Skip to content

GitLab

Commit 4d763c4d authored by Mark Friedrichs's avatar Mark Friedrichs
Browse files

Fixed bug in direct PME 

Added diagnostics for handling reorder of atoms
parent 8b8defe8
Expand all Show whitespace changes
Inline Side-by-side
Showing with 546 additions and 313 deletions
plugins/amoeba/openmmapi/src/AmoebaMultipoleForce.cpp
View file @ 4d763c4d
......@@ -38,7 +38,7 @@ using namespace OpenMM;
using std::string;
using std::vector;
AmoebaMultipoleForce::AmoebaMultipoleForce() : nonbondedMethod(NoCutoff), pmeBSplineOrder(5), cutoffDistance(1.0), ewaldErrorTol(5e-4), mutualInducedIterationMethod(SOR), mutualInducedMaxIterations(60),
AmoebaMultipoleForce::AmoebaMultipoleForce() : nonbondedMethod(NoCutoff), pmeBSplineOrder(5), cutoffDistance(1.0), ewaldErrorTol(1e-4), mutualInducedIterationMethod(SOR), mutualInducedMaxIterations(60),
mutualInducedTargetEpsilon(1.0e-05), scalingDistanceCutoff(100.0), electricConstant(138.9354558456) {
}
......
plugins/amoeba/platforms/cuda/src/AmoebaCudaKernels.cpp
View file @ 4d763c4d
......@@ -779,17 +779,15 @@ static void computeAmoebaMultipoleForce( AmoebaCudaData& data ) {
if( data.getMultipoleForceCount() == 0 ){
gpuCopyWorkUnit( gpu );
}
//if( data.getApplyCutoff() && (data.getMultipoleForceCount() % 100) == 0 ){
//gpuReorderAtoms(gpu->gpuContext);
//}
data.incrementMultipoleForceCount();
data.initializeGpu();
if( 0 && data.getLog() ){
(void) fprintf( data.getLog(), "computeAmoebaMultipoleForce\n" );
(void) fprintf( data.getLog(), "In computeAmoebaMultipoleForce\n" );
(void) fflush( data.getLog());
}
data.initializeGpu();
// calculate Born radii
if( data.getHasAmoebaGeneralizedKirkwood() ){
......@@ -974,23 +972,27 @@ void CudaCalcAmoebaMultipoleForceKernel::initialize(const System& system, const
nb.setCutoffDistance(force.getCutoffDistance());
std::vector<int> pmeGridDimension;
force.getPmeGridDimensions( pmeGridDimension );
if( 1 || pmeGridDimension[0] == 0 ){
int pmeParametersSetBasedOnEwaldErrorTolerance;
if( pmeGridDimension[0] == 0 ){
NonbondedForceImpl::calcPMEParameters(system, nb, alpha, xsize, ysize, zsize);
/*
alpha = 5.446;
xsize = 60;
ysize = 48;
zsize = 48;
*/
pmeParametersSetBasedOnEwaldErrorTolerance = 1;
} else {
alpha = force.getAEwald();
xsize = pmeGridDimension[0];
ysize = pmeGridDimension[1];
zsize = pmeGridDimension[2];
pmeParametersSetBasedOnEwaldErrorTolerance = 0;
}
if( data.getLog() ){
(void) fprintf( data.getLog(), "AmoebaMultipoleForce: PME parameters tol=%12.3e cutoff=%12.3f alpha=%12.3f [%d %d %d]\n",
(void) fprintf( data.getLog(), "AmoebaMultipoleForce: PME parameters tol=%12.3e cutoff=%12.3f alpha=%12.3f [%d %d %d] -",
force.getEwaldErrorTolerance(), force.getCutoffDistance(), alpha, xsize, ysize, zsize );
if( pmeParametersSetBasedOnEwaldErrorTolerance ){
(void) fprintf( data.getLog(), " parameters set based on error tolerance and OpenMM algorithm.\n" );
} else {
double impliedTolerance = alpha*force.getCutoffDistance();
impliedTolerance = 0.5*exp( -(impliedTolerance*impliedTolerance) );
(void) fprintf( data.getLog(), " using input parameters implied tolerance=%12.3e\n", impliedTolerance );
}
(void) fflush( data.getLog() );
}
gpuSetAmoebaPMEParameters(data.getAmoebaGpu(), (float) alpha, xsize, ysize, zsize);
......
plugins/amoeba/platforms/cuda/src/kernels/amoebaCudaGpu.cpp
View file @ 4d763c4d
......@@ -3909,6 +3909,33 @@ void cudaWriteVectorOfDoubleVectorsToFile( char* fname, std::vector<int>& fileId
(void) fclose( filePtr );
}
CUDAStream<float>* reorderFloat( amoebaGpuContext amoebaGpu, CUDAStream<float>* arrayToReorder ){
gpuContext gpu = amoebaGpu->gpuContext;
CUDAStream<float>* reorderdArray = new CUDAStream<float>(amoebaGpu->gpuContext->sim.paddedNumberOfAtoms, 1, "TempReorder");
int* order = gpu->psAtomIndex->_pSysData;
for( int ii = 0; ii < gpu->natoms; ii++ ){
reorderdArray->_pSysStream[0][order[ii]] = arrayToReorder->_pSysStream[0][ii];
}
return reorderdArray;
}
CUDAStream<float4>* reorderFloat4( amoebaGpuContext amoebaGpu, CUDAStream<float4>* arrayToReorder ){
gpuContext gpu = amoebaGpu->gpuContext;
CUDAStream<float4>* reorderdArray = new CUDAStream<float4>(amoebaGpu->gpuContext->sim.paddedNumberOfAtoms, 1, "TempReorder4");
int* order = gpu->psAtomIndex->_pSysData;
for( int ii = 0; ii < gpu->natoms; ii++ ){
reorderdArray->_pSysStream[0][order[ii]].x = arrayToReorder->_pSysStream[0][ii].x;
reorderdArray->_pSysStream[0][order[ii]].y = arrayToReorder->_pSysStream[0][ii].y;
reorderdArray->_pSysStream[0][order[ii]].z = arrayToReorder->_pSysStream[0][ii].z;
reorderdArray->_pSysStream[0][order[ii]].w = arrayToReorder->_pSysStream[0][ii].w;
}
return reorderdArray;
}
/**---------------------------------------------------------------------------------------
Load contents of arrays into vector
......@@ -3920,7 +3947,9 @@ void cudaWriteVectorOfDoubleVectorsToFile( char* fname, std::vector<int>& fileId
--------------------------------------------------------------------------------------- */
void cudaLoadCudaFloatArray( int numberOfParticles, int entriesPerParticle, CUDAStream<float>* array, VectorOfDoubleVectors& outputVector )
void cudaLoadCudaFloatArray( int numberOfParticles, int entriesPerParticle,
CUDAStream<float>* array, VectorOfDoubleVectors& outputVector,
int* order )
{
// ---------------------------------------------------------------------------------------
......@@ -3929,14 +3958,19 @@ void cudaLoadCudaFloatArray( int numberOfParticles, int entriesPerParticle, CUDA
// ---------------------------------------------------------------------------------------
array->Download();
int runningIndex = 0;
int orderIndex = 0;
outputVector.resize( numberOfParticles );
for( int ii = 0; ii < numberOfParticles; ii++ ){
if( order ){
orderIndex = order[ii];
} else {
orderIndex = ii;
}
for( int jj = 0; jj < entriesPerParticle; jj++ ) {
outputVector[ii].push_back( array->_pSysStream[0][runningIndex++] );
outputVector[orderIndex].push_back( array->_pSysStream[0][entriesPerParticle*ii+jj] );
}
}
}
......@@ -3975,6 +4009,7 @@ void cudaLoadCudaFloat2Array( int numberOfParticles, int entriesPerParticle, CUD
}
}
/**---------------------------------------------------------------------------------------
Load contents of arrays into vector
......@@ -3983,10 +4018,12 @@ void cudaLoadCudaFloat2Array( int numberOfParticles, int entriesPerParticle, CUD
@param entriesPerParticle entries/particles array
@param array cuda array
@param outputVector output vector
@param order if set, reorder entries
--------------------------------------------------------------------------------------- */
void cudaLoadCudaFloat4Array( int numberOfParticles, int entriesPerParticle, CUDAStream<float4>* array, VectorOfDoubleVectors& outputVector )
void cudaLoadCudaFloat4Array( int numberOfParticles, int entriesPerParticle, CUDAStream<float4>* array,
VectorOfDoubleVectors& outputVector, int* order )
{
// ---------------------------------------------------------------------------------------
......@@ -3996,21 +4033,27 @@ void cudaLoadCudaFloat4Array( int numberOfParticles, int entriesPerParticle, CUD
array->Download();
int runningIndex = 0;
int orderIndex;
outputVector.resize( numberOfParticles );
for( int ii = 0; ii < numberOfParticles; ii++ ){
if( order ){
orderIndex = order[runningIndex];
} else {
orderIndex = runningIndex;
}
if( entriesPerParticle > 0 ){
outputVector[ii].push_back( array->_pSysStream[0][runningIndex].x );
outputVector[orderIndex].push_back( array->_pSysStream[0][ii].x );
}
if( entriesPerParticle > 1 ){
outputVector[ii].push_back( array->_pSysStream[0][runningIndex].y );
outputVector[orderIndex].push_back( array->_pSysStream[0][ii].y );
}
if( entriesPerParticle > 2 ){
outputVector[ii].push_back( array->_pSysStream[0][runningIndex].z );
outputVector[orderIndex].push_back( array->_pSysStream[0][ii].z );
}
if( entriesPerParticle > 3 ){
outputVector[ii].push_back( array->_pSysStream[0][runningIndex].w );
outputVector[orderIndex].push_back( array->_pSysStream[0][ii].w );
}
runningIndex++;
}
......@@ -4353,8 +4396,8 @@ void trackMutualInducedIterations( amoebaGpuContext amoebaGpu, int iteration){
}
(void) fflush( amoebaGpu->log );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psVelm4, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, NULL );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psVelm4, outputVector, NULL );
/*
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector );
......
plugins/amoeba/platforms/cuda/src/kernels/amoebaCudaKernels.h
View file @ 4d763c4d
......@@ -147,9 +147,9 @@ extern void cudaWriteFloat1AndFloat1ArraysToFile( int numberOfAtoms, char* fname
int entriesPerAtom2, CUDAStream<float>* array2 );
extern void readFile( std::string fileName, StringVectorVector& fileContents );
extern void cudaLoadCudaFloatArray( int numberOfParticles, int entriesPerParticle, CUDAStream<float>* array, VectorOfDoubleVectors& outputVector );
extern void cudaLoadCudaFloatArray( int numberOfParticles, int entriesPerParticle, CUDAStream<float>* array, VectorOfDoubleVectors& outputVector, int* order );
extern void cudaLoadCudaFloat2Array( int numberOfParticles, int entriesPerParticle, CUDAStream<float2>* array, VectorOfDoubleVectors& outputVector );
extern void cudaLoadCudaFloat4Array( int numberOfParticles, int entriesPerParticle, CUDAStream<float4>* array, VectorOfDoubleVectors& outputVector );
extern void cudaLoadCudaFloat4Array( int numberOfParticles, int entriesPerParticle, CUDAStream<float4>* array, VectorOfDoubleVectors& outputVector, int* order );
extern void cudaWriteVectorOfDoubleVectorsToFile( char* fname, std::vector<int>& fileId, VectorOfDoubleVectors& outputVector );
extern void kClearFloat( amoebaGpuContext amoebaGpu, unsigned int entries, CUDAStream<float>* fieldToClear );
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaElectrostatic.cu
View file @ 4d763c4d
......@@ -947,9 +947,9 @@ void cudaComputeAmoebaElectrostatic( amoebaGpuContext amoebaGpu )
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psForce, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector);
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psForce, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector, NULL);
cudaWriteVectorOfDoubleVectorsToFile( "CudaForceTorque", fileId, outputVector );
}
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaFixedEAndGkFields.cu
View file @ 4d763c4d
......@@ -562,9 +562,9 @@ void cudaComputeAmoebaFixedEAndGkFields( amoebaGpuContext amoebaGpu )
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector);
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psGk_Field, outputVector);
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector, NULL);
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psGk_Field, outputVector, NULL);
cudaWriteVectorOfDoubleVectorsToFile( "CudaEAndGkField", fileId, outputVector );
}
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaFixedEField.cu
View file @ 4d763c4d
......@@ -307,8 +307,8 @@ void cudaComputeAmoebaFixedEField( amoebaGpuContext amoebaGpu )
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
//cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector);
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector, NULL);
cudaWriteVectorOfDoubleVectorsToFile( "CudaEField", fileId, outputVector );
}
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaKirkwood.cu
View file @ 4d763c4d
......@@ -1748,10 +1748,10 @@ static void kReduceToBornForcePrefactor( amoebaGpuContext amoebaGpu )
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( amoebaGpu->gpuContext->natoms, 3, amoebaGpu->gpuContext->psPosq4, outputVector );
cudaLoadCudaFloatArray( amoebaGpu->gpuContext->natoms, 1, amoebaGpu->gpuContext->psBornRadii, outputVector );
cudaLoadCudaFloat2Array( amoebaGpu->gpuContext->natoms, 2, amoebaGpu->gpuContext->psObcData, outputVector );
cudaLoadCudaFloatArray( amoebaGpu->gpuContext->natoms, 1, amoebaGpu->gpuContext->psBornForce, outputVector );
cudaLoadCudaFloat4Array( amoebaGpu->gpuContext->natoms, 3, amoebaGpu->gpuContext->psPosq4, outputVector, NULL );
cudaLoadCudaFloatArray( amoebaGpu->gpuContext->natoms, 1, amoebaGpu->gpuContext->psBornRadii, outputVector, NULL );
cudaLoadCudaFloat2Array( amoebaGpu->gpuContext->natoms, 2, amoebaGpu->gpuContext->psObcData, outputVector, NULL );
cudaLoadCudaFloatArray( amoebaGpu->gpuContext->natoms, 1, amoebaGpu->gpuContext->psBornForce, outputVector, NULL );
cudaWriteVectorOfDoubleVectorsToFile( "CudaBornForce", fileId, outputVector );
(void) fprintf( amoebaGpu->log, "kReduceToBornForcePrefactor: exiting.\n" );
(void) fprintf( stderr, "kReduceToBornForcePrefactor: exiting.\n" ); (void) fflush( stderr );
......@@ -2069,8 +2069,8 @@ void kCalculateAmoebaKirkwood( amoebaGpuContext amoebaGpu )
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psKirkwoodForce, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector);
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psKirkwoodForce, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector, NULL);
cudaWriteVectorOfDoubleVectorsToFile( "CudaForceTorque", fileId, outputVector );
}
......@@ -2113,8 +2113,8 @@ void kCalculateAmoebaKirkwood( amoebaGpuContext amoebaGpu )
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psKirkwoodForce, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psKirkwoodForce, outputVector, NULL );
cudaWriteVectorOfDoubleVectorsToFile( "CudaKirkwoodForce", fileId, outputVector );
}
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaKirkwoodEDiff.cu
View file @ 4d763c4d
......@@ -1215,9 +1215,9 @@ void kCalculateAmoebaKirkwoodEDiff( amoebaGpuContext amoebaGpu )
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psKirkwoodEDiffForce, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector);
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psKirkwoodEDiffForce, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector, NULL);
cudaWriteVectorOfDoubleVectorsToFile( "CudaForceTorque", fileId, outputVector );
}
......@@ -1260,8 +1260,8 @@ void kCalculateAmoebaKirkwoodEDiff( amoebaGpuContext amoebaGpu )
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psKirkwoodEDiffForce, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psKirkwoodEDiffForce, outputVector, NULL );
cudaWriteVectorOfDoubleVectorsToFile( "CudaKirkwoodEDiffForce", fileId, outputVector );
}
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaMapTorques.cu
View file @ 4d763c4d
......@@ -829,8 +829,8 @@ void cudaComputeAmoebaMapTorques( amoebaGpuContext amoebaGpu, CUDAStream<float>*
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psForce, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector);
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psForce, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector, NULL);
cudaWriteVectorOfDoubleVectorsToFile( "CudaVacuumElecForce", fileId, outputVector );
}
#endif
......@@ -1031,9 +1031,9 @@ void cudaComputeAmoebaMapTorquesAndAddTotalForce( amoebaGpuContext amoebaGpu,
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
//cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 4, gpu->psForce4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, psForce, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, psTorque, outputVector);
cudaLoadCudaFloat4Array( gpu->natoms, 4, gpu->psForce4, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, psForce, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, psTorque, outputVector, NULL);
cudaWriteVectorOfDoubleVectorsToFile( "CudaVacuumElecForce", fileId, outputVector );
}
#endif
......@@ -1123,9 +1123,9 @@ void cudaComputeAmoebaMapTorquesAndAddTotalForce2( amoebaGpuContext amoebaGpu,
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
//cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 4, gpu->psForce4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psForce, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector);
cudaLoadCudaFloat4Array( gpu->natoms, 4, gpu->psForce4, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psForce, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psTorque, outputVector, NULL);
cudaWriteVectorOfDoubleVectorsToFile( "CudaVacuumElecForce", fileId, outputVector );
}
#endif
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaMutualInducedAndGkFields.cu
View file @ 4d763c4d
......@@ -925,9 +925,9 @@ static void cudaComputeAmoebaMutualInducedAndGkFieldBySOR( amoebaGpuContext amoe
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector, NULL );
cudaWriteVectorOfDoubleVectorsToFile( "CudaMI_GK", fileId, outputVector );
}
#endif
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaMutualInducedField.cu
View file @ 4d763c4d
......@@ -594,8 +594,8 @@ static void cudaComputeAmoebaMutualInducedFieldBySOR( amoebaGpuContext amoebaGpu
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
// cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector, NULL );
cudaWriteVectorOfDoubleVectorsToFile( "CudaMI", fileId, outputVector );
}
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaPmeDirectElectrostatic.h
View file @ 4d763c4d
......@@ -43,7 +43,7 @@ void METHOD_NAME(kCalculateAmoebaPmeDirectElectrostatic, Forces_kernel)(
){
#ifdef AMOEBA_DEBUG
int maxPullIndex = 2;
int maxPullIndex = 7;
float4 pullBack[12];
#endif
......@@ -56,6 +56,7 @@ void METHOD_NAME(kCalculateAmoebaPmeDirectElectrostatic, Forces_kernel)(
unsigned int end = (warp+1)*numWorkUnits/totalWarps;
unsigned int lasty = 0xFFFFFFFF;
float totalEnergy = 0.0f;
float4 forceTorqueEnergy[3];
float scalingFactors[LastScalingIndex];
......@@ -90,10 +91,7 @@ void METHOD_NAME(kCalculateAmoebaPmeDirectElectrostatic, Forces_kernel)(
localParticle.torque[1] = 0.0f;
localParticle.torque[2] = 0.0f;
scalingFactors[PScaleIndex] = 1.0f;
scalingFactors[DScaleIndex] = 1.0f;
scalingFactors[UScaleIndex] = 1.0f;
scalingFactors[MScaleIndex] = 1.0f;
if (x == y) // Handle diagonals uniquely at 50% efficiency
{
......@@ -116,9 +114,6 @@ void METHOD_NAME(kCalculateAmoebaPmeDirectElectrostatic, Forces_kernel)(
for (unsigned int j = 0; j < GRID; j++)
{
float force[3];
float3 torque[2];
unsigned int atomJ = y + j;
// set scale factors
......@@ -132,9 +127,7 @@ void METHOD_NAME(kCalculateAmoebaPmeDirectElectrostatic, Forces_kernel)(
// force
float energy;
calculatePmeDirectElectrostaticPairIxn_kernel( localParticle, psA[j],
scalingFactors, force, torque, &energy
calculatePmeDirectElectrostaticPairIxn_kernel( localParticle, psA[j], scalingFactors, forceTorqueEnergy
#ifdef AMOEBA_DEBUG
, pullBack
#endif
......@@ -143,28 +136,25 @@ void METHOD_NAME(kCalculateAmoebaPmeDirectElectrostatic, Forces_kernel)(
// nan*0.0 = nan not 0.0, so explicitly exclude (atomI == atomJ) contribution
// by setting match flag
unsigned int mask = ( (atomI == atomJ) || (atomI >= cAmoebaSim.numberOfAtoms) || (atomJ >= cAmoebaSim.numberOfAtoms) ) ? 0 : 1;
// add to field at atomI the field due atomJ's charge/dipole/quadrupole
if( (atomI != atomJ) && (atomI < cAmoebaSim.numberOfAtoms) && (atomJ < cAmoebaSim.numberOfAtoms) )
{
localParticle.force[0] += forceTorqueEnergy[0].x;
localParticle.force[1] += forceTorqueEnergy[0].y;
localParticle.force[2] += forceTorqueEnergy[0].z;
localParticle.force[0] += mask ? force[0] : 0.0f;
localParticle.force[1] += mask ? force[1] : 0.0f;
localParticle.force[2] += mask ? force[2] : 0.0f;
localParticle.torque[0] += forceTorqueEnergy[1].x;
localParticle.torque[1] += forceTorqueEnergy[1].y;
localParticle.torque[2] += forceTorqueEnergy[1].z;
localParticle.torque[0] += mask ? torque[0].x : 0.0f;
localParticle.torque[1] += mask ? torque[0].y : 0.0f;
localParticle.torque[2] += mask ? torque[0].z : 0.0f;
// energy for each diagonal-block ixn included twice, hence factor of 0.5
totalEnergy += mask ? 0.5*energy : 0.0f;
totalEnergy += 0.5*forceTorqueEnergy[0].w;
}
#ifdef AMOEBA_DEBUG
/*
energy = mask ? 0.5*energy : 0.0f;
if( atomI < 200 && (fabs( energy ) > 1.0e+8 || energy != energy) ){
debugSetup( atomI, atomJ, debugArray, pullBack );
} */
if( atomI == targetAtom || atomJ == targetAtom ){
if( atomI == targetAtom ){
unsigned int mask = ( (atomI == atomJ) || (atomI >= cAmoebaSim.numberOfAtoms) || (atomJ >= cAmoebaSim.numberOfAtoms) ) ? 0 : 1;
unsigned int index = (atomI == targetAtom) ? atomJ : atomI;
float blockId = 1.0f;
......@@ -174,23 +164,24 @@ if( atomI == targetAtom ){
debugArray[index].w = blockId;
index += cAmoebaSim.paddedNumberOfAtoms;
debugArray[index].x = mask ? force[0] : 0.0f;
debugArray[index].y = mask ? force[1] : 0.0f;
debugArray[index].z = mask ? force[2] : 0.0f;
debugArray[index].w = blockId;
debugArray[index].x = mask ? forceTorqueEnergy[0].x : 0.0f;
debugArray[index].y = mask ? forceTorqueEnergy[0].y : 0.0f;
debugArray[index].z = mask ? forceTorqueEnergy[0].z : 0.0f;
debugArray[index].w = mask ? forceTorqueEnergy[0].w : 0.0f;
index += cAmoebaSim.paddedNumberOfAtoms;
debugArray[index].x = mask ? torque[0].x : 0.0f;
debugArray[index].y = mask ? torque[0].y : 0.0f;
debugArray[index].z = mask ? torque[0].z : 0.0f;
debugArray[index].w = mask ? energy : 0.0f;
debugArray[index].x = mask ? forceTorqueEnergy[1].x : 0.0f;
debugArray[index].y = mask ? forceTorqueEnergy[1].y : 0.0f;
debugArray[index].z = mask ? forceTorqueEnergy[1].z : 0.0f;
float offsetF = (float)(3*(x + tgx + (x >> GRIDBITS) * cAmoebaSim.paddedNumberOfAtoms));
debugArray[index].w = offsetF;
index += cAmoebaSim.paddedNumberOfAtoms;
debugArray[index].x = mask ? torque[0].x : 0.0f;
debugArray[index].y = mask ? torque[0].y : 0.0f;
debugArray[index].z = mask ? torque[0].z : 0.0f;
debugArray[index].w = (float) (blockIdx.x * blockDim.x + threadIdx.x);
debugArray[index].x = mask ? forceTorqueEnergy[2].x : 0.0f;
debugArray[index].y = mask ? forceTorqueEnergy[2].y : 0.0f;
debugArray[index].z = mask ? forceTorqueEnergy[2].z : 0.0f;
debugArray[index].w = offsetF;
for( int pullIndex = 0; pullIndex < maxPullIndex; pullIndex++ ){
index += cAmoebaSim.paddedNumberOfAtoms;
......@@ -290,14 +281,11 @@ if( atomI == targetAtom ){
for (unsigned int j = 0; j < GRID; j++)
{
if ((flags&(1<<j)) != 0)
if( (flags & (1<<j) ) != 0)
{
unsigned int jIdx = (flags == 0xFFFFFFFF) ? tj : j;
unsigned int atomJ = y + jIdx;
float force[3];
float3 torque[2];
// set scale factors
if( bExclusionFlag )
......@@ -309,9 +297,8 @@ if( atomI == targetAtom ){
// force
float energy;
calculatePmeDirectElectrostaticPairIxn_kernel( localParticle, psA[jIdx],
scalingFactors, force, torque, &energy
scalingFactors, forceTorqueEnergy
#ifdef AMOEBA_DEBUG
, pullBack
#endif
......@@ -319,41 +306,41 @@ if( atomI == targetAtom ){
// check if atoms out-of-bounds
unsigned int mask = ( (atomI >= cAmoebaSim.numberOfAtoms) || (atomJ >= cAmoebaSim.numberOfAtoms) ) ? 0 : 1;
if( (atomI < cAmoebaSim.numberOfAtoms) && (atomJ < cAmoebaSim.numberOfAtoms) )
{
// add force and torque to atom I due atom J
localParticle.force[0] += mask ? force[0] : 0.0f;
localParticle.force[1] += mask ? force[1] : 0.0f;
localParticle.force[2] += mask ? force[2] : 0.0f;
localParticle.force[0] += forceTorqueEnergy[0].x;
localParticle.force[1] += forceTorqueEnergy[0].y;
localParticle.force[2] += forceTorqueEnergy[0].z;
localParticle.torque[0] += mask ? torque[0].x : 0.0f;
localParticle.torque[1] += mask ? torque[0].y : 0.0f;
localParticle.torque[2] += mask ? torque[0].z : 0.0f;
totalEnergy += forceTorqueEnergy[0].w;
totalEnergy += mask ? energy : 0.0f;
localParticle.torque[0] += forceTorqueEnergy[1].x;
localParticle.torque[1] += forceTorqueEnergy[1].y;
localParticle.torque[2] += forceTorqueEnergy[1].z;
// add force and torque to atom J due atom I
if( flags == 0xFFFFFFFF ){
psA[jIdx].force[0] -= mask ? force[0] : 0.0f;
psA[jIdx].force[1] -= mask ? force[1] : 0.0f;
psA[jIdx].force[2] -= mask ? force[2] : 0.0f;
psA[jIdx].force[0] -= forceTorqueEnergy[0].x;
psA[jIdx].force[1] -= forceTorqueEnergy[0].y;
psA[jIdx].force[2] -= forceTorqueEnergy[0].z;
psA[jIdx].torque[0] += mask ? torque[1].x : 0.0f;
psA[jIdx].torque[1] += mask ? torque[1].y : 0.0f;
psA[jIdx].torque[2] += mask ? torque[1].z : 0.0f;
psA[jIdx].torque[0] += forceTorqueEnergy[2].x;
psA[jIdx].torque[1] += forceTorqueEnergy[2].y;
psA[jIdx].torque[2] += forceTorqueEnergy[2].z;
} else {
sA[threadIdx.x].tempForce[0] = mask ? 0.0f : force[0];
sA[threadIdx.x].tempForce[1] = mask ? 0.0f : force[1];
sA[threadIdx.x].tempForce[2] = mask ? 0.0f : force[2];
sA[threadIdx.x].tempForce[0] = forceTorqueEnergy[0].x;
sA[threadIdx.x].tempForce[1] = forceTorqueEnergy[1].y;
sA[threadIdx.x].tempForce[2] = forceTorqueEnergy[2].z;
sA[threadIdx.x].tempTorque[0] = mask ? 0.0f : torque[1].x;
sA[threadIdx.x].tempTorque[1] = mask ? 0.0f : torque[1].y;
sA[threadIdx.x].tempTorque[2] = mask ? 0.0f : torque[1].z;
sA[threadIdx.x].tempTorque[0] = forceTorqueEnergy[2].x;
sA[threadIdx.x].tempTorque[1] = forceTorqueEnergy[2].y;
sA[threadIdx.x].tempTorque[2] = forceTorqueEnergy[2].z;
if( tgx % 2 == 0 ){
sumTempBuffer( sA[threadIdx.x], sA[threadIdx.x+1] );
......@@ -379,15 +366,58 @@ if( atomI == targetAtom ){
psA[jIdx].torque[2] += sA[threadIdx.x].tempTorque[2] + sA[threadIdx.x+16].tempTorque[2];
}
}
}
} // end of atoms out-of-bounds
} // end of flags&(1<<j block
#ifdef AMOEBA_DEBUG
unsigned int jIdx = (flags == 0xFFFFFFFF) ? tj : j;
unsigned int atomJ = y + jIdx;
unsigned int mask = ( (atomI >= cAmoebaSim.numberOfAtoms) || (atomJ >= cAmoebaSim.numberOfAtoms) ) ? 0 : 1;
if( atomI == targetAtom || atomJ == targetAtom ){
unsigned int index = (atomI == targetAtom) ? atomJ : atomI;
debugArray[index].x = (float) atomI;
debugArray[index].y = (float) atomJ;
debugArray[index].z = (float) y;
debugArray[index].w = (flags == 0xFFFFFFFF) ? (float) -141.0f : -151.0f;
index += cAmoebaSim.paddedNumberOfAtoms;
debugArray[index].x = mask ? forceTorqueEnergy[0].x : 0.0f;
debugArray[index].y = mask ? forceTorqueEnergy[0].y : 0.0f;
debugArray[index].z = mask ? forceTorqueEnergy[0].z : 0.0f;
debugArray[index].w = mask ? forceTorqueEnergy[0].w : 0.0f;
index += cAmoebaSim.paddedNumberOfAtoms;
debugArray[index].x = mask ? forceTorqueEnergy[1].x : 0.0f;
debugArray[index].y = mask ? forceTorqueEnergy[1].y : 0.0f;
debugArray[index].z = mask ? forceTorqueEnergy[1].z : 0.0f;
float offsetF = (float)(3*(y + tgx + (x >> GRIDBITS) * cAmoebaSim.paddedNumberOfAtoms));
debugArray[index].w = offsetF;
index += cAmoebaSim.paddedNumberOfAtoms;
debugArray[index].x = mask ? forceTorqueEnergy[2].x : 0.0f;
debugArray[index].y = mask ? forceTorqueEnergy[2].y : 0.0f;
debugArray[index].z = mask ? forceTorqueEnergy[2].z : 0.0f;
offsetF = (float) (3*(x + tgx + (y >> GRIDBITS) * cAmoebaSim.paddedNumberOfAtoms));
debugArray[index].w = offsetF;
for( int pullIndex = 0; pullIndex < maxPullIndex; pullIndex++ ){
index += cAmoebaSim.paddedNumberOfAtoms;
debugArray[index].x = pullBack[pullIndex].x;
debugArray[index].y = pullBack[pullIndex].y;
debugArray[index].z = pullBack[pullIndex].z;
debugArray[index].w = pullBack[pullIndex].w;
}
}
#endif
tj = (tj + 1) & (GRID - 1);
} // end of j-loop
// Write results
#ifdef USE_OUTPUT_BUFFER_PER_WARP
#ifdef USE_OUTPUT_BUFFER_PER_WARP
float of;
unsigned int offset = 3*(x + tgx + warp*cAmoebaSim.paddedNumberOfAtoms);
......@@ -441,7 +471,7 @@ if( atomI == targetAtom ){
of += sA[threadIdx.x].torque[2];
outputTorque[offset+2] = of;
#else
#else
unsigned int offset = 3*(x + tgx + (y >> GRIDBITS) * cAmoebaSim.paddedNumberOfAtoms);
outputForce[offset] = localParticle.force[0];
......@@ -462,7 +492,7 @@ if( atomI == targetAtom ){
outputTorque[offset+1] = sA[threadIdx.x].torque[1];
outputTorque[offset+2] = sA[threadIdx.x].torque[2];
#endif
#endif
lasty = y;
} // end of pInteractionFlag block
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaPmeFixedEField.cu
View file @ 4d763c4d
......@@ -565,9 +565,9 @@ if( fabs(debugArray->_pSysStream[0][jj+3*paddedNumberOfAtoms].x) > 0.0 ){
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector);
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector, gpu->psAtomIndex->_pSysData );
cudaWriteVectorOfDoubleVectorsToFile( "CudaEField", fileId, outputVector );
}
delete debugArray;
......@@ -578,9 +578,9 @@ if( fabs(debugArray->_pSysStream[0][jj+3*paddedNumberOfAtoms].x) > 0.0 ){
std::vector<int> fileId;
fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector);
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector, gpu->psAtomIndex->_pSysData);
cudaWriteVectorOfDoubleVectorsToFile( "CudaEField", fileId, outputVector );
}
......@@ -590,4 +590,15 @@ void cudaComputeAmoebaPmeFixedEField( amoebaGpuContext amoebaGpu )
{
kCalculateAmoebaPMEFixedMultipoles( amoebaGpu );
cudaComputeAmoebaPmeDirectFixedEField( amoebaGpu );
if( 0 ){
gpuContext gpu = amoebaGpu->gpuContext;
std::vector<int> fileId;
fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_Field, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psE_FieldPolar, outputVector, gpu->psAtomIndex->_pSysData );
cudaWriteVectorOfDoubleVectorsToFile( "CudaEField", fileId, outputVector );
}
}
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaPmeMutualInducedField.cu
View file @ 4d763c4d
......@@ -193,6 +193,7 @@ __device__ void calculatePmeDirectMutualInducedFieldPairIxn_kernel( MutualInduce
fields[2].z = 0.0f;
fields[2].w = 0.0f;
}
/*
#ifdef AMOEBA_DEBUG
pullBack[0].x = xr;
pullBack[0].y = yr;
......@@ -204,7 +205,6 @@ __device__ void calculatePmeDirectMutualInducedFieldPairIxn_kernel( MutualInduce
pullBack[1].z = bn2;
pullBack[1].w = exp2a;
/*
pullBack[1].x = atomJ.x - atomI.x;
pullBack[1].y = atomJ.y - atomI.y;
pullBack[1].z = atomJ.z - atomI.z;
......@@ -212,8 +212,8 @@ __device__ void calculatePmeDirectMutualInducedFieldPairIxn_kernel( MutualInduce
pullBack[1].x = scale3;
pullBack[1].y = scale5;
pullBack[1].z = scale7;
*/
#endif
*/
}
// Include versions of the kernels for N^2 calculations.
......@@ -385,8 +385,7 @@ static void cudaComputeAmoebaPmeMutualInducedFieldMatrixMultiply( amoebaGpuConte
static const char* methodName = "cudaComputeAmoebaPmeMutualInducedFieldMatrixMultiply";
static int iteration = 1;
if( 1 && amoebaGpu->log ){
(void) fprintf( amoebaGpu->log, "%s: scalingDistanceCutoff=%.5f\n",
methodName, amoebaGpu->scalingDistanceCutoff );
(void) fprintf( amoebaGpu->log, "%s\n", methodName );
(void) fflush( amoebaGpu->log );
}
int paddedNumberOfAtoms = amoebaGpu->gpuContext->sim.paddedNumberOfAtoms;
......@@ -748,9 +747,9 @@ static void cudaComputeAmoebaPmeMutualInducedFieldBySOR( amoebaGpuContext amoeba
std::vector<int> fileId;
fileId.push_back( iteration );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector, gpu->psAtomIndex->_pSysData );
cudaWriteVectorOfDoubleVectorsToFile( "CudaPmeMI", fileId, outputVector );
}
......@@ -780,9 +779,9 @@ fflush( amoebaGpu->log );
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector );
//cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipole, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( gpu->natoms, 3, amoebaGpu->psInducedDipolePolar, outputVector, gpu->psAtomIndex->_pSysData );
cudaWriteVectorOfDoubleVectorsToFile( "CudaPmeMI", fileId, outputVector );
}
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaPmeMutualInducedField.h
View file @ 4d763c4d
......@@ -122,6 +122,7 @@ void METHOD_NAME(kCalculateAmoebaPmeMutualInducedField, _kernel)(
fieldPolarSum[1] += mask ? ijField[1].z : 0.0f;
fieldPolarSum[2] += mask ? ijField[2].z : 0.0f;
/*
#ifdef AMOEBA_DEBUG
if( atomI == targetAtom || (y+j) == targetAtom ){
unsigned int index = atomI == targetAtom ? (y+j) : atomI;
......@@ -173,7 +174,6 @@ if( atomI == targetAtom || (y+j) == targetAtom ){
debugArray[index].z = ijField[indexJ+1][2];
debugArray[index].w = flag;
/*
index += cAmoebaSim.paddedNumberOfAtoms;
index += cAmoebaSim.paddedNumberOfAtoms;
......@@ -189,10 +189,10 @@ if( atomI == targetAtom || (y+j) == targetAtom ){
debugArray[index].y = scaleMask.x & mask ? 1.0f : -1.0f;
debugArray[index].z = scaleMask.y & mask ? 1.0f : -1.0f;
debugArray[index].w = + 10.0f;
*/
}
#endif
*/
}
// Write results
......@@ -308,6 +308,7 @@ if( atomI == targetAtom || (y+j) == targetAtom ){
}
/*
#ifdef AMOEBA_DEBUG
if( atomI == targetAtom || (y+jIdx) == targetAtom ){
unsigned int index = atomI == targetAtom ? (y+jIdx) : atomI;
......@@ -360,6 +361,7 @@ if( atomI == targetAtom || (y+jIdx) == targetAtom ){
debugArray[index].w = flag;
}
#endif
*/
}
tj = (tj + 1) & (GRID - 1);
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaRotateFrame.cu
View file @ 4d763c4d
......@@ -528,8 +528,8 @@ void cudaComputeAmoebaLabFrameMoments( amoebaGpuContext amoebaGpu )
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( particles, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( particles, 9, amoebaGpu->psRotationMatrix, outputVector );
cudaLoadCudaFloat4Array( particles, 3, gpu->psPosq4, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( particles, 9, amoebaGpu->psRotationMatrix, outputVector, gpu->psAtomIndex->_pSysData );
cudaWriteVectorOfDoubleVectorsToFile( "CudaRotationMatrices", fileId, outputVector );
}
if( 0 ){
......@@ -539,9 +539,9 @@ void cudaComputeAmoebaLabFrameMoments( amoebaGpuContext amoebaGpu )
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( particles, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( particles, 3, amoebaGpu->psLabFrameDipole, outputVector );
cudaLoadCudaFloatArray( particles, 9, amoebaGpu->psLabFrameQuadrupole, outputVector );
cudaLoadCudaFloat4Array( particles, 3, gpu->psPosq4, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( particles, 3, amoebaGpu->psLabFrameDipole, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloatArray( particles, 9, amoebaGpu->psLabFrameQuadrupole, outputVector, gpu->psAtomIndex->_pSysData );
cudaWriteVectorOfDoubleVectorsToFile( "CudaRotatedMoments", fileId, outputVector );
}
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaVdw14_7.cu
View file @ 4d763c4d
......@@ -715,8 +715,8 @@ void kCalculateAmoebaVdw14_7Forces( amoebaGpuContext amoebaGpu, int applyCutoff
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, psTempForce, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, gpu->psAtomIndex->_pSysData );
cudaLoadCudaFloat4Array( gpu->natoms, 3, psTempForce, outputVector, gpu->psAtomIndex->_pSysData );
cudaWriteVectorOfDoubleVectorsToFile( "CudaVdw", fileId, outputVector );
delete psTempForce;
//exit(0);
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaWcaDispersion.cu
View file @ 4d763c4d
......@@ -631,8 +631,8 @@ void kCalculateAmoebaWcaDispersionForces( amoebaGpuContext amoebaGpu )
std::vector<int> fileId;
//fileId.push_back( 0 );
VectorOfDoubleVectors outputVector;
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector );
cudaLoadCudaFloatArray( gpu->natoms, 3, psTempForce, outputVector );
cudaLoadCudaFloat4Array( gpu->natoms, 3, gpu->psPosq4, outputVector, NULL );
cudaLoadCudaFloatArray( gpu->natoms, 3, psTempForce, outputVector, NULL );
cudaWriteVectorOfDoubleVectorsToFile( "CudaWca", fileId, outputVector );
delete psTempForce;
//exit(0);
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment