[PATCH] MIPS: 32-bit microMIPS DSP ASE support

[Richard Sandiford]

"Maciej W. Rozycki" <macro@codesourcery.com> writes:
> Hi Richard,
>
>> > @@ -6397,10 +6398,15 @@ macro (struct mips_cl_insn *ip)
>> >  	case 2:
>> >  	  macro_build (NULL, "packrl.ph", "d,s,t", treg, treg, sreg);
>> >  	  break;
>> > -	default:
>> > +	case 1:
>> > +	case 3:
>> >  	  macro_build (NULL, "balign", "t,s,2", treg, sreg,
>> >  		       (int) imm_expr.X_add_number);
>> >  	  break;
>> > +	default:
>> > +	  as_bad (_("BALIGN immediate not 0,1,2,3 (%lu)"),
>> > +		  (unsigned long) imm_expr.X_add_number);
>> 
>> "0, 1, 2 or 3" reads better to me and ties in well with:
>> 
>> 		  as_bad (_("BALIGN immediate not 1 or 3 (%lu)")
>> 
>> > @@ -10728,8 +10744,7 @@ mips_ip (char *str, struct mips_cl_insn 
>> >  		return;
>> >  	      break;
>> >  
>> > -	    case '2': /* DSP 2-bit unsigned immediate in bit 11.  */
>> > -	      gas_assert (!mips_opts.micromips);
>> > +	    case '2': /* DSP 2-bit unsigned immediate in bit 11 or 14.  */
>> 
>> "bit 11 (for normal MIPS) or bit 14 (for microMIPS)".  Same for the others.
>> 
>> OK with those changes, thanks.
>
>  Resending just to be sure this is still all right with you -- here's a 
> new version with your suggestions applied, except that I used "for 
> standard MIPS code" and "for microMIPS code" for consistency with the 
> wording used elsewhere and per the earlier discussion re nouns vs 
> adjectives.
>
>  Another update is the drop of the "l" formatting output modifier 
> throughout print_insn_micromips reflecting the change elsewhere to the 
> type used.  No other changes.
>
>  Still no regressions for mips-sde-elf/mips-linux-gnu and sorry it took so 
> long.  OK to apply?

OK, thanks.

Richard