mips SNaN/QNaN is swapped

Richard Henderson rth@redhat.com
Tue Apr 1 17:45:00 GMT 2003


On Tue, Apr 01, 2003 at 11:13:46AM -0300, Alexandre Oliva wrote:
> @@ -2569,6 +2574,10 @@ real_hash (r)
>        /* FALLTHRU */
>  
>      case rvc_nan:
> +      if (r->signalling)
> +	h ^= (unsigned int)-1;
> +      if (r->canonical)
> +	break;
>        if (sizeof(unsigned long) > sizeof(unsigned int))
>  	for (i = 0; i < SIGSZ; ++i)

Fallthru from rvc_normal needs to be fixed before testing these.

Otherwise ok.


r~



More information about the Gcc-patches mailing list