Re: egcs 1.0.1 miscompiles Linux 2.0.33

Jeffrey A Law (law@hurl.cygnus.com)
Fri, 27 Feb 1998 19:05:13 -0700


In message <199802270947.CAA20510@nyx10.nyx.net>you write:
> Regarding the documented behaviour of GCC w.r.t. clobbered registers,
> Jeffrey A. Law wrote:
> > The documentation is a little ambigious, but the natural
> > meaning of a clobber would be that it can't be used for an input
> > operand *or* in the address of an input operand.
>
> Actually, it's not ambiguous at all. From the 2.7.2 docs (which have not,
> I understand, changed this section for 2.8):
Read it more carefully.

Note that it says the clobber won't be used for an input operand,
or an output operand's address. Note that it never mentions an
input operand address.

Though it does mention you can read/write clobbers as many times as
you like. Which *implies* that it can't be used for an input address.

But I don't think we need to rehash all of this again. The natural
way to think of a clobber is that it's completely unavailable for use
by the compiler inside the asm. My job for this weekend is to see
if it's possible to make gcc "do the right thing".

jeff

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu