optimization/1945 fix

[Richard Henderson]

On Thu, Feb 15, 2001 at 05:07:19AM -0800, Shane Nay wrote:
> +         /* We need to check to make sure that we should be
> +            emitting shift code, because if the argument isn't
> +            getting used, it is considered a bug to emit
> +            preparation code for it in mips_expand_prologue.
> +         */

Ug.  I don't think you want to be scanning the entire function.
I'd have said check "BASIC_BLOCK (0)->global_live_at_start", but
just remembered that that isn't valid at this point.

As much as I dislike the thing, I think you'll have to add a
REG_MAYBE_UNUSED note to these shift instructions.


r~