This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix vector handling in simplify-rtx.c (PR rtl-optimization/82973)
- From: Jeff Law <law at redhat dot com>
- To: Jakub Jelinek <jakub at redhat dot com>, Eric Botcazou <ebotcazou at adacore dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Wed, 20 Dec 2017 22:28:53 -0700
- Subject: Re: [PATCH] Fix vector handling in simplify-rtx.c (PR rtl-optimization/82973)
- Authentication-results: sourceware.org; auth=none
- References: <20171220221850.GR2353@tucnak>
On 12/20/2017 03:18 PM, Jakub Jelinek wrote:
> Hi!
>
> In rtl.texi we say:
> @findex const_vector
> @item (const_vector:@var{m} [@var{x0} @var{x1} @dots{}])
> Represents a vector constant. The square brackets stand for the vector
> containing the constant elements. @var{x0}, @var{x1} and so on are
> the @code{const_int}, @code{const_wide_int}, @code{const_double} or
> @code{const_fixed} elements.
> and it is a very reasonable requirement.
> simplify_const_{unary,binary}_operation can violate that though, because
> the recursion can return also other expressions, like in this case
> a (mult (const_double) (const_double)) that can't be simplified because
> of -frounding-math. We need to punt on those.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
>
> 2017-12-20 Jakub Jelinek <jakub@redhat.com>
>
> PR rtl-optimization/82973
> * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
> CONST_VECTOR if some element isn't simplified into CONST_{,WIDE_}INT,
> CONST_DOUBLE or CONST_FIXED.
> (simplify_const_binary_operation): Likewise. Use CONST_FIXED_P macro
> instead of GET_CODE == CONST_FIXED.
> (simplify_subreg): Use CONST_FIXED_P macro instead of
> GET_CODE == CONST_FIXED.
>
> * gfortran.dg/pr82973.f90: New test.
OK.
jeff