gen_highpart() & volatile

[Dmitry]

Thanks Richard,

Will this be really bad if I mark an operand as non-volatile befor 
gen_highpart() and then restore its state?

Cheers,
~d 

On Tuesday 18 February 2003 21:25, Richard Henderson wrote:
> On Tue, Feb 18, 2003 at 07:19:42PM +0300, Dmitry wrote:
> > why gen_highpart(HImode, op); where op is mem:SI and declared as volatile
> > fails to produce a correct subreg?
>
> Because we're not allowed to change the access width of
> a volatile memory.
>
>
> r~