[PATCH] [AArch64] Add add/sub/cmp/cmn aliases to MC AsmParser

Jim Grosbach grosbach at apple.com
Fri Jul 11 13:34:11 PDT 2014


> On Jul 11, 2014, at 1:03 PM, Renato Golin <renato.golin at linaro.org> wrote:
> 
> On 11 July 2014 20:52, Jim Grosbach <grosbach at apple.com> wrote:
>> I have the same question about a number of the other proposed patches of this nature. If the docs say we should support them, then I’m all for it. If they don’t, I’m very hesitant to do so. I don’t like adding undocumented extensions to the assembler syntax.
> 
> I see, makes sense.
> 
> It is possible that GCC already does that and maybe that's the reason
> for the patch? (Arnaud?)
> 
> I don't have any interest in that kind of patch myself, unless GCC is
> generating them and we need to consume, which would make a bit of
> sense to add as a GNU extension.

That’s where the discussion gets interesting. I’d be more inclined to tell the GCC folks to stop generating bogus assembly. Not sure how far that’d get, though.

> About documentation, I don't think it's realistic to expect GNU folks
> from documenting every extension in the assembler, as they have been
> specifically lenient in that same area... :/

Yeah, I’m more curious about the official ARM docs.

> Back to the patch, the only explicit aliasing rules I've seen is
> between CMP/CMN and ADDS/SUBS, but that doesn't automatically allow us
> to convert negative immediate in the reverse operation.

Agreed.

I don’t have any strong philosophical objection to any of this in the abstract. If these are useful aliases, we should work with the docs folks at ARM to get them added (assuming they’re not already there).

-Jim



More information about the llvm-commits mailing list