imports and python locking

[Martin Pool]

On 21 Jul 2005, Denys Duchier <duchier at ps.uni-sb.de> wrote:
> Martin Pool <mbp at sourcefrog.net> writes:
> 
> > The attached c file demonstrates the problem when compiled with "gcc
> > -lpthread".  Compiling with -march=686 (in the hope of using cmpxchg or
> > something) does not help.  Run it under strace and observe thousands of
> > pointless futex calls.
> 
> all I see are FUTEX_WAKE operations. AIUI, there's no way around those as the
> queue associated with a futex lives in the kernel.  What we don't see are
> FUTEX_WAIT operations (all sem_wait are non-contended and take place only in
> userspace).

According to the futex man page, there should only be a FUTEX_WAKE call
when there are other processes waiting for the resource:

       Waiting on a futex, to ’down’ it, is the reverse
       operation. Atomically decrement the counter and check if it
       changed to 0, in which case the operation is done and the futex
       was uncontended. In all other circumstances, the process should
       set the counter to -1 and request that the kernel wait for
       another process to up the futex. This is done using the
       FUTEX_WAIT operation.

But here is some remarkable information: if you build the test case
-static, then the futex wakes don't happen!

-- 
Martin