Re: [PATCH, test, i386] Fix for PR50155

[Kirill Yukhin <kirill.yukhin@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 627 bytes --]

Thanks for inputs.
Updated test attached.

K

On Mon, Aug 22, 2011 at 11:12 PM, Jakub Jelinek <jakub@redhat.com> wrote:
> On Mon, Aug 22, 2011 at 10:51:19PM +0400, Kirill Yukhin wrote:
>> testsuite/ChangeLog entry:
>> 2011-08-22  Kirill Yukhin  <kirill.yukhin@intel.com>
>>
>>         PR target/50155
>>         * gcc.target/i386/pr50155.c: New test.
>
> For the testcase, I think you want explicit -mno-avx2 if
> you want to check that vpaddd hasn't been emitted, otherwise
> anybody testing with --with-arch=corei7-avx2 tuned compiler
> ot --target_board=unix/-mavx2 will see failures.
>
>        Jakub
>

[-- Attachment #2: pr50155.test.patch --]
[-- Type: application/octet-stream, Size: 491 bytes --]

diff --git a/gcc/testsuite/gcc.target/i386/pr50155.c b/gcc/testsuite/gcc.target/i386/pr50155.c
new file mode 100644
index 0000000..c641d4c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr50155.c
@@ -0,0 +1,13 @@
+/* { dg-do compile } */
+/* { dg-options "-O3 -mavx -mno-avx2" } */
+
+void
+foo (int x, double *a, double *b, double c)
+{
+  int i;
+
+  for (i = 0; i < x; i++)
+    *a++ = *b++ * i / c;
+}
+
+/* { dg-final { scan-assembler-not "vpaddd\[ \\t\]+\[^\n\]*%ymm\[0-9\]" } } */