[SciPy-user] shared memory machines

[Gael Varoquaux]

On Tue, Feb 10, 2009 at 02:23:09AM +0100, Sturla Molden wrote:
> Ok, the work is basically done :)

Congratulation, Sturla!

I must admit that I am not too much enthousiastic about the
thread+polling to do the cleaning up. I don't really understand why it is
necessary. If the view of the array is the last to be decref, than
'buf.shm_nattch' should be 0, and as a result the freeing up of the
memory can happen in the dealloc. Or did I miss something?

Cheers,

Gaël