Project

General

Profile

Task #3158

Feature #2816: GPU offload / optimization for update&constraits, buffer ops and multi-gpu communication

Feature #2891: PME/PP GPU communications

use MPI_Isend() in sendFToPpCudaDirect and receiveCoordinatesFromPpCudaDirect

Added by Szilárd Páll about 1 month ago. Updated 12 days ago.

Status:
In Progress
Priority:
Normal
Assignee:
Category:
mdrun
Target version:
-
Difficulty:
uncategorized
Close

Description

Consider overlap between MPI and CUDA launches.

Associated revisions

Revision 5b594f3b (diff)
Added by Alan Gray about 1 month ago

GPU Receive for PME/PP GPU Force Communications

This change extends the PME/PP GPU force communication functionality
to allow the force buffer to be recieved direct to GPU memory on the
PP task.

Implements part of #2817
Refs #3158 #3159

Change-Id: I5b1cff1846c7c3bd966b6bf9c0af72769600ef18

Revision c5595a8e (diff)
Added by Alan Gray 29 days ago

GPU Coordinate PME/PP Communications

Extends PmePpCommGpu class to provide PP-side support for coordinate
transfers from either GPU or CPU to PME task, and adds new
PmeCoordinateReceiverGpu class to recieve coordinate data directly to
the GPU on the PME task.

Implements part of #2817
Refs TODOs #3157 #3158 #3159

Change-Id: Iefa2bdfd9813282ad8b07feeb7691f16880e61a2

Revision 6c0ab48c (diff)
Added by Alan Gray 4 days ago

Replace blocking with non-blocking receive in GPU PME coordinate receiver

Replaces MPI_Recv with MPI_Irecv in original coordinate receiver
method, and adds associated method containing MPI_Waitall which is
called to wait on data completion across all PP ranks.

Implements part of #3158

Change-Id: Ifd152973e4d9a4da53c0541e591b41d29be173bb

History

#1 Updated by Alan Gray 12 days ago

Implemented for coordinates in https://gerrit.gromacs.org/c/gromacs/+/14159

#2 Updated by Alan Gray 12 days ago

  • Status changed from New to In Progress

Also available in: Atom PDF