[PATCH] Avoid shifts using hard regs from likely spilled classes for BLKmode returning calls (PR rtl-optimization/41646)

Richard Guenther richard.guenther@gmail.com
Fri Oct 9 19:02:00 GMT 2009


On Fri, Oct 9, 2009 at 8:40 PM, Jakub Jelinek <jakub@redhat.com> wrote:
> Hi!
>
> This testcase ICEs on x86_64-linux, because in expand copy_blkmode_from_reg
> is called with %rax and thus a hard register is used in several shift insns.
> Later on combiner extends lifetime of that register across multiplication
> insns that require this register during reload and reload ICEs.
>
> Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux.
> Ok for trunk?

Ok.

Thanks,
Richard.

> 2009-10-09  Jakub Jelinek  <jakub@redhat.com>
>
>        PR rtl-optimization/41646
>        * calls.c (expand_call): For BLKmode types returned in registers
>        avoid likely spilled hard regs in copy_blkmode_from_reg generated
>        insns.
>
>        * gcc.c-torture/compile/pr41646.c: New test.
>
> --- gcc/calls.c.jj      2009-09-23 00:57:18.000000000 +0200
> +++ gcc/calls.c 2009-10-09 17:25:32.000000000 +0200
> @@ -3014,7 +3014,10 @@ expand_call (tree exp, rtx target, int i
>        }
>       else if (TYPE_MODE (rettype) == BLKmode)
>        {
> -         target = copy_blkmode_from_reg (target, valreg, rettype);
> +         rtx val = valreg;
> +         if (GET_MODE (val) != BLKmode)
> +           val = avoid_likely_spilled_reg (val);
> +         target = copy_blkmode_from_reg (target, val, rettype);
>
>          /* We can not support sibling calls for this case.  */
>          sibcall_failure = 1;
> --- gcc/testsuite/gcc.c-torture/compile/pr41646.c.jj    2009-10-09 17:29:51.000000000 +0200
> +++ gcc/testsuite/gcc.c-torture/compile/pr41646.c       2009-10-09 17:30:17.000000000 +0200
> @@ -0,0 +1,28 @@
> +/* PR rtl-optimization/41646 */
> +
> +struct A { unsigned long a; };
> +struct B { unsigned short b, c, d; };
> +struct B bar (unsigned long);
> +
> +char *
> +foo (char *a, struct A *x)
> +{
> +  struct B b = bar (x->a);
> +  unsigned char c;
> +  unsigned short d;
> +  a[3] = ((unsigned char) (b.b % 10) + 48);
> +  d = b.b / 10;
> +  a[2] = ((unsigned char) (d % 10) + 48);
> +  d = d / 10;
> +  a[1] = ((unsigned char) (d % 10) + 48);
> +  a[0] = ((unsigned char) ((d / 10) % 10) + 48);
> +  a[4] = 46;
> +  c = (unsigned char) b.c;
> +  a[6] = (c % 10 + 48);
> +  a[5] = ((c / 10) % 10 + 48);
> +  a[7] = 46;
> +  c = b.d;
> +  a[9] = (c % 10 + 48);
> +  a[8] = ((c / 10) % 10 + 48);
> +  return a + 10;
> +}
>
>        Jakub
>



More information about the Gcc-patches mailing list