locking problem with mips atomicity

[Richard Sandiford]

Richard Henderson <rth@redhat.com> writes:
> On Thu, Mar 18, 2004 at 11:04:19AM +0000, Richard Sandiford wrote:
>> Sorry if I've missed the explanation, but why is using a memory clobber
>> an improvement over an unused memory operand?
>
> To make sure that any memory protected by the mutex is not
> moved across the acquisition of the lock.

Ah, right, sorry for being dense ;)