Project

General

Profile

Task #3155

Unify DD and CPU-GPU gather & scatter

Added by Berk Hess 12 months ago. Updated 12 months ago.

Status:
Accepted
Priority:
Normal
Assignee:
Category:
core library
Difficulty:
uncategorized
Close

Description

With the simulation state owned by the GPU, lots of special algorithms, such as REMD, need to make sure the CPU and GPU states are synced. A simple and elegant solution would be to collect the DD gather/scatter and GPU D2H/H2D calls in wrapper functions. Then any algorithm needed such functionality will work now and automatically for future algorithms as well.

History

#1 Updated by Berk Hess 12 months ago

  • Subject changed from Unify DD and CPU gather & scatter to Unify DD and CPU-GPU gather & scatter

Also available in: Atom PDF