On Tue, 5 Dec 2000, Peter Samuelson wrote:
> [Tigran Aivazian]
> > The only reason one could think of was to "hold the lock for as short
> > time as possible" but a minute's thought reveals that such reason is
> > invalid (i.e. one is more likely to waste time spinning on the lock
> > than to save it by dropping/retaking it, given the relative duration
> > of the instructions we execute there without the lock).
>
> If there is no contention, you do not spin, no time wasted
wrong -- time is wasted -- 1 decb for lock and 1 movb for unlock. No time
is wasted on spinning but 2 instructions wasted for taking/dropping the
lock.
>
> If there *is* contention, you deserialize the routine just a little
> bit, which is generally a Good Thing.
>
> Whether a memset of 92 bytes (on 32-bit arch), plus an atomic_set(),
> are worth deserializing, I do not know.
>
Of course, they are worth it. Actually, I don't understand how can you
even doubt it? Even a single cycle of code executed for _no reason_ must
be removed, if for no other reason than to make the code easier to
understand and prevent people from asking questions like "why does X do Y
for no reason?" -- if there are no such items "Y" then the questions will
also cease.
Regards,
Tigran
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Dec 07 2000 - 21:00:13 EST