On Thu, 4 Feb 2021 16:03:44 +0100
Tobias Burnus <tob...@codesourcery.com> wrote:

> LGTM.
> 
> However, I'd feel better if there were a testcase, which actually
> checks that it works correctly. (I think that means a
> libgomp/testsuite/ run test.)

Thanks! FYI, I've committed this version with an additional runtime
test.

Julian
>From 9a4d32f85ccebc0ee4b24e6d9d7a4f11c04d7146 Mon Sep 17 00:00:00 2001
From: Julian Brown <jul...@codesourcery.com>
Date: Tue, 2 Feb 2021 03:44:34 -0800
Subject: [PATCH 3/4] openacc: Allow strided arrays in update directives

OpenACC 3.0 ("2.14.4. Update Directive") states:

  Noncontiguous subarrays may appear. It is implementation-specific
  whether noncontiguous regions are updated by using one transfer for
  each contiguous subregion, or whether the non-contiguous data is
  packed, transferred once, and unpacked, or whether one or more larger
  subarrays (no larger than the smallest contiguous region that contains
  the specified subarray) are updated.

This patch relaxes some conditions in the Fortran front-end so that
strided accesses are permitted for update directives.

gcc/fortran/
	* openmp.c (resolve_omp_clauses): Omit OpenACC update in
	contiguity check and stride-specified error.

gcc/testsuite/
	* gfortran.dg/goacc/array-with-dt-2.f90: New test.

libgomp/
	* testsuite/libgomp.oacc-fortran/array-stride-dt-1.f90: New test.
---
 gcc/fortran/openmp.c                          |  5 ++-
 .../gfortran.dg/goacc/array-with-dt-2.f90     | 10 +++++
 .../array-stride-dt-1.f90                     | 44 +++++++++++++++++++
 3 files changed, 57 insertions(+), 2 deletions(-)
 create mode 100644 gcc/testsuite/gfortran.dg/goacc/array-with-dt-2.f90
 create mode 100644 libgomp/testsuite/libgomp.oacc-fortran/array-stride-dt-1.f90

diff --git a/gcc/fortran/openmp.c b/gcc/fortran/openmp.c
index aab17f0589f..797f6c86b62 100644
--- a/gcc/fortran/openmp.c
+++ b/gcc/fortran/openmp.c
@@ -5192,7 +5192,8 @@ resolve_omp_clauses (gfc_code *code, gfc_omp_clauses *omp_clauses,
 			   array isn't contiguous.  An expression such as
 			   arr(-n:n,-n:n) could be contiguous even if it looks
 			   like it may not be.  */
-			if (list != OMP_LIST_CACHE
+			if (code->op != EXEC_OACC_UPDATE
+			    && list != OMP_LIST_CACHE
 			    && list != OMP_LIST_DEPEND
 			    && !gfc_is_simply_contiguous (n->expr, false, true)
 			    && gfc_is_not_contiguous (n->expr))
@@ -5230,7 +5231,7 @@ resolve_omp_clauses (gfc_code *code, gfc_omp_clauses *omp_clauses,
 			int i;
 			gfc_array_ref *ar = &array_ref->u.ar;
 			for (i = 0; i < ar->dimen; i++)
-			  if (ar->stride[i])
+			  if (ar->stride[i] && code->op != EXEC_OACC_UPDATE)
 			    {
 			      gfc_error ("Stride should not be specified for "
 					 "array section in %s clause at %L",
diff --git a/gcc/testsuite/gfortran.dg/goacc/array-with-dt-2.f90 b/gcc/testsuite/gfortran.dg/goacc/array-with-dt-2.f90
new file mode 100644
index 00000000000..807580d75a9
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/goacc/array-with-dt-2.f90
@@ -0,0 +1,10 @@
+type t
+   integer, allocatable :: A(:,:)
+end type t
+
+type(t), allocatable :: b(:)
+
+!$acc update host(b(::2))
+!$acc update host(b(1)%A(::3,::4))
+end
+
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/array-stride-dt-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/array-stride-dt-1.f90
new file mode 100644
index 00000000000..f04d76d583a
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/array-stride-dt-1.f90
@@ -0,0 +1,44 @@
+! { dg-do run }
+
+type t
+   integer, allocatable :: A(:,:)
+end type t
+
+type(t), allocatable :: b(:)
+
+integer :: i
+
+allocate(b(1:20))
+do i=1,20
+  allocate(b(i)%A(1:20,1:20))
+end do
+
+do i=1,20
+  b(i)%A(:,:) = 0
+end do
+
+!$acc enter data copyin(b)
+do i=1,20
+  !$acc enter data copyin(b(i)%A)
+end do
+
+b(1)%A(:,:) = 5
+
+!$acc update device(b(::2))
+!$acc update device(b(1)%A(::3,::4))
+
+do i=1,20
+  !$acc exit data copyout(b(i)%A)
+end do
+!$acc exit data copyout(b)
+
+! This is necessarily conservative because the "update" is allowed to copy
+! e.g. the whole of the containing block for a discontinuous update.
+! Try to ensure that the update covers a sufficient portion of the array.
+
+if (any(b(1)%A(::3,::4) .ne. 5)) stop 1
+do i=2,20
+  if (any(b(i)%A(:,:) .ne. 0)) stop 2
+end do
+
+end
-- 
2.29.2

Reply via email to