Re: [patch, testsuite] Fix vectorizer testcase

Re: [patch, testsuite] Fix vectorizer testcase

[Dorit Nuzman]

Revital1 Eres/Haifa/IBM wrote on 30/10/2007 00:13:40:

> OK for mainline?
>

it actually requires vect_long rather than vect_int.
I tested this fix on powerpc64-linux with -m32 and -m64.
A fix to require vect_long is pre-approved.

dorit

> Thanks,
> Revital
>
> * testsuite/gcc.dg/vect/pr33866.c: Require vect_int.
>
> Index: testsuite/gcc.dg/vect/pr33866.c
> ===================================================================
> --- testsuite/gcc.dg/vect/pr33866.c     (revision 129765)
> +++ testsuite/gcc.dg/vect/pr33866.c     (working copy)
> @@ -1,5 +1,6 @@
>  /* Testcase by Martin Michlmayr <tbm@cyrius.com> */
>  /* { dg-do compile } */
> +/* { dg-require-effective-target vect_int } */
>
>  typedef struct
>  {

Re: [patch, testsuite] Fix vectorizer testcase

[Revital1 Eres]

Committed to r129767.

Thanks,
Revital

Dorit Nuzman/Haifa/IBM wrote on 30/10/2007 10:14:20:

> Revital1 Eres/Haifa/IBM wrote on 30/10/2007 00:13:40:
>
> > OK for mainline?
> >
>
> it actually requires vect_long rather than vect_int.
> I tested this fix on powerpc64-linux with -m32 and -m64.
> A fix to require vect_long is pre-approved.
>
> dorit
>
> > Thanks,
> > Revital
> >
> > * testsuite/gcc.dg/vect/pr33866.c: Require vect_int.
> >
> > Index: testsuite/gcc.dg/vect/pr33866.c
> > ===================================================================
> > --- testsuite/gcc.dg/vect/pr33866.c     (revision 129765)
> > +++ testsuite/gcc.dg/vect/pr33866.c     (working copy)
> > @@ -1,5 +1,6 @@
> >  /* Testcase by Martin Michlmayr <tbm@cyrius.com> */
> >  /* { dg-do compile } */
> > +/* { dg-require-effective-target vect_int } */
> >
> >  typedef struct
> >  {

[patch, testsuite] Fix vectorizer testcase

[Revital1 Eres]

OK for mainline?

Thanks,
Revital

* testsuite/gcc.dg/vect/pr33866.c: Require vect_int.

Index: testsuite/gcc.dg/vect/pr33866.c
===================================================================
--- testsuite/gcc.dg/vect/pr33866.c     (revision 129765)
+++ testsuite/gcc.dg/vect/pr33866.c     (working copy)
@@ -1,5 +1,6 @@
 /* Testcase by Martin Michlmayr <tbm@cyrius.com> */
 /* { dg-do compile } */
+/* { dg-require-effective-target vect_int } */

 typedef struct
 {

[patch, testsuite] Fix vectorizer testcase

[Ira Rosen]

In vect-strided-store-u16-i4.c there was a conditional call to abort() in
initialization loop  in order to avoid its vectorization. However, now the
call is optimized out before the vectorizer and the loop gets vectorized
anyway.  I removed the call to abort() and fixed dg-final checks.

Tested on ppc-linux. Committed as obvious.

Ira

ChangeLog:

       * gcc.dg/vect/vect-strided-store-u16-i4.c: Remove abort() from input
      initialization and fix dg-final check - the initialization loop is
      vectorizable.

Index: testsuite/gcc.dg/vect/vect-strided-store-u16-i4.c
===================================================================
--- testsuite/gcc.dg/vect/vect-strided-store-u16-i4.c   (revision 126162)
+++ testsuite/gcc.dg/vect/vect-strided-store-u16-i4.c   (working copy)
@@ -58,8 +58,6 @@ int main (void)
       a[i] = i;
       b[i] = i * 2;
       c[i] = 17;
-      if (a[i] == 178)
-         abort();
     }

   main1 (arr);
@@ -67,7 +65,8 @@ int main (void)
   return 0;
 }

-/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect"  {
target { vect_interleave } } } } */
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 2 "vect"  {
target { vect_interleave } } } } */
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect"  {
target { ! { vect_interleave } } } } } */
 /* { dg-final { cleanup-tree-dump "vect" } } */