Re: [patch] spinlocks: remove 'volatile'

[Albert Cahalan]

Linus Torvalds writes:

> AND I POINTED OUT THAT EVEN IN YOUR TRIVIAL EXAMPLE, VOLATILE WAS
> ACTUALLY THE WRONG THING TO DO.
>
> And that's _exactly_ because the language environment (in this case
> the CPU itself) has evolved past the point it was 30 years ago.

Key thing being "language environment", meaning gcc.

By the language spec, volatile is a low-performance way to
get the job done. Following the language spec is damn hard
these days, considering some of the evil things PCI does in
the name of performance. The compiler probably should offer
an option to call a function for read/write to volatile mem.
Such a function could take a lock, determine if a PCI read
is required to enforce ordering, perform the operation, and
then release the lock.

That's all theoretical though. Today, gcc's volatile does
not follow the C standard on modern hardware. Bummer.
It'd be low-performance anyway though.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/