This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: MS/CW-style inline assembly for GCC


Richard Henderson wrote:

On Mon, May 03, 2004 at 06:32:17PM -0700, Mike Stump wrote:

Would you be willing entertain letting it in and let Apple people maintain it?


We can certainly arrange for the actual asm parsing bits to be cleanly separated. Indeed, if pressed I could pull out the MSVC stuff from Cygnus' second implementation and you could re-use that. The parsing bits were the only thing I was moderately happy about in the result of that project.

HOWEVER, I disbelieve that will be the end of the line.  I predict
that random little turds will appear all over the optimizers, and
this prospect does not please me at all.  This eventuallity means
that we /can't/ just let Apple people maintain it; it becomes
everyone's problem.

I agree that we don't want to go there. If the functionality is limited
to what GCC asm statements already provide, that would hopefully limit
the entropy growth rate. :-)

If you can convince me that my MSVC experience (see response to Stan)
won't apply to CW, I might be ameanable.  But didn't Stan say yall
want MSVC support for darwin/x86 too?  You'd have to come up with a
fantastically better implementation than we did at Cygnus for me to
want to let that in.

Ironically, I just threw out the MSVC suggestion as a way to knock off two
user-requested features at once, and to keep the addition from being too
CodeWarrior/Mac-centric. I would be perfectly happy to do a CW-only
version if everybody would prefer that; I don't personally know of any
requests for MS-style asm on x86 darwin systems, so that seems too
hypothetical to drive requirements.

Stan


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]