On Mon, Feb 11, 2019 at 10:00:03AM +0100, Hans-Peter Nilsson wrote:
> > Date: Mon, 11 Feb 2019 07:38:14 +0100
> > From: Richard Biener <richard.guent...@gmail.com>
>
> > >+ HOST_WIDE_INT min_parm_align
> > >+ = STRICT_ALIGNMENT ? BITS_PER_WORD : PREFERRED_STACK_BOUNDARY;
> >
> > Shouldn't it be MIN (...) of BOTH?
>
> That *does* seem logical... Take 2 as follows, in testing as
> before.
>
> Ok to commit, if testing passes?
Is PREFERRED_STACK_BOUNDARY what we want here? Shouldn't that be
STACK_BOUNDARY, or PARM_BOUNDARY? Though, PARM_BOUNDARY on cris is 32...
> gcc:
> * function.c (assign_parm_setup_block): Align the
> parameter to a minimum of PREFERRED_STACK_BOUNDARY and
> BITS_PER_WORD instead of always BITS_PER_WORD.
>
> --- gcc/function.c.orig2 Sat Feb 9 00:53:17 2019
> +++ gcc/function.c Mon Feb 11 09:37:28 2019
> @@ -2912,7 +2912,10 @@ assign_parm_setup_block (struct assign_p
> size_stored = CEIL_ROUND (size, UNITS_PER_WORD);
> if (stack_parm == 0)
> {
> - SET_DECL_ALIGN (parm, MAX (DECL_ALIGN (parm), BITS_PER_WORD));
> + HOST_WIDE_INT min_parm_align
> + = MIN (BITS_PER_WORD, PREFERRED_STACK_BOUNDARY);
> +
> + SET_DECL_ALIGN (parm, MAX (DECL_ALIGN (parm), min_parm_align));
> if (DECL_ALIGN (parm) > MAX_SUPPORTED_STACK_ALIGNMENT)
> {
> rtx allocsize = gen_int_mode (size_stored, Pmode);
Jakub
