Hi!
On Mon, 04 Apr 2016 12:39:37 +0200, I wrote:
> [...] gomp-4_0-branch [...] additional (cleanup) changes [...]
> libgomp/
> [...]
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gang-np-1.c:
> Merge this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gv-np-1.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gw-np-1.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-1.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-2.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-3.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-4.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-1.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-2.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-worker-p-1.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-1.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-2.c:
> ... this file, and...
> * testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-3.c:
> ... this file into...
> * testsuite/libgomp.oacc-c-c++-common/reduction-7.c: ... this new
> file. Use dg-warning directives instead of specifying the -w
> compiler option.
> [...]
Cesar didn't pick these up in his recent trunk commit, so now applied
this to trunk in r234899:
commit 40495bd0847a05aa76cc37e05292cf937449f9dd
Author: tschwinge <tschwinge@138bc75d-0d04-0410-961f-82ee72b054a4>
Date: Tue Apr 12 11:02:32 2016 +0000
Merge libgomp.oacc-c-c++-common/loop-reduction-*.c into
libgomp.oacc-c-c++-common/reduction-7.c
libgomp/
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-gang-np-1.c:
Merge this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-gv-np-1.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-gw-np-1.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-1.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-2.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-3.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-4.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-1.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-2.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-worker-p-1.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-1.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-2.c:
... this file, and...
* testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-3.c:
... this file into...
* testsuite/libgomp.oacc-c-c++-common/reduction-7.c: ... this
file.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@234899
138bc75d-0d04-0410-961f-82ee72b054a4
---
libgomp/ChangeLog | 29 ++
.../loop-reduction-gang-np-1.c | 45 ---
.../loop-reduction-gv-np-1.c | 30 --
.../loop-reduction-gw-np-1.c | 30 --
.../loop-reduction-gwv-np-1.c | 28 --
.../loop-reduction-gwv-np-2.c | 34 --
.../loop-reduction-gwv-np-3.c | 33 --
.../loop-reduction-gwv-np-4.c | 55 ----
.../loop-reduction-vector-p-1.c | 43 ---
.../loop-reduction-vector-p-2.c | 41 ---
.../loop-reduction-worker-p-1.c | 43 ---
.../loop-reduction-wv-p-1.c | 41 ---
.../loop-reduction-wv-p-2.c | 45 ---
.../loop-reduction-wv-p-3.c | 38 ---
.../libgomp.oacc-c-c++-common/reduction-7.c | 351 +++++++++++++++++++++
15 files changed, 380 insertions(+), 506 deletions(-)
diff --git libgomp/ChangeLog libgomp/ChangeLog
index 6071b23..1716ba0 100644
--- libgomp/ChangeLog
+++ libgomp/ChangeLog
@@ -1,5 +1,34 @@
2016-04-12 Thomas Schwinge <tho...@codesourcery.com>
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gang-np-1.c:
+ Merge this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gv-np-1.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gw-np-1.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-1.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-2.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-3.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-4.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-1.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-2.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-worker-p-1.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-1.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-2.c:
+ ... this file, and...
+ * testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-3.c:
+ ... this file into...
+ * testsuite/libgomp.oacc-c-c++-common/reduction-7.c: ... this
+ file.
+
* testsuite/libgomp.oacc-c-c++-common/par-loop-comb-reduction-2.c:
Make failure observable.
diff --git
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gang-np-1.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gang-np-1.c
deleted file mode 100644
index 55ab3c9..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gang-np-1.c
+++ /dev/null
@@ -1,45 +0,0 @@
-/* { dg-additional-options "-w" } */
-
-#include <assert.h>
-
-/* Test of reduction on loop directive (gangs, non-private reduction
- variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, arr[1024], res = 0, hres = 0;
-
- for (i = 0; i < 1024; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- copy(res)
- {
- #pragma acc loop gang reduction(+:res)
- for (i = 0; i < 1024; i++)
- res += arr[i];
- }
-
- for (i = 0; i < 1024; i++)
- hres += arr[i];
-
- assert (res == hres);
-
- res = hres = 1;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- copy(res)
- {
- #pragma acc loop gang reduction(*:res)
- for (i = 0; i < 12; i++)
- res *= arr[i];
- }
-
- for (i = 0; i < 12; i++)
- hres *= arr[i];
-
- assert (res == hres);
-
- return 0;
-}
diff --git libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gv-np-1.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gv-np-1.c
deleted file mode 100644
index d4341e9..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gv-np-1.c
+++ /dev/null
@@ -1,30 +0,0 @@
-/* { dg-additional-options "-w" } */
-
-#include <assert.h>
-
-/* Test of reduction on loop directive (gangs and vectors, non-private
- reduction variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, arr[1024], res = 0, hres = 0;
-
- for (i = 0; i < 1024; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- copy(res)
- {
- #pragma acc loop gang vector reduction(+:res)
- for (i = 0; i < 1024; i++)
- res += arr[i];
- }
-
- for (i = 0; i < 1024; i++)
- hres += arr[i];
-
- assert (res == hres);
-
- return 0;
-}
diff --git libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gw-np-1.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gw-np-1.c
deleted file mode 100644
index 2e5668b..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gw-np-1.c
+++ /dev/null
@@ -1,30 +0,0 @@
-/* { dg-additional-options "-w" } */
-
-#include <assert.h>
-
-/* Test of reduction on loop directive (gangs and workers, non-private
- reduction variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, arr[1024], res = 0, hres = 0;
-
- for (i = 0; i < 1024; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- copy(res)
- {
- #pragma acc loop gang worker reduction(+:res)
- for (i = 0; i < 1024; i++)
- res += arr[i];
- }
-
- for (i = 0; i < 1024; i++)
- hres += arr[i];
-
- assert (res == hres);
-
- return 0;
-}
diff --git
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-1.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-1.c
deleted file mode 100644
index d610373..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-1.c
+++ /dev/null
@@ -1,28 +0,0 @@
-#include <assert.h>
-
-/* Test of reduction on loop directive (gangs, workers and vectors, non-private
- reduction variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, arr[1024], res = 0, hres = 0;
-
- for (i = 0; i < 1024; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- copy(res)
- {
- #pragma acc loop gang worker vector reduction(+:res)
- for (i = 0; i < 1024; i++)
- res += arr[i];
- }
-
- for (i = 0; i < 1024; i++)
- hres += arr[i];
-
- assert (res == hres);
-
- return 0;
-}
diff --git
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-2.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-2.c
deleted file mode 100644
index ea5c151..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-2.c
+++ /dev/null
@@ -1,34 +0,0 @@
-#include <assert.h>
-
-/* Test of reduction on loop directive (gangs, workers and vectors, non-private
- reduction variable: separate gang and worker/vector loops). */
-
-int
-main (int argc, char *argv[])
-{
- int i, j, arr[32768], res = 0, hres = 0;
-
- for (i = 0; i < 32768; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- copy(res)
- {
- #pragma acc loop gang reduction(+:res)
- for (j = 0; j < 32; j++)
- {
- #pragma acc loop worker vector reduction(+:res)
- for (i = 0; i < 1024; i++)
- res += arr[j * 1024 + i];
- }
- /* "res" is non-private, and is not available until after the parallel
- region. */
- }
-
- for (i = 0; i < 32768; i++)
- hres += arr[i];
-
- assert (res == hres);
-
- return 0;
-}
diff --git
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-3.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-3.c
deleted file mode 100644
index 0056f3c..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-3.c
+++ /dev/null
@@ -1,33 +0,0 @@
-#include <assert.h>
-
-/* Test of reduction on loop directive (gangs, workers and vectors, non-private
- reduction variable: separate gang and worker/vector loops). */
-
-int
-main (int argc, char *argv[])
-{
- int i, j;
- double arr[32768], res = 0, hres = 0;
-
- for (i = 0; i < 32768; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- copyin(arr) copy(res)
- {
- #pragma acc loop gang reduction(+:res)
- for (j = 0; j < 32; j++)
- {
- #pragma acc loop worker vector reduction(+:res)
- for (i = 0; i < 1024; i++)
- res += arr[j * 1024 + i];
- }
- }
-
- for (i = 0; i < 32768; i++)
- hres += arr[i];
-
- assert (res == hres);
-
- return 0;
-}
diff --git
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-4.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-4.c
deleted file mode 100644
index e69d0ec..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-gwv-np-4.c
+++ /dev/null
@@ -1,55 +0,0 @@
-#include <assert.h>
-
-/* Test of reduction on loop directive (gangs, workers and vectors, multiple
- non-private reduction variables, float type). */
-
-int
-main (int argc, char *argv[])
-{
- int i, j;
- float arr[32768];
- float res = 0, mres = 0, hres = 0, hmres = 0;
-
- for (i = 0; i < 32768; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- copy(res, mres)
- {
- #pragma acc loop gang reduction(+:res) reduction(max:mres)
- for (j = 0; j < 32; j++)
- {
- #pragma acc loop worker vector reduction(+:res) reduction(max:mres)
- for (i = 0; i < 1024; i++)
- {
- res += arr[j * 1024 + i];
- if (arr[j * 1024 + i] > mres)
- mres = arr[j * 1024 + i];
- }
-
- #pragma acc loop worker vector reduction(+:res) reduction(max:mres)
- for (i = 0; i < 1024; i++)
- {
- res += arr[j * 1024 + (1023 - i)];
- if (arr[j * 1024 + (1023 - i)] > mres)
- mres = arr[j * 1024 + (1023 - i)];
- }
- }
- }
-
- for (j = 0; j < 32; j++)
- for (i = 0; i < 1024; i++)
- {
- hres += arr[j * 1024 + i];
- hres += arr[j * 1024 + (1023 - i)];
- if (arr[j * 1024 + i] > hmres)
- hmres = arr[j * 1024 + i];
- if (arr[j * 1024 + (1023 - i)] > hmres)
- hmres = arr[j * 1024 + (1023 - i)];
- }
-
- assert (res == hres);
- assert (mres == hmres);
-
- return 0;
-}
diff --git
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-1.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-1.c
deleted file mode 100644
index 31e4366..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-1.c
+++ /dev/null
@@ -1,43 +0,0 @@
-/* { dg-additional-options "-w" } */
-
-#include <assert.h>
-
-/* Test of reduction on loop directive (vectors, private reduction
- variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, j, arr[1024], out[32], res = 0, hres = 0;
-
- for (i = 0; i < 1024; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- private(res) copyout(out)
- {
- #pragma acc loop gang
- for (j = 0; j < 32; j++)
- {
- res = 0;
-
- #pragma acc loop vector reduction(+:res)
- for (i = 0; i < 32; i++)
- res += arr[j * 32 + i];
-
- out[j] = res;
- }
- }
-
- for (j = 0; j < 32; j++)
- {
- hres = 0;
-
- for (i = 0; i < 32; i++)
- hres += arr[j * 32 + i];
-
- assert (out[j] == hres);
- }
-
- return 0;
-}
diff --git
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-2.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-2.c
deleted file mode 100644
index 15f0053..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-vector-p-2.c
+++ /dev/null
@@ -1,41 +0,0 @@
-#include <assert.h>
-
-/* Test of reduction on loop directive (vector reduction in
- gang-partitioned/worker-partitioned mode, private reduction variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, j, k;
- double ina[1024], inb[1024], out[1024], acc;
-
- for (j = 0; j < 32; j++)
- for (i = 0; i < 32; i++)
- {
- ina[j * 32 + i] = (i == j) ? 2.0 : 0.0;
- inb[j * 32 + i] = (double) (i + j);
- }
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- private(acc) copyin(ina, inb) copyout(out)
- {
- #pragma acc loop gang worker
- for (k = 0; k < 32; k++)
- for (j = 0; j < 32; j++)
- {
- acc = 0;
-
- #pragma acc loop vector reduction(+:acc)
- for (i = 0; i < 32; i++)
- acc += ina[k * 32 + i] * inb[i * 32 + j];
-
- out[k * 32 + j] = acc;
- }
- }
-
- for (j = 0; j < 32; j++)
- for (i = 0; i < 32; i++)
- assert (out[j * 32 + i] == (i + j) * 2);
-
- return 0;
-}
diff --git
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-worker-p-1.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-worker-p-1.c
deleted file mode 100644
index 4a92503..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-worker-p-1.c
+++ /dev/null
@@ -1,43 +0,0 @@
-/* { dg-additional-options "-w" } */
-
-#include <assert.h>
-
-/* Test of reduction on loop directive (workers, private reduction
- variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, j, arr[1024], out[32], res = 0, hres = 0;
-
- for (i = 0; i < 1024; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- private(res) copyout(out)
- {
- #pragma acc loop gang
- for (j = 0; j < 32; j++)
- {
- res = 0;
-
- #pragma acc loop worker reduction(+:res)
- for (i = 0; i < 32; i++)
- res += arr[j * 32 + i];
-
- out[j] = res;
- }
- }
-
- for (j = 0; j < 32; j++)
- {
- hres = 0;
-
- for (i = 0; i < 32; i++)
- hres += arr[j * 32 + i];
-
- assert (out[j] == hres);
- }
-
- return 0;
-}
diff --git libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-1.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-1.c
deleted file mode 100644
index 1bfb284..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-1.c
+++ /dev/null
@@ -1,41 +0,0 @@
-#include <assert.h>
-
-/* Test of reduction on loop directive (workers and vectors, private reduction
- variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, j, arr[1024], out[32], res = 0, hres = 0;
-
- for (i = 0; i < 1024; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- private(res) copyout(out)
- {
- #pragma acc loop gang
- for (j = 0; j < 32; j++)
- {
- res = 0;
-
- #pragma acc loop worker vector reduction(+:res)
- for (i = 0; i < 32; i++)
- res += arr[j * 32 + i];
-
- out[j] = res;
- }
- }
-
- for (j = 0; j < 32; j++)
- {
- hres = 0;
-
- for (i = 0; i < 32; i++)
- hres += arr[j * 32 + i];
-
- assert (out[j] == hres);
- }
-
- return 0;
-}
diff --git libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-2.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-2.c
deleted file mode 100644
index 93ab78f..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-2.c
+++ /dev/null
@@ -1,45 +0,0 @@
-#include <assert.h>
-
-/* Test of reduction on loop directive (workers and vectors, private reduction
- variable). */
-
-int
-main (int argc, char *argv[])
-{
- int i, j, arr[32768], out[32], res = 0, hres = 0;
-
- for (i = 0; i < 32768; i++)
- arr[i] = i;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- private(res) copyout(out)
- {
- #pragma acc loop gang
- for (j = 0; j < 32; j++)
- {
- res = j;
-
- #pragma acc loop worker reduction(+:res)
- for (i = 0; i < 1024; i++)
- res += arr[j * 1024 + i];
-
- #pragma acc loop vector reduction(+:res)
- for (i = 1023; i >= 0; i--)
- res += arr[j * 1024 + i];
-
- out[j] = res;
- }
- }
-
- for (j = 0; j < 32; j++)
- {
- hres = j;
-
- for (i = 0; i < 1024; i++)
- hres += arr[j * 1024 + i] * 2;
-
- assert (out[j] == hres);
- }
-
- return 0;
-}
diff --git libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-3.c
libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-3.c
deleted file mode 100644
index 298e25c..0000000
--- libgomp/testsuite/libgomp.oacc-c-c++-common/loop-reduction-wv-p-3.c
+++ /dev/null
@@ -1,38 +0,0 @@
-#include <assert.h>
-
-/* Test of reduction on loop directive (workers and vectors, private reduction
- variable: gang-redundant mode). */
-
-int
-main (int argc, char *argv[])
-{
- int i, arr[1024], out[32], res = 0, hres = 0;
-
- for (i = 0; i < 1024; i++)
- arr[i] = i ^ 33;
-
- #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
- private(res) copyin(arr) copyout(out)
- {
- /* Private variables aren't initialized by default in openacc. */
- res = 0;
-
- /* "res" should be available at the end of the following loop (and should
- have the same value redundantly in each gang). */
- #pragma acc loop worker vector reduction(+:res)
- for (i = 0; i < 1024; i++)
- res += arr[i];
-
- #pragma acc loop gang (static: 1)
- for (i = 0; i < 32; i++)
- out[i] = res;
- }
-
- for (i = 0; i < 1024; i++)
- hres += arr[i];
-
- for (i = 0; i < 32; i++)
- assert (out[i] == hres);
-
- return 0;
-}
diff --git libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-7.c
libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-7.c
index b23c758..76c33e4 100644
--- libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-7.c
+++ libgomp/testsuite/libgomp.oacc-c-c++-common/reduction-7.c
@@ -118,12 +118,363 @@ void gwv_np_1()
}
+/* Test of reduction on loop directive (gangs, workers and vectors, non-private
+ reduction variable: separate gang and worker/vector loops). */
+
+void gwv_np_2()
+{
+ int i, j, arr[32768], res = 0, hres = 0;
+
+ for (i = 0; i < 32768; i++)
+ arr[i] = i;
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ copy(res)
+ {
+ #pragma acc loop gang reduction(+:res)
+ for (j = 0; j < 32; j++)
+ {
+ #pragma acc loop worker vector reduction(+:res)
+ for (i = 0; i < 1024; i++)
+ res += arr[j * 1024 + i];
+ }
+ /* "res" is non-private, and is not available until after the parallel
+ region. */
+ }
+
+ for (i = 0; i < 32768; i++)
+ hres += arr[i];
+
+ assert (res == hres);
+}
+
+
+/* Test of reduction on loop directive (gangs, workers and vectors, non-private
+ reduction variable: separate gang and worker/vector loops). */
+
+void gwv_np_3()
+{
+ int i, j;
+ double arr[32768], res = 0, hres = 0;
+
+ for (i = 0; i < 32768; i++)
+ arr[i] = i;
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ copyin(arr) copy(res)
+ {
+ #pragma acc loop gang reduction(+:res)
+ for (j = 0; j < 32; j++)
+ {
+ #pragma acc loop worker vector reduction(+:res)
+ for (i = 0; i < 1024; i++)
+ res += arr[j * 1024 + i];
+ }
+ }
+
+ for (i = 0; i < 32768; i++)
+ hres += arr[i];
+
+ assert (res == hres);
+}
+
+
+/* Test of reduction on loop directive (gangs, workers and vectors, multiple
+ non-private reduction variables, float type). */
+
+void gwv_np_4()
+{
+ int i, j;
+ float arr[32768];
+ float res = 0, mres = 0, hres = 0, hmres = 0;
+
+ for (i = 0; i < 32768; i++)
+ arr[i] = i;
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ copy(res, mres)
+ {
+ #pragma acc loop gang reduction(+:res) reduction(max:mres)
+ for (j = 0; j < 32; j++)
+ {
+ #pragma acc loop worker vector reduction(+:res) reduction(max:mres)
+ for (i = 0; i < 1024; i++)
+ {
+ res += arr[j * 1024 + i];
+ if (arr[j * 1024 + i] > mres)
+ mres = arr[j * 1024 + i];
+ }
+
+ #pragma acc loop worker vector reduction(+:res) reduction(max:mres)
+ for (i = 0; i < 1024; i++)
+ {
+ res += arr[j * 1024 + (1023 - i)];
+ if (arr[j * 1024 + (1023 - i)] > mres)
+ mres = arr[j * 1024 + (1023 - i)];
+ }
+ }
+ }
+
+ for (j = 0; j < 32; j++)
+ for (i = 0; i < 1024; i++)
+ {
+ hres += arr[j * 1024 + i];
+ hres += arr[j * 1024 + (1023 - i)];
+ if (arr[j * 1024 + i] > hmres)
+ hmres = arr[j * 1024 + i];
+ if (arr[j * 1024 + (1023 - i)] > hmres)
+ hmres = arr[j * 1024 + (1023 - i)];
+ }
+
+ assert (res == hres);
+ assert (mres == hmres);
+}
+
+
+/* Test of reduction on loop directive (vectors, private reduction
+ variable). */
+
+void v_p_1()
+{
+ int i, j, arr[1024], out[32], res = 0, hres = 0;
+
+ for (i = 0; i < 1024; i++)
+ arr[i] = i;
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ private(res) copyout(out)
+ {
+ #pragma acc loop gang
+ for (j = 0; j < 32; j++)
+ {
+ res = 0;
+
+ #pragma acc loop vector reduction(+:res)
+ for (i = 0; i < 32; i++)
+ res += arr[j * 32 + i];
+
+ out[j] = res;
+ }
+ }
+
+ for (j = 0; j < 32; j++)
+ {
+ hres = 0;
+
+ for (i = 0; i < 32; i++)
+ hres += arr[j * 32 + i];
+
+ assert (out[j] == hres);
+ }
+}
+
+
+/* Test of reduction on loop directive (vector reduction in
+ gang-partitioned/worker-partitioned mode, private reduction variable). */
+
+void v_p_2()
+{
+ int i, j, k;
+ double ina[1024], inb[1024], out[1024], acc;
+
+ for (j = 0; j < 32; j++)
+ for (i = 0; i < 32; i++)
+ {
+ ina[j * 32 + i] = (i == j) ? 2.0 : 0.0;
+ inb[j * 32 + i] = (double) (i + j);
+ }
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ private(acc) copyin(ina, inb) copyout(out)
+ {
+ #pragma acc loop gang worker
+ for (k = 0; k < 32; k++)
+ for (j = 0; j < 32; j++)
+ {
+ acc = 0;
+
+ #pragma acc loop vector reduction(+:acc)
+ for (i = 0; i < 32; i++)
+ acc += ina[k * 32 + i] * inb[i * 32 + j];
+
+ out[k * 32 + j] = acc;
+ }
+ }
+
+ for (j = 0; j < 32; j++)
+ for (i = 0; i < 32; i++)
+ assert (out[j * 32 + i] == (i + j) * 2);
+}
+
+
+/* Test of reduction on loop directive (workers, private reduction
+ variable). */
+
+void w_p_1()
+{
+ int i, j, arr[1024], out[32], res = 0, hres = 0;
+
+ for (i = 0; i < 1024; i++)
+ arr[i] = i;
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ private(res) copyout(out)
+ {
+ #pragma acc loop gang
+ for (j = 0; j < 32; j++)
+ {
+ res = 0;
+
+ #pragma acc loop worker reduction(+:res)
+ for (i = 0; i < 32; i++)
+ res += arr[j * 32 + i];
+
+ out[j] = res;
+ }
+ }
+
+ for (j = 0; j < 32; j++)
+ {
+ hres = 0;
+
+ for (i = 0; i < 32; i++)
+ hres += arr[j * 32 + i];
+
+ assert (out[j] == hres);
+ }
+}
+
+
+/* Test of reduction on loop directive (workers and vectors, private reduction
+ variable). */
+
+void wv_p_1()
+{
+ int i, j, arr[1024], out[32], res = 0, hres = 0;
+
+ for (i = 0; i < 1024; i++)
+ arr[i] = i;
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ private(res) copyout(out)
+ {
+ #pragma acc loop gang
+ for (j = 0; j < 32; j++)
+ {
+ res = 0;
+
+ #pragma acc loop worker vector reduction(+:res)
+ for (i = 0; i < 32; i++)
+ res += arr[j * 32 + i];
+
+ out[j] = res;
+ }
+ }
+
+ for (j = 0; j < 32; j++)
+ {
+ hres = 0;
+
+ for (i = 0; i < 32; i++)
+ hres += arr[j * 32 + i];
+
+ assert (out[j] == hres);
+ }
+}
+
+
+/* Test of reduction on loop directive (workers and vectors, private reduction
+ variable). */
+
+void wv_p_2()
+{
+ int i, j, arr[32768], out[32], res = 0, hres = 0;
+
+ for (i = 0; i < 32768; i++)
+ arr[i] = i;
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ private(res) copyout(out)
+ {
+ #pragma acc loop gang
+ for (j = 0; j < 32; j++)
+ {
+ res = j;
+
+ #pragma acc loop worker reduction(+:res)
+ for (i = 0; i < 1024; i++)
+ res += arr[j * 1024 + i];
+
+ #pragma acc loop vector reduction(+:res)
+ for (i = 1023; i >= 0; i--)
+ res += arr[j * 1024 + i];
+
+ out[j] = res;
+ }
+ }
+
+ for (j = 0; j < 32; j++)
+ {
+ hres = j;
+
+ for (i = 0; i < 1024; i++)
+ hres += arr[j * 1024 + i] * 2;
+
+ assert (out[j] == hres);
+ }
+}
+
+
+/* Test of reduction on loop directive (workers and vectors, private reduction
+ variable: gang-redundant mode). */
+
+void wv_p_3()
+{
+ int i, arr[1024], out[32], res = 0, hres = 0;
+
+ for (i = 0; i < 1024; i++)
+ arr[i] = i ^ 33;
+
+ #pragma acc parallel num_gangs(32) num_workers(32) vector_length(32) \
+ private(res) copyin(arr) copyout(out)
+ {
+ /* Private variables aren't initialized by default in openacc. */
+ res = 0;
+
+ /* "res" should be available at the end of the following loop (and should
+ have the same value redundantly in each gang). */
+ #pragma acc loop worker vector reduction(+:res)
+ for (i = 0; i < 1024; i++)
+ res += arr[i];
+
+ #pragma acc loop gang (static: 1)
+ for (i = 0; i < 32; i++)
+ out[i] = res;
+ }
+
+ for (i = 0; i < 1024; i++)
+ hres += arr[i];
+
+ for (i = 0; i < 32; i++)
+ assert (out[i] == hres);
+}
+
+
int main()
{
g_np_1();
gv_np_1();
gw_np_1();
gwv_np_1();
+ gwv_np_2();
+ gwv_np_3();
+ gwv_np_4();
+ v_p_1();
+ v_p_2();
+ w_p_1();
+ wv_p_1();
+ wv_p_2();
+ wv_p_3();
return 0;
}
Grüße
Thomas
