Max number of torsion-torsion grids increased to 12

plugins/amoeba/platforms/cuda/src/AmoebaCudaKernels.cpp

@@ -50,8 +50,8 @@ using namespace std;
  * -------------------------------------------------------------------------- */
 
 static void computeAmoebaLocalForces( AmoebaCudaData& data ) {
-    amoebaGpuContext gpu = data.getAmoebaGpu();
 
+    amoebaGpuContext gpu = data.getAmoebaGpu();
     if( 0 && data.getLog() ){
         (void) fprintf( data.getLog(), "computeAmoebaLocalForces\n" ); (void) fflush( data.getLog() );
     }
@@ -755,7 +755,9 @@ void CudaCalcAmoebaTorsionTorsionForceKernel::initialize(const System& system, c
             AmoebaTorsionTorsionForceImpl::reorderGrid( grid, reorderedGrid );
             reorder = 1;
             if( data.getLog() ){
-                (void) fprintf( data.getLog(), "CudaCalcAmoebaTorsionTorsionForceKernel::initialize: reordering torsion-torsion grid %d.\n", gridIndex );
+                (void) fprintf( data.getLog(), "CudaCalcAmoebaTorsionTorsionForceKernel Reordered torsion-torsion grid %4d [%u %u] %12.3f %12.3f   [%u %u] %12.3f %12.3f.\n",
+                                gridIndex, static_cast<unsigned int>(grid.size()), static_cast<unsigned int>(grid[0].size()), grid[0][0][0], grid[0][1][0],
+                                static_cast<unsigned int>(reorderedGrid.size() ),  static_cast<unsigned int>(reorderedGrid[0].size() ), reorderedGrid[0][0][0], reorderedGrid[0][1][0] );
             }
         }
         for (unsigned int ii = 0; ii < grid.size(); ii++) {
@@ -765,9 +767,11 @@ void CudaCalcAmoebaTorsionTorsionForceKernel::initialize(const System& system, c
 
                 floatGrids[gridIndex][ii][jj].resize( grid[ii][jj].size() );
                 if( reorder ){
+
                     for( unsigned int kk = 0; kk < grid[ii][jj].size(); kk++) {
                         floatGrids[gridIndex][ii][jj][kk] = static_cast<float>(reorderedGrid[ii][jj][kk]);
                     }
+
                 } else {
                     for( unsigned int kk = 0; kk < grid[ii][jj].size(); kk++) {
                         floatGrids[gridIndex][ii][jj][kk] = static_cast<float>(grid[ii][jj][kk]);
@@ -1121,7 +1125,6 @@ static void computeAmoebaVdwForce( AmoebaCudaData& data ) {
     data.initializeGpu();
 
     // Vdw14_7F
-
     kCalculateAmoebaVdw14_7Forces(gpu, data.getUseVdwNeighborList());
 }
 

plugins/amoeba/platforms/cuda/src/kernels/amoebaCudaGpu.cpp

@@ -1170,6 +1170,11 @@ void gpuSetAmoebaTorsionTorsionGrids(amoebaGpuContext amoebaGpu, const std::vect
     unsigned int torsionTorsionGrids                     = floatGrids.size(); // number of grids
     unsigned int totalGridEntries                        = 0;                 // total number of entries over all grids
                                                                               // used to allocate single memory buffer for grids 
+    if( floatGrids.size() > AMOEBA_MAX_TORSION_TORSION_GRIDS ){
+        std::stringstream message;
+        message  << "Number of slots for TorsionTorsionGrids is too small -- should be increased to at least " << floatGrids.size();
+        throw OpenMM::OpenMMException( message.str() );
+    }
 
     // 4 (grids) * (25 *25 grid)*(2 +4 a1, a2, f, f1,f2, f12) = 15000
 
@@ -1178,9 +1183,11 @@ void gpuSetAmoebaTorsionTorsionGrids(amoebaGpuContext amoebaGpu, const std::vect
 
     for (unsigned int ii = 0; ii < floatGrids.size(); ii++) {
 
+        unsigned int lastIndex                          = floatGrids[ii][0].size()-1;
+        float range                                     = floatGrids[ii][0][lastIndex][1] - floatGrids[ii][0][0][1];
         amoebaGpu->amoebaSim.amoebaTorTorGridOffset[ii] = (totalGridEntries/4);
         amoebaGpu->amoebaSim.amoebaTorTorGridBegin[ii]  = floatGrids[ii][0][0][0];
-        amoebaGpu->amoebaSim.amoebaTorTorGridDelta[ii]  = 360.0f/static_cast<float>(floatGrids[ii].size()-1);
+        amoebaGpu->amoebaSim.amoebaTorTorGridDelta[ii]  = range/static_cast<float>(floatGrids[ii].size()-1);
         amoebaGpu->amoebaSim.amoebaTorTorGridNy[ii]     = floatGrids[ii].size();
 
         for (unsigned int jj = 0; jj < floatGrids[ii].size(); jj++) {

plugins/amoeba/platforms/cuda/src/kernels/amoebaCudaTypes.h

@@ -48,6 +48,7 @@ enum CudaAmoebaNonbondedMethod
 };
 
 static const int AMOEBA_PME_ORDER = 5;
+static const int AMOEBA_MAX_TORSION_TORSION_GRIDS = 12;
 
 struct cudaAmoebaGmxSimulation {
 
@@ -120,10 +121,10 @@ struct cudaAmoebaGmxSimulation {
     unsigned int    amoebaTorsionTorsion_offset;    // Offset to end of torsion torsions
 
                                                     // grids
-    int   amoebaTorTorGridOffset[4];                // grid offset
-    int   amoebaTorTorGridNy[4];                    // 25
-    float amoebaTorTorGridBegin[4];                 // -180.0
-    float amoebaTorTorGridDelta[4];                 // 15.0
+    int   amoebaTorTorGridOffset[AMOEBA_MAX_TORSION_TORSION_GRIDS];                // grid offset
+    int   amoebaTorTorGridNy[AMOEBA_MAX_TORSION_TORSION_GRIDS];                    // 25
+    float amoebaTorTorGridBegin[AMOEBA_MAX_TORSION_TORSION_GRIDS];                 // -180.0
+    float amoebaTorTorGridDelta[AMOEBA_MAX_TORSION_TORSION_GRIDS];                 // 15.0
     float4*          pAmoebaTorsionTorsionGrids;    // torsion torsion grids
 
     unsigned int    amoebaUreyBradleys;             // Number of UB ixns