From 38f9b8e92cabb3638d3c4d7412045426357108a9 Mon Sep 17 00:00:00 2001
From: Giovanni Bussi <giovanni.bussi@gmail.com>
Date: Thu, 16 Jan 2020 00:02:02 +0100
Subject: [PATCH] Fix #410 - DOC ONLY
---
user-doc/tutorials/hrex.txt | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/user-doc/tutorials/hrex.txt b/user-doc/tutorials/hrex.txt
index 71f28719d..f42425e00 100644
--- a/user-doc/tutorials/hrex.txt
+++ b/user-doc/tutorials/hrex.txt
@@ -129,7 +129,7 @@ do
# -maxwarn is often needed because box could be charged
grompp_mpi_d -maxwarn 1 -o topol$i.tpr -f grompp$i.mdp -p topol$i.top
done
-mpirun -np $nrep gmx_mpi mdrun_d -v -plumed plumed.dat -multi $nrep -replex 100 -nsteps 15000000 -hrex
+mpirun -np $nrep gmx_mpi mdrun_d -v -plumed plumed.dat -multi $nrep -replex 100 -nsteps 15000000 -hrex -dlb no
\endverbatim
Notice that total cell could be charged. This happens whenever the scaled portion of the system
@@ -156,6 +156,9 @@ Warnings:
unnoticed errors in the final results.
- Choose neighbor list update (nstlist) that divides replex. Notice that running with GPUs
GROMACS is going to change nstlist automatically, be sure that it still divides replex.
+- It seems that when using multiple processes per replica it is necessary to switch off
+ dynamic load balancing (`-dlb no`) otherwise the simulation could crash randomly,
+ see \issue{410}.
- Option -hrex requires also option -plumed. If you do not care about plumed, just provide an empty
plumed.dat file.
- It should work correctly if replicas have different force-field, temperature, lambda, pressure,
--
GitLab