Hi!
On 2024-11-06T16:18:50+0000, Andrew Stubbs <a...@baylibre.com> wrote:
> On 06/11/2024 15:41, Jakub Jelinek wrote:
>> On Wed, Nov 06, 2024 at 03:27:22PM +0000, Andrew Stubbs wrote:
>>> [...] requires enabling the offload-dump scanning features previously only
>>> used
>>> in the libgomp testsuite. The automake scheme used there isn't a good fit
>>> here, so we probe the known devices manually.
>>>
>>> gcc/testsuite/ChangeLog:
>>>
>>> * gcc.dg/gomp/gomp.exp: Load scanoffload.exp and scanoffloadtree.exp.
>>> Set offload_targets when available.
>>> * gcc.dg/gomp/max_vf-1.c: New test.
>>> * gcc.dg/gomp/max_vf-2.c: New test.
>>> * gcc.dg/gomp/max_vf-3.c: New test.
>>
>> I don't see how this can work. gomp.exp isn't prepared to find the libgomp
>> directory nor add -B options etc.
ACK.
>> Perhaps it appears to work if you have
>> your system gcc's libgomp installed, but that isn't the library that should
>> be used.
>> So, max_vf-1.c test can stay where it is, but the gomp.exp changes shouldn't
>> be done and max_vf-{2,3}.c should move to libgomp/testsuite/libgomp.c/
>
> It worked for me, but I might have an unusual configuration that allows
> me to test installed toolchains with remote devices, rather than build
> trees with local devices.
Right. (And, of course, boths way should work.)
> --- /dev/null
> +++ b/libgomp/testsuite/libgomp.c/max_vf-1.c
> @@ -0,0 +1,47 @@
> +/* Test that omp parallel simd schedule uses the correct max_vf for the
> + host system, when target directives are present. */
> +
> +/* { dg-require-effective-target offloading_enabled } */
> +
> +[...]
> +
> +/* Make sure that the max_vf is used as an IFN.
> +{ dg-final { scan-tree-dump-times {GOMP_MAX_VF} 2 "ompexp" { target {
> x86_64-*-* i?86-*-* } } } } */
> +
> +/* Make sure the max_vf is passed as a temporary variable.
> +{ dg-final { scan-tree-dump-times
> {__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \(.*, D\.[0-9]*, 0\);} 1
> "ompexp" { target { x86_64-*-* i?86-*-* } } } } */
> +
> +/* Test SIMD offload devices
> +{ dg-final { scan-offload-tree-dump-times
> {__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \(.*, 64, 0\);} 1
> "optimized" { target { offload_gcn } } } }
> +{ dg-final { scan-offload-tree-dump-times
> {__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \(.*, 7, 0\);} 1
> "optimized" { target { offload_nvptx } } } } */
> +
> +int main() {}
For configurations where both GCN and nvptx offloading are enabled, we
get FAILs here. Avoid these via 'only_for_offload_target [...]'.
Pushed to trunk branch commit 730f28b081bea4a749f9b82902446731ec8faa93
"Adjust 'libgomp.c/max_vf-*.c'", see attached.
Grüße
Thomas
>From 730f28b081bea4a749f9b82902446731ec8faa93 Mon Sep 17 00:00:00 2001
From: Thomas Schwinge <tschwi...@baylibre.com>
Date: Sat, 9 Nov 2024 13:37:53 +0100
Subject: [PATCH] Adjust 'libgomp.c/max_vf-*.c'
For configurations where both GCN and nvptx offloading are enabled, we get:
PASS: libgomp.c/max_vf-1.c (test for excess errors)
PASS: libgomp.c/max_vf-1.c scan-tree-dump-times ompexp "GOMP_MAX_VF" 2
PASS: libgomp.c/max_vf-1.c scan-tree-dump-times ompexp "__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \\(.*, D\\.[0-9]*, 0\\);" 1
PASS: libgomp.c/max_vf-1.c scan-amdgcn-amdhsa-offload-tree-dump-times optimized "__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \\(.*, 64, 0\\);" 1
FAIL: libgomp.c/max_vf-1.c scan-nvptx-none-offload-tree-dump-times optimized "__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \\(.*, 64, 0\\);" 1
FAIL: libgomp.c/max_vf-1.c scan-amdgcn-amdhsa-offload-tree-dump-times optimized "__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \\(.*, 7, 0\\);" 1
PASS: libgomp.c/max_vf-1.c scan-nvptx-none-offload-tree-dump-times optimized "__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \\(.*, 7, 0\\);" 1
Avoid these FAILs via 'only_for_offload_target [...]'. Also, for consistency
with other libgomp test cases, use effective-target specifiers of the libgomp
test suite. Fix-up for recent commit d334f729e53867b838e867375b3f475ba793d96e
"openmp: Add testcases for omp_max_vf".
libgomp/
* testsuite/libgomp.c/max_vf-1.c: Adjust.
* testsuite/libgomp.c/max_vf-2.c: Likewise.
---
libgomp/testsuite/libgomp.c/max_vf-1.c | 6 +++---
libgomp/testsuite/libgomp.c/max_vf-2.c | 4 ++--
2 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/libgomp/testsuite/libgomp.c/max_vf-1.c b/libgomp/testsuite/libgomp.c/max_vf-1.c
index 9c8d5dc0af9..70f6b86c614 100644
--- a/libgomp/testsuite/libgomp.c/max_vf-1.c
+++ b/libgomp/testsuite/libgomp.c/max_vf-1.c
@@ -1,7 +1,7 @@
/* Test that omp parallel simd schedule uses the correct max_vf for the
host system, when target directives are present. */
-/* { dg-require-effective-target offloading_enabled } */
+/* { dg-require-effective-target offload_target_any } */
/* { dg-do link } */
/* { dg-options "-fopenmp -O2 -fdump-tree-ompexp -foffload=-fdump-tree-optimized" } */
@@ -41,7 +41,7 @@ f3 (int *a, int *b, int *c)
{ dg-final { scan-tree-dump-times {__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \(.*, D\.[0-9]*, 0\);} 1 "ompexp" { target { x86_64-*-* i?86-*-* } } } } */
/* Test SIMD offload devices
-{ dg-final { scan-offload-tree-dump-times {__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \(.*, 64, 0\);} 1 "optimized" { target { offload_gcn } } } }
-{ dg-final { scan-offload-tree-dump-times {__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \(.*, 7, 0\);} 1 "optimized" { target { offload_nvptx } } } } */
+{ dg-final { only_for_offload_target amdgcn-amdhsa scan-offload-tree-dump-times {__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \(.*, 64, 0\);} 1 "optimized" { target offload_target_amdgcn } } }
+{ dg-final { only_for_offload_target nvptx-none scan-offload-tree-dump-times {__builtin_GOMP_parallel_loop_nonmonotonic_dynamic \(.*, 7, 0\);} 1 "optimized" { target offload_target_nvptx } } } */
int main() {}
diff --git a/libgomp/testsuite/libgomp.c/max_vf-2.c b/libgomp/testsuite/libgomp.c/max_vf-2.c
index 91744c309df..545cf1d2138 100644
--- a/libgomp/testsuite/libgomp.c/max_vf-2.c
+++ b/libgomp/testsuite/libgomp.c/max_vf-2.c
@@ -1,7 +1,7 @@
/* Ensure that the default safelen is set correctly for the larger of the host
and offload device, to prevent defeating the vectorizer. */
-/* { dg-require-effective-target offloading_enabled } */
+/* { dg-require-effective-target offload_target_any } */
/* { dg-do link } */
/* { dg-options "-fopenmp -O2 -fdump-tree-omplower" } */
@@ -16,6 +16,6 @@ int f(float *a, float *b, int n)
}
/* Make sure that the max_vf used is suitable for the offload device.
-{ dg-final { scan-tree-dump-times {omp simd safelen\(64\)} 1 "omplower" { target { offload_gcn } } } } */
+{ dg-final { scan-tree-dump-times {omp simd safelen\(64\)} 1 "omplower" { target offload_target_amdgcn } } } */
int main() {}
--
2.34.1
