Re: [PATCH] x86: allow 1- and 0-operand forms of fcomip and fucomip

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Thu, Aug 28, 2008 at 10:39:36AM +0100, Jan Beulich wrote:
> >>> "H.J. Lu" <hjl.tools@gmail.com> 27.08.08 18:35 >>>
> >On Wed, Aug 27, 2008 at 9:29 AM, Jan Beulich <jbeulich@novell.com> wrote:
> >>>I really don't like those x87 gas extensions. It makes it harder to look them
> >>>up in IA32/Intel64 SDM. I don't think we should add more unless we can't
> >>>live without them.
> >>
> >> So ease of porting code from other assemblers doesn't count? Also, other
> >
> >Can you name the assemblers which support this extension? Porting
> 
> Borland's TASM (which I happen to prefer of MASM).
> 
> >of assembly code is nice, but very low priority. For this particular extension,
> >I think assembly code should be changed according to IA32/Intel64 SDM.
> 
> So this basically means that I'll have to maintain the patch as a private
> one infinitely.

I want to keep those extensions as few as possible. I don't think
we should add a new one for TASM.

> 
> >BTW, I have some MASM assembly codes I'd like to assemble with gas.
> 
> As you know you have to adjust it. But ideally you'd only have to adjust
> the program structure (i.e. directives), but not any instructions. If that
> doesn't work then I'd have to look into it.

They are many directives in those MASM assembly codes. I gave them
up at the first try.

> 
> Btw., I just finished implementing a less ad-hoc Intel parser that I intend
> to commit in place of the current one once 2.19 got released. It needs
> some new hooks in the general expression parser, but otherwise
> completely uses that one rather than doing anything custom. Plus it's
> smaller and (I believe) gets closer to MASM than the current one.

That is very nice. We are trying to add a complete testsuite for Intel
syntax. But it may take a while.

Thanks.


H.J.