Messages in this thread | | | Date | Mon, 26 Oct 1998 09:53:56 -0800 (PST) | From | Linus Torvalds <> | Subject | Re: 2.2.0 and egcs 1.1 was Re: Sorry, wrong gcc-version |
| |
On Sun, 25 Oct 1998, David S. Miller wrote: > > For example, everybody in the egcs camp just decided that clobbers > and inputs must not overlap. Nobody told me why, and why they can't > just be automatically converted to early-clobbers inside gcc. > > You're telling the compiler two different things which conflict.
Which has worked before (and was certainly the _only_ way to do it in gcc-1.40 which I started with).
And which is something so obvious to the compiler that it really shouldn't generate bad code, it should just do the obvious thing. It's not as if the concept is ambiguous or anything like that.
If gcc internally doesn't want to handle the construct, then it should be truly trivial to make gcc turn the clobber into a early-clobber, which according to everybody _should_ work.
The reason I do not agree with the gcc people about having to use early-clobbers is that gcc also has limits on how many register specifiers you can use. At some point it was just seven, and I think somebody upped it to ten.
And doing early-clobbers by hand overflows that very very quickly, because it essentially means you have to double all your input specifiers.
Linus
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |