MS/CW-style inline assembly for GCC

[Stan Shebs]

Mark Mitchell wrote:

> Matt Austern wrote:
>
>> If you can come up with an assembly syntax that's genuinely better 
>> than CW's, and if we can convincingly tell our developers that it's 
>> better, then we really will consider telling our developers they 
>> should convert their code.
>
>
> Oh, throwing down the guantlet! :-)
>
> I'm in Palm Springs next week, I'll see if anything springs to mind -- 
> but I don't really expect to do better that CW.  I do understand why 
> people like that more than GCC's syntax.
>
Yep, it's not a take-it-or-leave-it situation - there will almost
certainly be points of CW compatibility that are too fragrant for
FSF GCC under any circumstances, but that Apple needs, and those
would remain local as usual. A good analogy is AltiVec, which has a
generic implementation built on on even more generic vector machinery,
but for which we have Apple-specific changes as well.

> To do CW-syntax (in an ideal) world, you'd want the assembler 
> integrated into the compiler.  It would then parse the assembly code, 
> emitting actual RTL insns into the instruction stream.  Right?  (That, 
> presumably, is how CW actually inserts code in and amongst these 
> assembly instructions; it just puts them into the instruction stream 
> and then happily schedules them along with everything else.)  Whereas 
> Apple's current implementation produces one big asm string for GCC, 
> annoated with constraints obtained by parsing the assembly code a 
> bit.  Right?

Right, although I make a separate asm statement for each line of
assembly (if nothing else, a single asm statement would have a
rather large number of inputs, outputs, and clobbers, could cause
reload problems).

Stan