Steven Dalton <[email protected]> writes:

> Since CUSP provides support for wrapping generic GPU memory
> into CUSP vectors it shouldn't be difficult to interface with the CUSP
> preconditioners either way.

Do you intend for CUSP to ever support distributed memory, or will it
always be purely local?

Attachment: pgpzAWuK9skYK.pgp
Description: PGP signature

Reply via email to