Project

General

Profile

Task #1694

large amount of serial code with p-coupling

Added by Szilárd Páll over 4 years ago. Updated over 3 years ago.

Status:
Closed
Priority:
Normal
Assignee:
-
Category:
-
Target version:
Difficulty:
uncategorized
Close

Description

The "rest" time with pressure coupling - especially with no DD and GPUs it can be 5-6x larger than without pcoupl; at the same time, the difference is only 2x with DD.

I suggest adding some multi-threading if possible.

Logs attached.

bench-hsw_1x12_gpu.log (24.4 KB) bench-hsw_1x12_gpu.log Szilárd Páll, 02/23/2015 02:01 AM
bench-hsw_6x1_gpu.log (28.4 KB) bench-hsw_6x1_gpu.log Szilárd Páll, 02/23/2015 02:01 AM
bench-hsw_1x12_gpu_pcoupl.log (24.1 KB) bench-hsw_1x12_gpu_pcoupl.log Szilárd Páll, 02/23/2015 02:03 AM
bench-hsw_6x1_gpu_pcoupl.log (28.8 KB) bench-hsw_6x1_gpu_pcoupl.log Szilárd Páll, 02/23/2015 02:03 AM

Associated revisions

Revision d18dc301 (diff)
Added by Berk Hess over 4 years ago

Reduced cost of Berendsen pressure coupling

The coordinate and box scaling for Berendsen pressure coupling is now
done only every nstpcouple steps. It now also uses OpenMP threading.

Partially solves #1694

Change-Id: Id6abbb7f60960076e683707277ebbe5ff8e75453

Revision 7b3f64db (diff)
Added by Berk Hess over 4 years ago

Removed reduction of dH/dl without FE

Removed call to sum_dhdl when not doing free-energy calculations.
This call could take measurable time at high parallelization.

Refs #1694

Change-Id: I268ffea8ea008f20e192c7ce683ce8f6bee9e759

History

#1 Updated by Szilárd Páll over 4 years ago

  • Description updated (diff)

#2 Updated by Roland Schulz over 4 years ago

#3 Updated by Gerrit Code Review Bot over 4 years ago

Gerrit received a related patchset '6' for Issue #1694.
Uploader: Szilárd Páll ()
Change-Id: Id6abbb7f60960076e683707277ebbe5ff8e75453
Gerrit URL: https://gerrit.gromacs.org/4492

#4 Updated by Gerrit Code Review Bot over 4 years ago

Gerrit received a related patchset '3' for Issue #1694.
Uploader: Szilárd Páll ()
Change-Id: I268ffea8ea008f20e192c7ce683ce8f6bee9e759
Gerrit URL: https://gerrit.gromacs.org/4491

#5 Updated by Szilárd Páll over 4 years ago

Gerrit Code Review Bot wrote:

Gerrit received a related patchset '6' for Issue #1694.
Uploader: Szilárd Páll ()
Change-Id: Id6abbb7f60960076e683707277ebbe5ff8e75453
Gerrit URL: https://gerrit.gromacs.org/4492

This drastically reduces the rest time with high thread count.

#6 Updated by Mark Abraham over 4 years ago

  • Status changed from New to In Progress

#7 Updated by Erik Lindahl over 4 years ago

  • Target version changed from 5.1 to future

The two first patches are in the repo. The final one might have to wait for post-5.1, since it has been at the RFC stage since March.

#8 Updated by Berk Hess over 4 years ago

  • Status changed from In Progress to Closed

The two merged changes significantly reduce the serial time with pressure coupling. The calcvir OpenMP parallelization in https://gerrit.gromacs.org/#/c/4272/ probably helps on Intel MIC, but gives mixed results on normal x86 CPUs. So we consider this issue closed.
We should continue to discuss which parts of https://gerrit.gromacs.org/#/c/4272/ should be merged (or modified).

#9 Updated by Mark Abraham over 3 years ago

  • Target version changed from future to 5.1

Also available in: Atom PDF