This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix up ix86_expand_set_or_movmem (PR target/61060)
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: Jakub Jelinek <jakub at redhat dot com>
- Cc: Jan Hubicka <hubicka at ucw dot cz>, Uros Bizjak <ubizjak at gmail dot com>, gcc-patches at gcc dot gnu dot org
- Date: Tue, 13 May 2014 09:13:23 +0200
- Subject: Re: [PATCH] Fix up ix86_expand_set_or_movmem (PR target/61060)
- Authentication-results: sourceware.org; auth=none
- References: <20140512194754 dot GT10386 at tucnak dot redhat dot com>
> Hi!
>
> As discussed in the PR, ix86_expand_set_or_movmem assumes that count == 0
> means !CONST_INT_P (count_exp), but that is not necessarily true with
> -O0 -ftree-ter, when the middle-end doesn't immediately see the length
> argument be integer_zerop, yet TER results in it being expanded as
> const0_rtx.
>
> There is nothing to set or move in that case, so this patch makes the
> backend more tolerand just by returning true immediately.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
OK,
Honza
>
> 2014-05-12 Jakub Jelinek <jakub@redhat.com>
>
> PR target/61060
> * config/i386/i386.c (ix86_expand_set_or_movmem): If count_exp
> is const0_rtx, return immediately. Don't test count == 0 when
> it is always true.
>
> * gcc.dg/pr61060.c: New test.
>
> --- gcc/config/i386/i386.c.jj 2014-05-11 22:21:28.000000000 +0200
> +++ gcc/config/i386/i386.c 2014-05-12 10:59:18.940416162 +0200
> @@ -24156,8 +24156,13 @@ ix86_expand_set_or_movmem (rtx dst, rtx
> align = MEM_ALIGN (dst) / BITS_PER_UNIT;
>
> if (CONST_INT_P (count_exp))
> - min_size = max_size = probable_max_size = count = expected_size
> - = INTVAL (count_exp);
> + {
> + min_size = max_size = probable_max_size = count = expected_size
> + = INTVAL (count_exp);
> + /* When COUNT is 0, there is nothing to do. */
> + if (!count)
> + return true;
> + }
> else
> {
> if (min_size_exp)
> @@ -24166,7 +24171,7 @@ ix86_expand_set_or_movmem (rtx dst, rtx
> max_size = INTVAL (max_size_exp);
> if (probable_max_size_exp)
> probable_max_size = INTVAL (probable_max_size_exp);
> - if (CONST_INT_P (expected_size_exp) && count == 0)
> + if (CONST_INT_P (expected_size_exp))
> expected_size = INTVAL (expected_size_exp);
> }
>
> --- gcc/testsuite/gcc.dg/pr61060.c.jj 2014-05-12 11:03:10.797222337 +0200
> +++ gcc/testsuite/gcc.dg/pr61060.c 2014-05-12 11:03:02.882263281 +0200
> @@ -0,0 +1,19 @@
> +/* PR target/61060 */
> +/* { dg-do compile } */
> +/* { dg-options "-O0 -ftree-ter" } */
> +
> +typedef __SIZE_TYPE__ size_t;
> +
> +extern inline __attribute__ ((gnu_inline, always_inline, artificial))
> +void *memset (void *dest, int ch, size_t len)
> +{
> + return __builtin_memset (dest, ch, len);
> +}
> +
> +char buf[10];
> +
> +void
> +foo (void)
> +{
> + memset (buf, sizeof (buf), 0);
> +}
>
> Jakub