Hi Simon,

Simon Perkins <[email protected]> writes:
> Here's the patch!

The patch looks good. One minor complaint is that in absence of an
allocator kwarg, your patch changes existing behavior. Specifically, the
allocator that was previously used was the one of the array passed to
the reduction, and now it's pycuda.driver.mem_alloc.

Andreas

Attachment: pgpLCX2Si8Hx_.pgp
Description: PGP signature

_______________________________________________
PyCUDA mailing list
[email protected]
http://lists.tiker.net/listinfo/pycuda

Reply via email to