Patch for h8300 port
graham stott
grahams@redhat.com
Mon Oct 9 08:13:00 GMT 2000
Will
Will Cohen wrote:
>
> I have made some revisions in the patch based on your comments.
>
> I found that the pattern generated for the sign extend and
> zero extend may have memory references. If the predicates
> were registers, then the compiler would fail when building
> the libraries. I have made the predicates general_operand
> and then use the constraints to get the values in the
> correct registers. Is this a reasonable approach?
>
> -Will
>
> Richard Henderson wrote:
>
> > You've reduced the number of alternatives to 1, which
> > means that you should no longer use the @ format for
> > the output template. So just write "exts.w %T0".
> >
> > The input operand has a matching constraint on the
> > output operand. That implies that the input operand
> > should be using the same predicate. I.e. use
> >
> > (sign_extend:HI (match_operand:QI 1 "register_operand" "0"))
> >
> > r~
>
>
I think you should use "nonimmediate_operand" for output operands, don't
forget general_operand allows almost anything. Didn't you get lots of new
"warning: destination operand 0 allows non-lvalue" warnings from genrecog.
Graham
More information about the Gcc-patches
mailing list