[Bug tree-optimization/59591] New: ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[Bug tree-optimization/59591] New: ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[izamyatin at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

            Bug ID: 59591
           Summary: ICE in vect_get_vec_def_for_stmt_copy, at
                    tree-vect-stmts.c:156 for -march=core-avx2
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: izamyatin at gmail dot com
            Target: x86

Started after r206069 and reproduced on 481.wrf from spec2006

Reduced testcase attached

Options for reproducing 
gfortran -O2 -ftree-vectorize  -march=core-avx2 -c

[Bug tree-optimization/59591] ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[izamyatin at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

--- Comment #1 from Igor Zamyatin <izamyatin at gmail dot com> ---
Created attachment 31509
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=31509&action=edit
small testcase

[Bug tree-optimization/59591] [4.9 Regression] ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.9.0
            Summary|ICE in                      |[4.9 Regression] ICE in
                   |vect_get_vec_def_for_stmt_c |vect_get_vec_def_for_stmt_c
                   |opy, at                     |opy, at
                   |tree-vect-stmts.c:156 for   |tree-vect-stmts.c:156 for
                   |-march=core-avx2            |-march=core-avx2

[Bug tree-optimization/59591] [4.9 Regression] ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-12-24
     Ever confirmed|0                           |1

[Bug tree-optimization/59591] [4.9 Regression] ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Slightly reduced fortran testcase for -O3 -mavx2:
subroutine foo (m, n, o, p, q, e, a, b, x, f)
  integer, intent (in) :: n, o, p, q, m, f
  integer, dimension (n:o) :: e, a, b
  integer, dimension (n:o, p:q) :: x
  integer :: j, i
  do j = 1, m
    do i = n, f
      if (e(i) .eq. 0) then
        a(i) = x(i, b(i))
      endif
    enddo
  enddo
end subroutine foo

and C testcase for -O3 -mavx2 -fopenmp-simd:
int p[256], q[256], r[256], t[256];

void
foo (void)
{
  int i;
  #pragma omp simd safelen(64)
  for (i = 0; i < 256; i++)
    if (r[i] > 32)
      t[i] = p[q[i] * 3L + 2L];
}

[Bug tree-optimization/59591] [4.9 Regression] ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 31525
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=31525&action=edit
gcc49-pr59591.patch

Untested fix.  Still need to turn the testcase into runtime testcase and look
at the other gather direction (the patch fixes the modifier = NARROW case, need
to look at modifier = WIDEN case).

[Bug tree-optimization/59591] [4.9 Regression] ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #31525|0                           |1
        is obsolete|                            |

--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 31540
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=31540&action=edit
gcc49-pr59591.patch

Updated patch with runtime testcases.  Unfortunately, I don't have access to
any CPUs with AVX2 right now, so can't test whether it works there properly.

[Bug tree-optimization/59591] [4.9 Regression] ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Mon Dec 30 17:05:10 2013
New Revision: 206248

URL: http://gcc.gnu.org/viewcvs?rev=206248&root=gcc&view=rev
Log:
    PR tree-optimization/59591
    * tree-vect-stmts.c (vectorizable_mask_load_store): Fix up handling
    of modifier = NARROW masked gathers.
    (permute_vec_elements): Use gimple_get_lhs instead of
    gimple_assign_lhs.

    * gcc.dg/vect/pr59591-1.c: New test.
    * gcc.dg/vect/pr59591-2.c: New test.
    * gcc.target/i386/pr59591-1.c: New test.
    * gcc.target/i386/pr59591-2.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/vect/pr59591-1.c
    trunk/gcc/testsuite/gcc.dg/vect/pr59591-2.c
    trunk/gcc/testsuite/gcc.target/i386/pr59591-1.c
    trunk/gcc/testsuite/gcc.target/i386/pr59591-2.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-vect-stmts.c

[Bug tree-optimization/59591] [4.9 Regression] ICE in vect_get_vec_def_for_stmt_copy, at tree-vect-stmts.c:156 for -march=core-avx2

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59591

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Fixed.