[Bug testsuite/102946] [12 Regression] gcc.dg/vect/pr101145_1.c etc. FAIL

2021-12-15 Thread guojiufu at gcc dot gnu.org via Gcc-bugs 
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102946

Jiu Fu Guo  changed:

   What|Removed |Added

 Resolution|--- |FIXED
 Status|UNCONFIRMED |RESOLVED

--- Comment #7 from Jiu Fu Guo  ---
In trunk cases updated to run on the target which supports the cases (e.g.
vec_char_add)

[Bug testsuite/102946] [12 Regression] gcc.dg/vect/pr101145_1.c etc. FAIL

2021-10-27 Thread guojiufu at gcc dot gnu.org via Gcc-bugs 
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102946

--- Comment #6 from Jiu Fu Guo  ---
Hi Rainer and Richard,
Thanks for working on this PR.

The intention of these test cases (pr101145*) is to test if the number 
of iterations can be calculated for the loop with the 'until wrap' 
condition.
So, I'm thinking we may be able to update the cases like:
-/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 2 "vect" } } */
+/* { dg-final { scan-tree-dump-times "Symbolic number of iterations is" 2
"vect" } } */

[Bug testsuite/102946] [12 Regression] gcc.dg/vect/pr101145_1.c etc. FAIL

2021-10-27 Thread ro at gcc dot gnu.org via Gcc-bugs 
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102946

Rainer Orth  changed:

   What|Removed |Added

  Attachment #51666|0   |1
is obsolete||

--- Comment #5 from Rainer Orth  ---
Created attachment 51673
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51673=edit
32-bit sparc-sun-solaris2.11 pr101145_1.c.170t.vect with a, b aligned

[Bug testsuite/102946] [12 Regression] gcc.dg/vect/pr101145_1.c etc. FAIL

2021-10-27 Thread ro at CeBiTec dot Uni-Bielefeld.DE via Gcc-bugs 
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102946

--- Comment #4 from ro at CeBiTec dot Uni-Bielefeld.DE  ---
> --- Comment #3 from Richard Biener  ---
> Very likely might be fixed with properly aligning the data like with adding
>
>  a = __builtin_assume_aligned (a, __BIGGEST_ALIGNMENT__);
>  b = __builtin_assume_aligned (b, __BIGGEST_ALIGNMENT__);
>
> at the top of each function in pr101145.inc - note the global variables
> would then need a similar aligned attribute for correctness.
>
> Rainer, if you can manage to test this and that works such change is
> pre-approved.

While this fixes the 64-bit SPARC case, the 32-bit failures remain.
pr101145_1.c dump attached.

[Bug testsuite/102946] [12 Regression] gcc.dg/vect/pr101145_1.c etc. FAIL

2021-10-27 Thread rguenth at gcc dot gnu.org via Gcc-bugs 
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102946

Richard Biener  changed:

   What|Removed |Added

Summary|gcc.dg/vect/pr101145_1.c|[12 Regression]
   |etc. FAIL   |gcc.dg/vect/pr101145_1.c
   ||etc. FAIL

--- Comment #3 from Richard Biener  ---
Very likely might be fixed with properly aligning the data like with adding

 a = __builtin_assume_aligned (a, __BIGGEST_ALIGNMENT__);
 b = __builtin_assume_aligned (b, __BIGGEST_ALIGNMENT__);

at the top of each function in pr101145.inc - note the global variables
would then need a similar aligned attribute for correctness.

Rainer, if you can manage to test this and that works such change is
pre-approved.