assign_parm_setup_block handling of parallels

[Richard Henderson]

On Thu, Dec 23, 2004 at 05:45:43PM -0400, Aldy Hernandez wrote:
> In my limited understanding of the problem I think we can allow the
> optimization for one-member parallels iff the hard register will span
> more than one register.  I think this will satisfy both PPC64 and
> Sparc64.

I can't decide whether the better check is

	nelts > 0
	|| hard_regno_nregs[REGNO (elt0)][GET_MODE (elt0)] > 1

or 

	|| GET_MODE_SIZE (parallel) == GET_MODE_SIZE (elt0)

or

	|| hard_regno_nregs[REGNO (elt0)][GET_MODE (elt0)]
	   != hard_regno_nregs[REGNO (elt0)][GET_MODE (parallel)]

All of these would avoid the Sparc problem.

But at that point, I wonder if we're trying to get too clever,
and the most correct solution is simply a target macro.



r~