From: Berk Hess <hess@kth.se>
Date: Fri, 26 Sep 2014 18:47:47 +0000 (+0200)
Subject: Avoid PME tuning decreasing rcoulomb
X-Git-Url: http://biod.pnpi.spb.ru/gitweb/?p=alexxy%2Fgromacs.git;a=commitdiff_plain;h=d3e2d78f5c42e4ad8c3d6dabf7a6d909b2a9c280

Avoid PME tuning decreasing rcoulomb

With mdrun input with the grid much finer than fourier-spacing,
PME tuning could reduce rcoulomb below the starting value.
Fixes #1606.

Change-Id: I3ae7fe38a4240251aadb96336199b306fc7dce40
---

diff --git a/src/programs/mdrun/pme_loadbal.c b/src/programs/mdrun/pme_loadbal.c
index 1c12f86435..d6fd68f372 100644
--- a/src/programs/mdrun/pme_loadbal.c
+++ b/src/programs/mdrun/pme_loadbal.c
@@ -261,6 +261,15 @@ static gmx_bool pme_loadbal_increase_cutoff(pme_load_balancing_t  pme_lb,
     while (sp <= 1.001*pme_lb->setup[pme_lb->cur].spacing || !grid_ok);
 
     set->rcut_coulomb = pme_lb->cut_spacing*sp;
+    if (set->rcut_coulomb < pme_lb->rcut_coulomb_start)
+    {
+        /* This is unlikely, but can happen when e.g. continuing from
+         * a checkpoint after equilibration where the box shrank a lot.
+         * We want to avoid rcoulomb getting smaller than rvdw
+         * and there might be more issues with decreasing rcoulomb.
+         */
+        set->rcut_coulomb = pme_lb->rcut_coulomb_start;
+    }
 
     if (pme_lb->cutoff_scheme == ecutsVERLET)
     {