llvm dev - Jun 2014 - [LLVMdev] Bug in LEA16r (X86InstrArithmetic.td) ??

Hi,

in X86InstrArithmetic.td, we have LEA16r defined as:


def LEA16r   : I<0x8D, MRMSrcMem,
                 (outs GR16:$dst), (ins i32mem:$src),
                 "lea{w}\t{$src|$dst}, {$dst|$src}", [],
IIC_LEA_16>,
OpSize16;


Please correct me if I am wrong, but I think "ins i32mem" should be
"ins
i16mem" because this is about 16bit register?

So is this a bug?


Thanks,
Jun
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20140624/4126f408/attachment.html>

Hi Jun,

On 24 June 2014 08:08, Jun Koi <junkoi2004 at gmail.com>
wrote:
> def LEA16r   : I<0x8D, MRMSrcMem,
>                  (outs GR16:$dst), (ins i32mem:$src),
>                  "lea{w}\t{$src|$dst}, {$dst|$src}", [],
IIC_LEA_16>,
> OpSize16;
>
>
> Please correct me if I am wrong, but I think "ins i32mem" should
be "ins
> i16mem" because this is about 16bit register?
I think there's something going on here, but it's probably more
complicated than that. The lea instruction just calculates an address,
so the resulting pointer doesn't really *have* a size.

Binutils seems to handle this reasonably by omitting the "dword
ptr"/"word ptr" part entirely when printing: "lea ax,
[eax]" instead
of "mov ax, word ptr [eax]". The assembler accepts all variants, which
may or may not be a good choice.

Cheers.

Tim.

On Tue, Jun 24, 2014 at 4:03 PM, Tim Northover <t.p.northover at
gmail.com>
wrote:
> Hi Jun,
>
> On 24 June 2014 08:08, Jun Koi <junkoi2004 at gmail.com> wrote:
> > def LEA16r   : I<0x8D, MRMSrcMem,
> >                  (outs GR16:$dst), (ins i32mem:$src),
> >                  "lea{w}\t{$src|$dst}, {$dst|$src}", [],
IIC_LEA_16>,
> > OpSize16;
> >
> >
> > Please correct me if I am wrong, but I think "ins i32mem"
should be "ins
> > i16mem" because this is about 16bit register?
>
> I think there's something going on here, but it's probably more
> complicated than that. The lea instruction just calculates an address,
> so the resulting pointer doesn't really *have* a size.
>
> Binutils seems to handle this reasonably by omitting the "dword
> ptr"/"word ptr" part entirely when printing: "lea ax,
[eax]" instead
> of "mov ax, word ptr [eax]". The assembler accepts all variants,
which
> may or may not be a good choice.
>
yes, i think this is a good idea. but then how can i modify LEA16r above to
eliminate "dword ptr" part?


thanks,
Jun
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20140624/a44d18ad/attachment.html>