Re: [PATCH 7/8] i386: bitops: Kill needless usage of __asm__ __volatile__

[Satyam Sharma <ssatyam@cse.iitk.ac.in>]

Hi Jeremy,


On Mon, 23 Jul 2007, Jeremy Fitzhardinge wrote:

> Satyam Sharma wrote:
> > From: Satyam Sharma <ssatyam@cse.iitk.ac.in>
> >
> > [7/8] i386: bitops: Kill needless usage of __asm__ __volatile__
> >
> > Another oddity I noticed in this file. The semantics of __volatile__
> > when used to qualify inline __asm__ are that the compiler will not
> > (1) elid, or, (2) reorder, or, (3) intersperse, our inline asm with
> > the rest of the generated code.
> >   
> 
> "asm volatile" does not mean that at all.  It only guarantees (1),


Actually, you're probably right about (2), but (3)?

>From the gcc manual:

<quote>

Similarly, you can't expect a sequence of volatile asm instructions to
remain perfectly consecutive. If you want consecutive output, use a
single asm. Also GCC will perform some optimizations across a volatile
asm instruction, GCC does not "forget everything" when it encounters a
volatile asm instruction the way some other compilers do.

</quote>

I'm reading "Similarly, you can't expect a sequence of volatile asm
instructions to remain perfectly consecutive" to mean they're talking
about something like:

asm volatile(...);
asm volatile(...);
asm volatile(...);

But "If you want consecutive output, use a single asm" probably means:

asm volatile(... multiple instructions here ...);

would actually ensure the code written in there would not be
interspersed ... at least that's how I read it.

[ BTW "Also GCC will perform some optimizations across a volatile
asm instruction ..." is exactly the kind of optimizations we must
allow the compiler to do, but that's not related to point at hand. ]


> and
> only then if the asm is ever reachable.

Yup, of course.

Satyam
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/