From bb0418cea20f4276fa60a5e1de7c39975765c461 Mon Sep 17 00:00:00 2001 From: Giovanni Bussi <giovanni.bussi@gmail.com> Date: Mon, 27 Jun 2016 10:55:31 +0200 Subject: [PATCH] clarified neighbor list explanation [makedoc] --- user-doc/Performances.txt | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/user-doc/Performances.txt b/user-doc/Performances.txt index 53c09e3ed..992a1ba0f 100644 --- a/user-doc/Performances.txt +++ b/user-doc/Performances.txt @@ -127,6 +127,16 @@ Collective variables that can be speed up making us of neighbour lists: By tuning the cut-off for the neighbour list and the frequency for the recalculation of the list it is possible to balance between accuracy and performances. +Notice that for \ref COORDINATION and \ref DHENERGY using a neighbor list could imply that a smaller +number of atoms are requested to the host MD engine. This is typically true when considering +\ref COORDINATION of a small number of atoms (e.g. a ligand) again many atoms (e.g. water). +When the neighbor list is used, only the water atoms close to the ligand will be requested at each step. + +\warning +Notice that the calculation of the neighbour list is not not parallelized for \ref COORDINATION and \ref DHENERGY. +As a consequence, if you run +with many processors and/or OpenMP threads, the neighbor list might even make the calculation slower. + \page Openmp OpenMP -- GitLab