Fix for isolated ions

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

@@ -93,7 +93,7 @@ KernelImpl* AmoebaCudaKernelFactory::createKernelImpl(std::string name, const Pl
     if( mapIterator == contextToAmoebaDataMap.end() ){
         amoebaCudaData                         = new AmoebaCudaData( cudaPlatformData );
         contextToAmoebaDataMap[&context]       = amoebaCudaData;
-        amoebaCudaData->setLog( stderr );
+        //amoebaCudaData->setLog( stderr );
         amoebaCudaData->setContextImpl( static_cast<void*>(&context) );
     } else {
         amoebaCudaData                         = mapIterator->second;

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

@@ -86,13 +86,14 @@ amoebaGpuContext amoebaGpuInit( _gpuContext* gpu )
 extern "C"
 void gpuPrintCudaStream( std::string name,
                          unsigned int length, unsigned int subStreams, unsigned int stride,
+                         unsigned int memoryFootprint,
                          void*  pSysStream, void* pDevStream,
                          void*  pSysData,   void* pDevData, FILE* log)
 {
    
-    (void) fprintf( log, "     %-35s [%8u %5u %8u] Stream[%p %p] Data[%16p %16p]\n",
+    (void) fprintf( log, "     %-35s [%8u %5u %8u %8u] Stream[%p %p] Data[%16p %16p]\n",
                     name.c_str(), length, subStreams,
-                    stride, pSysStream, pDevStream, pSysData, pDevData );
+                    stride, memoryFootprint, pSysStream, pDevStream, pSysData, pDevData );
 }
 
 extern "C"
@@ -102,6 +103,7 @@ void gpuPrintCudaStreamFloat( CUDAStream<float>* cUDAStream, FILE* log )
     if( cUDAStream == NULL )return;
     gpuPrintCudaStream( cUDAStream->_name.c_str(),
                         cUDAStream->_length, cUDAStream->_subStreams, cUDAStream->_stride,
+                        cUDAStream->_length*cUDAStream->_subStreams*sizeof( float ),
                         static_cast<void*>(cUDAStream->_pSysStream), static_cast<void*>(cUDAStream->_pDevStream),
                         static_cast<void*>(cUDAStream->_pSysData), static_cast<void*>(cUDAStream->_pDevData), log );
 }
@@ -113,6 +115,7 @@ void gpuPrintCudaStreamFloat2( CUDAStream<float2>* cUDAStream, FILE* log )
     if( cUDAStream == NULL )return;
     gpuPrintCudaStream( cUDAStream->_name.c_str(),
                         cUDAStream->_length, cUDAStream->_subStreams, cUDAStream->_stride,
+                        cUDAStream->_length*cUDAStream->_subStreams*sizeof( float2 ),
                         static_cast<void*>(cUDAStream->_pSysStream), static_cast<void*>(cUDAStream->_pDevStream),
                         static_cast<void*>(cUDAStream->_pSysData), static_cast<void*>(cUDAStream->_pDevData), log );
 }
@@ -124,6 +127,7 @@ void gpuPrintCudaStreamFloat4( CUDAStream<float4>* cUDAStream, FILE* log )
     if( cUDAStream == NULL )return;
     gpuPrintCudaStream( cUDAStream->_name.c_str(),
                         cUDAStream->_length, cUDAStream->_subStreams, cUDAStream->_stride,
+                        cUDAStream->_length*cUDAStream->_subStreams*sizeof( float4 ),
                         static_cast<void*>(cUDAStream->_pSysStream), static_cast<void*>(cUDAStream->_pDevStream),
                         static_cast<void*>(cUDAStream->_pSysData), static_cast<void*>(cUDAStream->_pDevData), log );
 }
@@ -135,6 +139,7 @@ void gpuPrintCudaStreamUnsignedInt( CUDAStream<unsigned int>* cUDAStream, FILE*
     if( cUDAStream == NULL )return;
     gpuPrintCudaStream( cUDAStream->_name.c_str(),
                         cUDAStream->_length, cUDAStream->_subStreams, cUDAStream->_stride,
+                        cUDAStream->_length*cUDAStream->_subStreams*sizeof( unsigned int ),
                         static_cast<void*>(cUDAStream->_pSysStream), static_cast<void*>(cUDAStream->_pDevStream),
                         static_cast<void*>(cUDAStream->_pSysData), static_cast<void*>(cUDAStream->_pDevData), log );
 }
@@ -146,6 +151,7 @@ void gpuPrintCudaStreamInt( CUDAStream<int>* cUDAStream, FILE* log )
     if( cUDAStream == NULL )return;
     gpuPrintCudaStream( cUDAStream->_name.c_str(),
                         cUDAStream->_length, cUDAStream->_subStreams, cUDAStream->_stride,
+                        cUDAStream->_length*cUDAStream->_subStreams*sizeof( int ),
                         static_cast<void*>(cUDAStream->_pSysStream), static_cast<void*>(cUDAStream->_pDevStream),
                         static_cast<void*>(cUDAStream->_pSysData), static_cast<void*>(cUDAStream->_pDevData), log );
 }
@@ -157,6 +163,7 @@ void gpuPrintCudaStreamInt2( CUDAStream<int2>* cUDAStream, FILE* log )
     if( cUDAStream == NULL )return;
     gpuPrintCudaStream( cUDAStream->_name.c_str(),
                         cUDAStream->_length, cUDAStream->_subStreams, cUDAStream->_stride,
+                        cUDAStream->_length*cUDAStream->_subStreams*sizeof( int2 ),
                         static_cast<void*>(cUDAStream->_pSysStream), static_cast<void*>(cUDAStream->_pDevStream),
                         static_cast<void*>(cUDAStream->_pSysData), static_cast<void*>(cUDAStream->_pDevData), log );
 }
@@ -168,6 +175,7 @@ void gpuPrintCudaStreamInt4( CUDAStream<int4>* cUDAStream, FILE* log )
     if( cUDAStream == NULL )return;
     gpuPrintCudaStream( cUDAStream->_name.c_str(),
                         cUDAStream->_length, cUDAStream->_subStreams, cUDAStream->_stride,
+                        cUDAStream->_length*cUDAStream->_subStreams*sizeof( int4 ),
                         static_cast<void*>(cUDAStream->_pSysStream), static_cast<void*>(cUDAStream->_pDevStream),
                         static_cast<void*>(cUDAStream->_pSysData), static_cast<void*>(cUDAStream->_pDevData), log );
 }
@@ -1277,15 +1285,6 @@ static void gpuRotationToLabFrameAllocate( amoebaGpuContext amoebaGpu )
    
     int paddedNumberOfAtoms = amoebaGpu->gpuContext->sim.paddedNumberOfAtoms;
 
-#ifdef AMOEBA_DEBUG
-    if( amoebaGpu->log ){
-        (void) fprintf( amoebaGpu->log,"%s: paddedNumberOfAtoms=%d\n",
-                        methodName.c_str(), paddedNumberOfAtoms ); (void) fflush( amoebaGpu->log );
-    }
-#endif
-
-    // work space
- 
     // parameters
  
     amoebaGpu->psMultipoleParticlesIdsAndAxisType                      = new CUDAStream<int4>(paddedNumberOfAtoms,    1, "MultipoleParticlesIdsAndAxisType");
@@ -1296,9 +1295,11 @@ static void gpuRotationToLabFrameAllocate( amoebaGpuContext amoebaGpu )
 
     amoebaGpu->psMolecularDipole                                       = new CUDAStream<float>(3*paddedNumberOfAtoms, 1, "MolecularDipole");
     amoebaGpu->amoebaSim.pMolecularDipole                              = amoebaGpu->psMolecularDipole->_pDevData;
+    memset( amoebaGpu->psMolecularDipole->_pSysData,      0, sizeof(float)*3*paddedNumberOfAtoms );
 
     amoebaGpu->psMolecularQuadrupole                                   = new CUDAStream<float>(9*paddedNumberOfAtoms, 1, "MolecularQuadrupole");
     amoebaGpu->amoebaSim.pMolecularQuadrupole                          = amoebaGpu->psMolecularQuadrupole->_pDevData;
+    memset( amoebaGpu->psMolecularQuadrupole->_pSysData,      0, sizeof(float)*9*paddedNumberOfAtoms );
  
     // output
  

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

@@ -47,38 +47,10 @@ typedef MapIntFloat::const_iterator MapIntFloatCI;
 struct _amoebaGpuContext {
     
     _gpuContext* gpuContext;
+    cudaAmoebaGmxSimulation amoebaSim;
  
     FILE* log;
 
-    //bool bOutputBufferPerWarp;
-    //unsigned int paddedNumberOfAtoms;
-    //unsigned int nonbondBlocks;
-    //unsigned int nonbondThreadsPerBlock;
-    //unsigned int nonbondOutputBuffers;
-    //unsigned int threadsPerBlock;
-    //unsigned int fieldReduceThreadsPerBlock;
-    //unsigned int outputBuffers; 
-    unsigned int workUnits; 
-
-    // workspace arrays
-
-    CUDAStream<float>*  psWorkArray_3_1; 
-    CUDAStream<float>*  psWorkArray_3_2; 
-    CUDAStream<float>*  psWorkArray_3_3; 
-    CUDAStream<float>*  psWorkArray_3_4; 
-
-    CUDAStream<float>*  psWorkArray_1_1; 
-    CUDAStream<float>*  psWorkArray_1_2; 
-
-    CUDAStream<unsigned int>*  psWorkUnit; 
-    CUDAStream<int>*  psScalingIndicesIndex; 
-    CUDAStream<int>*  ps_D_ScaleIndices; 
-    CUDAStream<int2>* ps_P_ScaleIndices; 
-    CUDAStream<int2>* ps_M_ScaleIndices; 
-
-    cudaAmoebaGmxSimulation amoebaSim;
-    int maxCovalentDegreeSz;
-
     CUDAStream<int4>*   psAmoebaBondID;
     CUDAStream<float2>* psAmoebaBondParameter;
 
@@ -116,6 +88,25 @@ struct _amoebaGpuContext {
     CUDAStream<int4>*   psAmoebaTorsionTorsionID3;
     CUDAStream<float4>* psAmoebaTorsionTorsionGrids;
 
+    unsigned int workUnits; 
+
+    // workspace arrays
+
+    CUDAStream<float>*  psWorkArray_3_1; 
+    CUDAStream<float>*  psWorkArray_3_2; 
+    CUDAStream<float>*  psWorkArray_3_3; 
+    CUDAStream<float>*  psWorkArray_3_4; 
+
+    CUDAStream<float>*  psWorkArray_1_1; 
+    CUDAStream<float>*  psWorkArray_1_2; 
+
+    CUDAStream<unsigned int>*  psWorkUnit; 
+    CUDAStream<int>*  psScalingIndicesIndex; 
+    CUDAStream<int>*  ps_D_ScaleIndices; 
+    CUDAStream<int2>* ps_P_ScaleIndices; 
+    CUDAStream<int2>* ps_M_ScaleIndices; 
+
+    int maxCovalentDegreeSz;
     float solventDielectric;
 
     // multipole parameters
@@ -126,7 +117,6 @@ struct _amoebaGpuContext {
     // buffer indices used for mapping torques onto forces 
 
     int maxTorqueBufferIndex;
-    int useNewTorqueMapScheme;
     int torqueMapForce4Delete;
     CUDAStream<int4>* psMultipoleParticlesTorqueBufferIndices;
     CUDAStream<float4>*  psTorqueMapForce4; 

plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaFixedEAndGkFields.cu

@@ -6,7 +6,8 @@
 #include "amoebaCudaKernels.h"
 #include "kCalculateAmoebaCudaUtilities.h"
 
-#define AMOEBA_DEBUG
+//#define AMOEBA_DEBUG
+#undef AMOEBA_DEBUG
 
 static __constant__ cudaGmxSimulation cSim;
 static __constant__ cudaAmoebaGmxSimulation cAmoebaSim;
@@ -480,7 +481,7 @@ void cudaComputeAmoebaFixedEAndGkFields( amoebaGpuContext amoebaGpu )
 
         // write results to file
 
-        if( 1 ){
+        if( 0 ){
             std::vector<int> fileId;
             //fileId.push_back( 0 );
             VectorOfDoubleVectors outputVector;

plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaKirkwood.cu

@@ -1730,7 +1730,7 @@ static void kReduceToBornForcePrefactor( amoebaGpuContext amoebaGpu )
     }
     LAUNCHERROR("kReduceToBornForcePrefactor");
 
-#define AMOEBA_DEBUG
+//#define AMOEBA_DEBUG
 #ifdef AMOEBA_DEBUG
     if( amoebaGpu->log ){
 
@@ -1748,7 +1748,7 @@ static void kReduceToBornForcePrefactor( amoebaGpuContext amoebaGpu )
         }
         (void) fflush( amoebaGpu->log );
 */
-        if( 1 ){
+        if( 0 ){
             std::vector<int> fileId;
             //fileId.push_back( 0 );
             VectorOfDoubleVectors outputVector;
@@ -1943,7 +1943,7 @@ void kCalculateAmoebaKirkwood( amoebaGpuContext amoebaGpu )
 
     cudaComputeAmoebaMapTorqueAndAddToForce( amoebaGpu, amoebaGpu->psTorque );
 
-    if( 1 ){
+    if( 0 ){
         std::vector<int> fileId;
         VectorOfDoubleVectors outputVector;
         //cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4,            outputVector, NULL, 1.0f );
@@ -1956,7 +1956,7 @@ void kCalculateAmoebaKirkwood( amoebaGpuContext amoebaGpu )
 
     kCalculateObcGbsaForces2( amoebaGpu->gpuContext );
 
-    if( 1 ){
+    if( 0 ){
         std::vector<int> fileId;
         VectorOfDoubleVectors outputVector;
         //cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4,            outputVector, NULL, 1.0f );

plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaMutualInducedAndGkFields.cu

@@ -33,8 +33,8 @@ void GetCalculateAmoebaCudaMutualInducedAndGkFieldsSim(amoebaGpuContext amoebaGp
     RTERROR(status, "GetCalculateAmoebaCudaMutualInducedAndGkFieldSim: cudaMemcpyFromSymbol: SetSim copy from cAmoebaSim failed");
 }
 
-#define AMOEBA_DEBUG
-//#undef AMOEBA_DEBUG
+//#define AMOEBA_DEBUG
+#undef AMOEBA_DEBUG
 
 #define GK
 #include "kCalculateAmoebaCudaMutualInducedParticle.h"
@@ -923,7 +923,7 @@ static void cudaComputeAmoebaMutualInducedAndGkFieldBySOR( amoebaGpuContext amoe
     }
 
 #ifdef AMOEBA_DEBUG
-    if( 1 ){
+    if( 0 ){
         std::vector<int> fileId;
         //fileId.push_back( 0 );
         VectorOfDoubleVectors outputVector;