I agree that they should be documented, if for no other reason than to
prevent a user from trying to use multiple cores on a primitive that
already does so.
Only +/ . * is currently parallelized among the mainstream J primitives.
A candidate for multithreading has lots of computation compared to
memory access. Not too many primitives qualify. Perhaps x I. y,
division, some cases in the i.-family, and extended-precision multiply
using FFT would.
Henry Rich
On 1/22/2023 7:52 AM, Raul Miller wrote:
I thought that the point of having an explicit threading primitive was
to support / give J developers the decision making process in deciding
how to manage threads?
How is this supposed to work if the threads doing the work are not the
threads they are delegating the tasks to, in some undocumented
fashion?
Thanks,
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
