What's the reasoning behind using ld64 for gccXX?

[vincent habchi]

On 19 avr. 2012, at 22:33, Jack Howarth <howarth at bromo.med.uc.edu> wrote:

> On Thu, Apr 19, 2012 at 10:11:12PM +0200, vincent habchi wrote:
>> On 19 avr. 2012, at 22:07, Jack Howarth <howarth at bromo.med.uc.edu> wrote:
>> 
>>> On Thu, Apr 19, 2012 at 09:47:13PM +0200, vincent habchi wrote:
>>>> Folks,
>>>> 
>>>> What about the upcoming LLD?
>>>> 
>>>> Vincent
>>> If lld is the proposed replacement for the existing ld linker tool, what is proposed as the replacement assmebler for compilers outside of clang?
>> 
>> llvm-mc?
> 
> But will llvm-mc pretend to be gas (like clang pretends to be gcc) in terms of understanding option flags and such so
> that it can be a drop in replacement for the current 'as'?

Frankly, I have no idea. At this point of time, I’d say no. I have tried to build a kind of "combo" with gcc-mp-4.6 as frontend and clang -x assembler-with-gcc as backend, which is essentially the same as llvm-mc. Mostly works, but gcc emits some heterodox instructions (especially AVX) that llvm-mc rejects as non-Intel compliant. 

Yet,