[SciPy-user] shared memory machines
Tue Feb 10 08:24:03 CST 2009
The Windows version seems to be working correctly on my computers.
I will put updates here:
Testing (particularly on Unix and friends) is appreciated. If you have
comments, encounter bugs, or have corrections, please send them to my email.
import sharedmem as shm
Now shm.zeros, shm.ones, and shm.empty should work like their numpy
equivalents. They are pickled and depickled by name (hidden from sight),
meaning only metadata is stored in the pickle. Call .copy() if you need
the pickle to contain a copy of the data as well. Unlike
multiprocessing.Array, these shared memory arrays can be sent through a
multiprocessing.Queue, tcp, or any other IPC you may think of. That's it.
I am done spamming this list with this for now.
More information about the SciPy-user