Re: define_constraints patch, re-revised

[Bernd Schmidt <bernds_cb1 at t-online dot de>]

Zack Weinberg wrote:
> :ADDPATCH machine description:

:REVIEWMAIL:

> +If you put the pseudo-Texinfo directive @samp{@@internal} at the
> +beginning of a docstring, then (in the future) it will appear only in
> +the internals manual's version of the machine-specific constraint
> +tables.  Use this for constraints that should not appear in @code{asm}
> +statements.  (The compiler itself is unaware of the distinction; this
> +is not a way to give users a helpful diagnostic instead of a
> +@code{reload} failure.)

Can we say "The compiler may in the future use this to give a helpful 
diagnostic instead of a reload failure"?

> +#if 0
> +{
> +	"  enum machine_mode mode = GET_MODE (op);\n"
> +	"  HOST_WIDE_INT ival = 0;\n"
> +	"  HOST_WIDE_INT hval = 0;\n"
> +	"  unsigned HOST_WIDE_INT lval = 0;\n"
> +	"  const REAL_VALUE_TYPE *rval = 0;\n"
> +	"\n"
> +	"  if (GET_CODE (op) == CONST_INT)\n"
> +	"    ival = INTVAL (op);\n"
> +	"  else if (GET_CODE (op) == CONST_DOUBLE)\n"
> +	"    {\n"
> +	"      if (mode == VOIDmode)\n"
> +	"        hval = CONST_DOUBLE_HIGH (op),\n"
> +	"        lval = CONST_DOUBLE_LOW  (op);\n"
> +	"      else\n"
> +	"        rval = CONST_DOUBLE_REAL_VALUE (op);\n"
> +	"    }\n"
> +	"\n"
> +      if (c->is_register)	
> +	printf ("reg_fits_class_p (op, %s, 0, mode)", c->regclass);
> +      else
> +	write_predicate_expr (c->exp);
> +      fputs (";\n\n", stdout);
> +	  ;
> +}
> +#endif

Lose this.

Otherwise ok, minus the h8300 changes - I'd prefer you get buyin and 
maybe testing from whoever is maintaining it.

Thanks!


Bernd