Project

General

Profile

Task #2469

implement GPU timer reduction for reporting

Added by Szilárd Páll over 1 year ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
mdrun
Target version:
-
Difficulty:
simple
Close

Description

Unlike in CUDA where device-side timing has severe limitations (can't time reliably with multiple streams), in OpenCL the same restrictions do not apply. While we have all facilities to correctly time GPU-side execution, the reporting is missing the accumulation/reduction.

A simple reduction across ranks should allow unrestricted reporting.


Related issues

Related to GROMACS - Bug #2468: incorrect GPU timing reported with OpenCL and domain decompositionNew

History

#1 Updated by Szilárd Páll over 1 year ago

  • Related to Bug #2468: incorrect GPU timing reported with OpenCL and domain decomposition added

Also available in: Atom PDF