This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH: [4.4/4.5 Regression] -fschedule-insns causes FAIL: gcc.target/i386/vararg-1.c execution test
> How about this patch?
Well, I am still not happy for mainline about this because it increases
alignment for no real need, but it is OK for release branches.
The reason why we did not hit this earlier is that the va_list variable
itself bumps up stack alignment needs to 128 and thus we (almost) always
have variadic functions requiring 128bit stack alignment.
This is bug, I am testing more involved patch for mainline.
Patch is OK for all release branches.
Honza
>
>
> --
> H.J.
> gcc/
>
> 2010-04-06 H.J. Lu <hongjiu.lu@intel.com>
>
> PR target/43668
> * config/i386/i386.c (setup_incoming_varargs_64): Align stack to
> 16byte for FP register save area.
>
> gcc/testsuite/
>
> 2010-04-06 H.J. Lu <hongjiu.lu@intel.com>
>
> PR target/43668
> * gcc.target/i386/pr43668.c: New.
>
> diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
> index c945a54..5176eb2 100644
> --- a/gcc/config/i386/i386.c
> +++ b/gcc/config/i386/i386.c
> @@ -6785,6 +6785,10 @@ setup_incoming_varargs_64 (CUMULATIVE_ARGS *cum)
>
> if (ix86_varargs_fpr_size)
> {
> + /* Stack must be aligned to 16byte for FP register save area. */
> + if (crtl->stack_alignment_needed < 128)
> + crtl->stack_alignment_needed = 128;
> +
> /* Now emit code to save SSE registers. The AX parameter contains number
> of SSE parameter registers used to call this function. We use
> sse_prologue_save insn template that produces computed jump across
> --- /dev/null 2010-03-31 15:02:03.585768648 -0700
> +++ gcc/gcc/testsuite/gcc.target/i386/pr43668.c 2010-04-06 22:50:40.000000000 -0700
> @@ -0,0 +1,10 @@
> +/* PR target/43668 */
> +/* { dg-do run } */
> +/* { dg-options "-fschedule-insns" } */
> +
> +int foo(int i, ...) {
> + return i;
> +}
> +int main() {
> + return foo(0, 0.0);
> +}