test4
+    integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+    integer :: v1, v2, v3, v4, v5, v6, v7, v8, v9
+    logical :: l, r
+    l = .false.
+    r = .false.
+    a(:, :, :) = 0
+    b(:, :, :) = 0
+    v1 = 2
+    v2 = 6
+    v3 = -2
+    v4 = 4
+    v5 = 13
+    v6 = 18
+    v7 = 1
+    v8 = 1
+    v9 = 1
+    !$acc parallel pcopyin (v1, v2, v3, v4, v5, v6, v7, v8, v9) reduction (.or.:l)
+    !$acc loop reduction (.or.:l) collapse (3)
+      do i = v1, v2, v7
+        do j = v3, v4, v8
+          do k = v5, v6, v9
+            l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+            l = l.or.k.lt.13.or.k.gt.18
+            if (.not.l) a(i, j, k) = a(i, j, k) + 1
+          end do
+        end do
+      end do
+    !$acc end parallel
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+          r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          r = r.or.k.lt.13.or.k.gt.18
+          if (.not.r) b(i, j, k) = b(i, j, k) + 1
+        end do
+      end do
+    end do
+    if (l .neqv. r) call abort
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+         do k = v5, v6, v9
+           if (a(i, j, k) .ne. b(i, j, k)) call abort
+         end do
+      end do
+    end do
+  end subroutine test4
+
+  subroutine test5(v1, v2, v3, v4, v5, v6)
+    integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+    integer :: v1, v2, v3, v4, v5, v6
+    logical :: l, r
+    l = .false.
+    r = .false.
+    a(:, :, :) = 0
+    b(:, :, :) = 0
+    !$acc parallel pcopyin (v1, v2, v3, v4, v5, v6) reduction (.or.:l)
+    !$acc loop reduction (.or.:l) collapse (3)
+      do i = v1, v2
+        do j = v3, v4
+          do k = v5, v6
+            l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+            l = l.or.k.lt.13.or.k.gt.18
+            if (.not.l) a(i, j, k) = a(i, j, k) + 1
+          end do
+        end do
+      end do
+    !$acc end parallel
+    do i = v1, v2
+      do j = v3, v4
+        do k = v5, v6
+          r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          r = r.or.k.lt.13.or.k.gt.18
+          if (.not.r) b(i, j, k) = b(i, j, k) + 1
+        end do
+      end do
+    end do
+    if (l .neqv. r) call abort
+    do i = v1, v2
+      do j = v3, v4
+        do k = v5, v6
+           if (a(i, j, k) .ne. b(i, j, k)) call abort
+        end do
+      end do
+    end do
+  end subroutine test5
+
+  subroutine test6(v1, v2, v3, v4, v5, v6, v7, v8, v9)
+    integer :: i, j, k, a(1:7, -3:5, 12:19), b(1:7, -3:5, 12:19)
+    integer :: v1, v2, v3, v4, v5, v6, v7, v8, v9
+    logical :: l, r
+    l = .false.
+    r = .false.
+    a(:, :, :) = 0
+    b(:, :, :) = 0
+    !$acc parallel pcopyin (v1, v2, v3, v4, v5, v6, v7, v8, v9) reduction (.or.:l)
+    !$acc loop reduction (.or.:l) collapse (3)
+      do i = v1, v2, v7
+        do j = v3, v4, v8
+          do k = v5, v6, v9
+            l = l.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+            l = l.or.k.lt.13.or.k.gt.18
+            if (.not.l) a(i, j, k) = a(i, j, k) + 1
+            m = i * 100 + j * 10 + k
+          end do
+        end do
+      end do
+    !$acc end parallel
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+          r = r.or.i.lt.2.or.i.gt.6.or.j.lt.-2.or.j.gt.4
+          r = r.or.k.lt.13.or.k.gt.18
+          if (.not.r) b(i, j, k) = b(i, j, k) + 1
+        end do
+      end do
+    end do
+    if (l .neqv. r) call abort
+    do i = v1, v2, v7
+      do j = v3, v4, v8
+        do k = v5, v6, v9
+           if (a(i, j, k) .ne. b(i, j, k)) call abort
+        end do
+      end do
+    end do
+  end subroutine test6
+
+end program sub_collapse_3