Avoid PME tuning decreasing rcoulomb

[alexxy/gromacs.git] / src / programs / mdrun / pme_loadbal.c

diff --git a/src/programs/mdrun/pme_loadbal.c b/src/programs/mdrun/pme_loadbal.c
index 4606f735cc18e72751509d980a2fd135d9a501b5..d6fd68f37269d10692d625172706519c053c2cf3 100644 (file)
--- a/src/programs/mdrun/pme_loadbal.c
+++ b/src/programs/mdrun/pme_loadbal.c
@@ -36,7 +36,8 @@
 #include <config.h>
 #endif
 
-#include "smalloc.h"
+#include "gromacs/utility/smalloc.h"
+#include "types/commrec.h"
 #include "network.h"
 #include "calcgrid.h"
 #include "pme.h"
@@ -260,6 +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)
     {
@@ -317,6 +327,9 @@ static gmx_bool pme_loadbal_increase_cutoff(pme_load_balancing_t  pme_lb,
     /* The Ewald coefficient is inversly proportional to the cut-off */
     set->ewaldcoeff_q =
         pme_lb->setup[0].ewaldcoeff_q*pme_lb->setup[0].rcut_coulomb/set->rcut_coulomb;
+    /* We set ewaldcoeff_lj in set, even when LJ-PME is not used */
+    set->ewaldcoeff_lj =
+        pme_lb->setup[0].ewaldcoeff_lj*pme_lb->setup[0].rcut_coulomb/set->rcut_coulomb;
 
     set->count   = 0;
     set->cycles  = 0;
@@ -662,10 +675,27 @@ gmx_bool pme_load_balance(pme_load_balancing_t pme_lb,
     ic->rlistlong    = set->rlistlong;
     ir->nstcalclr    = set->nstcalclr;
     ic->ewaldcoeff_q = set->ewaldcoeff_q;
+    /* TODO: centralize the code that sets the potentials shifts */
     if (ic->coulomb_modifier == eintmodPOTSHIFT)
     {
         ic->sh_ewald = gmx_erfc(ic->ewaldcoeff_q*ic->rcoulomb);
     }
+    if (EVDW_PME(ic->vdwtype))
+    {
+        /* We have PME for both Coulomb and VdW, set rvdw equal to rcoulomb */
+        ic->rvdw            = set->rcut_coulomb;
+        ic->ewaldcoeff_lj   = set->ewaldcoeff_lj;
+        if (ic->vdw_modifier == eintmodPOTSHIFT)
+        {
+            real crc2;
+
+            ic->dispersion_shift.cpot = -pow(ic->rvdw, -6.0);
+            ic->repulsion_shift.cpot  = -pow(ic->rvdw, -12.0);
+            ic->sh_invrc6             = -ic->dispersion_shift.cpot;
+            crc2                      = sqr(ic->ewaldcoeff_lj*ic->rvdw);
+            ic->sh_lj_ewald           = (exp(-crc2)*(1 + crc2 + 0.5*crc2*crc2) - 1)*pow(ic->rvdw, -6.0);
+        }
+    }
 
     bUsesSimpleTables = uses_simple_tables(ir->cutoff_scheme, nbv, 0);
     if (pme_lb->cutoff_scheme == ecutsVERLET &&
@@ -692,17 +722,12 @@ gmx_bool pme_load_balance(pme_load_balancing_t pme_lb,
         }
 #endif  /* GMX_THREAD_MPI */
     }
-    else
-    {
-        init_interaction_const_tables(NULL, ic, bUsesSimpleTables,
-                                      rtab);
-    }
 
-    if (pme_lb->cutoff_scheme == ecutsVERLET && nbv->ngrp > 1)
-    {
-        init_interaction_const_tables(NULL, ic, bUsesSimpleTables,
-                                      rtab);
-    }
+    /* Usually we won't need the simple tables with GPUs.
+     * But we do with hybrid acceleration and with free energy.
+     * To avoid bugs, we always re-initialize the simple tables here.
+     */
+    init_interaction_const_tables(NULL, ic, bUsesSimpleTables, rtab);
 
     if (cr->duty & DUTY_PME)
     {
@@ -814,8 +839,8 @@ static void print_pme_loadbal_settings(pme_load_balancing_t pme_lb,
     {
         md_print_warn(cr, fplog,
                       "NOTE: PME load balancing increased the non-bonded workload by more than 50%%.\n"
-                      "      For better performance use (more) PME nodes (mdrun -npme),\n"
-                      "      or in case you are beyond the scaling limit, use less nodes in total.\n");
+                      "      For better performance, use (more) PME ranks (mdrun -npme),\n"
+                      "      or if you are beyond the scaling limit, use fewer total ranks (or nodes).\n");
     }
     else
     {