This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch to improve x86 integer -> FP conversions
- To: John Wehle <john at feith dot com>
- Subject: Re: Patch to improve x86 integer -> FP conversions
- From: Jeffrey A Law <law at upchuck dot cygnus dot com>
- Date: Fri, 02 Apr 1999 16:07:08 -0700
- cc: egcs-patches at egcs dot cygnus dot com
- Reply-To: law at cygnus dot com
In message <199904022136.QAA14050@jwlab.FEITH.COM>you write:
> > I think this resolved all the issues.
> >
> > Can you please check in the patch? Thanks!
>
> I hesitate to mention this given last time, however since posting the
> patch I've tweaked the floatdiXX constraints. :-)
>
> For example I've changed:
>
> [(set (match_operand:SF 0 "register_operand" "=f,f")
> (float:SF (match_operand:DI 1 "nonimmediate_operand" "m,!r")))
> (clobber (match_operand:DI 2 "memory_operand" "m,m"))])]
>
> to:
>
> [(set (match_operand:SF 0 "register_operand" "=f,f")
> (float:SF (match_operand:DI 1 "nonimmediate_operand" "m,!r")))
> (clobber (match_operand:DI 2 "memory_operand" "m,o"))])]
>
> And:
>
> [(parallel [(set (match_operand:SF 0 "register_operand" "=f")
> (float:SF (match_operand:DI 1 "register_operand" "r")))
> (clobber (match_operand:DI 2 "memory_operand" "m"))])]
>
> to:
>
> [(parallel [(set (match_operand:SF 0 "register_operand" "=f")
> (float:SF (match_operand:DI 1 "register_operand" "r")))
> (clobber (match_operand:DI 2 "memory_operand" "o"))])]
>
> Is this okay?
Should be fine. Presumably you did this for a reason though? :-) "o" is
more restrictive than "m", so we'd prefer to use "m" unless there's a reason
we can't.
jeff