Skip to content

GitLab

Commit 9298d000 authored by Mark Friedrichs's avatar Mark Friedrichs
Browse files

Fix for particles w/o multipole axis (e.g. ions) 

-This line, and th -se below, will be ignored--

M    amoebaCudaGpu.cpp
M    kCalculateAmoebaCudaMapTorques.cu
M    kCalculateAmoebaCudaRotateFrame.cu
parent 29c09776
Hide whitespace changes
Inline Side-by-side
Showing with 41 additions and 22 deletions
plugins/amoeba/platforms/cuda/src/kernels/amoebaCudaGpu.cpp
View file @ 9298d000
...@@ -1646,8 +1646,8 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect ...@@ -1646,8 +1646,8 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect
amoebaGpu->amoebaSim.dielec = 1.0f; amoebaGpu->amoebaSim.dielec = 1.0f;
} }
static const int maxAxisType = 5; static const int maxAxisType = 6;
int axisTypeCount[maxAxisType+1] = { 0, 0, 0, 0, 0, 0 }; int axisTypeCount[maxAxisType+1] = { 0, 0, 0, 0, 0, 0, 0 };
for( int ii = 0; ii < static_cast<int>(charges.size()); ii++ ){ for( int ii = 0; ii < static_cast<int>(charges.size()); ii++ ){
// axis type & multipole particles ids // axis type & multipole particles ids
...@@ -1666,19 +1666,23 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect ...@@ -1666,19 +1666,23 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect
// and need test system // and need test system
int axisParticleIndex = multipoleParticleZ[ii]; int axisParticleIndex = multipoleParticleZ[ii];
if( maxIndices[axisParticleIndex] < ii ){ if( axisParticleIndex > -1 ){
maxIndices[axisParticleIndex] = ii; if( maxIndices[axisParticleIndex] < ii ){
} maxIndices[axisParticleIndex] = ii;
if( amoebaGpu->psMultipoleAxisOffset->_pSysData[axisParticleIndex] > ii ){ }
amoebaGpu->psMultipoleAxisOffset->_pSysData[axisParticleIndex] = ii; if( amoebaGpu->psMultipoleAxisOffset->_pSysData[axisParticleIndex] > ii ){
amoebaGpu->psMultipoleAxisOffset->_pSysData[axisParticleIndex] = ii;
}
} }
axisParticleIndex = multipoleParticleX[ii]; axisParticleIndex = multipoleParticleX[ii];
if( maxIndices[axisParticleIndex] < ii ){ if( axisParticleIndex > -1 ){
maxIndices[axisParticleIndex] = ii; if( maxIndices[axisParticleIndex] < ii ){
} maxIndices[axisParticleIndex] = ii;
if( amoebaGpu->psMultipoleAxisOffset->_pSysData[axisParticleIndex] > ii ){ }
amoebaGpu->psMultipoleAxisOffset->_pSysData[axisParticleIndex] = ii; if( amoebaGpu->psMultipoleAxisOffset->_pSysData[axisParticleIndex] > ii ){
amoebaGpu->psMultipoleAxisOffset->_pSysData[axisParticleIndex] = ii;
}
} }
axisParticleIndex = multipoleParticleY[ii]; axisParticleIndex = multipoleParticleY[ii];
...@@ -1692,7 +1696,7 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect ...@@ -1692,7 +1696,7 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect
} }
if( 0 && amoebaGpu->log ) if( 0 && amoebaGpu->log )
fprintf( amoebaGpu->log, "Z1 %4d [%4d %4d] %4d %4d %4d %4d %d %d\n", ii, fprintf( amoebaGpu->log, "Z1 %4d %d [%4d %4d] %4d %4d %4d %4d %d %d\n", ii, axisType[ii],
multipoleParticleX[ii], multipoleParticleY[ii], multipoleParticleZ[ii], multipoleParticleX[ii], multipoleParticleY[ii], multipoleParticleZ[ii],
amoebaGpu->psMultipoleAxisOffset->_pSysData[multipoleParticleZ[ii]], amoebaGpu->psMultipoleAxisOffset->_pSysData[multipoleParticleZ[ii]],
maxIndices[multipoleParticleZ[ii]], maxIndices[multipoleParticleZ[ii]],
...@@ -1931,7 +1935,7 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect ...@@ -1931,7 +1935,7 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect
} }
if( amoebaGpu->log ){ if( amoebaGpu->log ){
std::string axisLabel[maxAxisType+1] = { "ZThenX", "Bisector", "ZBisect", "ThreeFold", "ZOnly", "Unknown"}; std::string axisLabel[maxAxisType+1] = { "ZThenX", "Bisector", "ZBisect", "ThreeFold", "ZOnly", "NoAxisType", "Unknown"};
for( unsigned int kk = 0; kk < (maxAxisType+1); kk++ ){ for( unsigned int kk = 0; kk < (maxAxisType+1); kk++ ){
(void) fprintf( amoebaGpu->log, "%2u %10s atom count=%d\n", kk, axisLabel[kk].c_str(), axisTypeCount[kk] ); (void) fprintf( amoebaGpu->log, "%2u %10s atom count=%d\n", kk, axisLabel[kk].c_str(), axisTypeCount[kk] );
} }
...@@ -1969,6 +1973,7 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect ...@@ -1969,6 +1973,7 @@ void gpuSetAmoebaMultipoleParameters(amoebaGpuContext amoebaGpu, const std::vect
int diff = maxIndices[ii] - amoebaGpu->psMultipoleAxisOffset->_pSysData[ii]; int diff = maxIndices[ii] - amoebaGpu->psMultipoleAxisOffset->_pSysData[ii];
if( diff > amoebaGpu->maxMapTorqueDifference ){ if( diff > amoebaGpu->maxMapTorqueDifference ){
//fprintf( stderr, "maxMapTorqueDifference %d maxInd=%d off=%d diff=%d\n", ii, maxIndices[ii], amoebaGpu->psMultipoleAxisOffset->_pSysData[ii], diff );
amoebaGpu->maxMapTorqueDifference = diff; amoebaGpu->maxMapTorqueDifference = diff;
} }
} }
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaMapTorques.cu
View file @ 9298d000
...@@ -101,6 +101,18 @@ void amoebaMapTorqueToForce_kernel( float* torque, int maxDiff, float* tempElecF ...@@ -101,6 +101,18 @@ void amoebaMapTorqueToForce_kernel( float* torque, int maxDiff, float* tempElecF
int axisAtom = multiPoleAtoms[threadId].z; int axisAtom = multiPoleAtoms[threadId].z;
int axisType = multiPoleAtoms[threadId].w; int axisType = multiPoleAtoms[threadId].w;
// NoAxisType
if( axisType == 5 )
{
int min = multiPoleAtomOffset[threadId];
int offset = 3*(threadId*(maxDiff + 1)-min);
tempElecForce[offset ] = 0.0f;
tempElecForce[offset+1] = 0.0f;
tempElecForce[offset+2] = 0.0f;
return;
}
vector[U][0] = atomCoord[threadId].x - atomCoord[axisAtom].x; vector[U][0] = atomCoord[threadId].x - atomCoord[axisAtom].x;
vector[U][1] = atomCoord[threadId].y - atomCoord[axisAtom].y; vector[U][1] = atomCoord[threadId].y - atomCoord[axisAtom].y;
vector[U][2] = atomCoord[threadId].z - atomCoord[axisAtom].z; vector[U][2] = atomCoord[threadId].z - atomCoord[axisAtom].z;
...@@ -689,7 +701,7 @@ void cudaComputeAmoebaMapTorques( amoebaGpuContext amoebaGpu, CUDAStream<float>* ...@@ -689,7 +701,7 @@ void cudaComputeAmoebaMapTorques( amoebaGpuContext amoebaGpu, CUDAStream<float>*
// check that BLOCK_SIZE >= amoebaGpu->maxMapTorqueDifference // check that BLOCK_SIZE >= amoebaGpu->maxMapTorqueDifference
if( amoebaGpu->maxMapTorqueDifference > BLOCK_SIZE ){ if( amoebaGpu->maxMapTorqueDifference > BLOCK_SIZE ){
(void) fprintf( amoebaGpu->log, "block size (%d) in amoebaMapTorqueReduce_kernel is too small ( > %d)! -- aborting.\n", (void) fprintf( amoebaGpu->log, "cudaComputeAmoebaMapTorques: block size (%d) in amoebaMapTorqueReduce_kernel is too small ( > %d)! -- aborting.\n",
BLOCK_SIZE, amoebaGpu->maxMapTorqueDifference ); BLOCK_SIZE, amoebaGpu->maxMapTorqueDifference );
exit(-1); exit(-1);
} }
...@@ -859,7 +871,7 @@ void cudaComputeAmoebaMapTorquesAndAddTotalForce( amoebaGpuContext amoebaGpu, ...@@ -859,7 +871,7 @@ void cudaComputeAmoebaMapTorquesAndAddTotalForce( amoebaGpuContext amoebaGpu,
// check that BLOCK_SIZE >= amoebaGpu->maxMapTorqueDifference // check that BLOCK_SIZE >= amoebaGpu->maxMapTorqueDifference
if( amoebaGpu->maxMapTorqueDifference > BLOCK_SIZE ){ if( amoebaGpu->maxMapTorqueDifference > BLOCK_SIZE ){
(void) fprintf( amoebaGpu->log, "block size (%d) in amoebaMapTorqueReduce_kernel is too small ( > %d)! -- aborting.\n", (void) fprintf( amoebaGpu->log, "cudaComputeAmoebaMapTorquesAndAddTotalForce: block size (%d) in amoebaMapTorqueReduce_kernel is too small ( > %d)! -- aborting.\n",
BLOCK_SIZE, amoebaGpu->maxMapTorqueDifference ); BLOCK_SIZE, amoebaGpu->maxMapTorqueDifference );
exit(-1); exit(-1);
} }
...@@ -1061,7 +1073,7 @@ void cudaComputeAmoebaMapTorquesAndAddTotalForce2( amoebaGpuContext amoebaGpu, ...@@ -1061,7 +1073,7 @@ void cudaComputeAmoebaMapTorquesAndAddTotalForce2( amoebaGpuContext amoebaGpu,
// check that BLOCK_SIZE >= amoebaGpu->maxMapTorqueDifference // check that BLOCK_SIZE >= amoebaGpu->maxMapTorqueDifference
if( amoebaGpu->maxMapTorqueDifference > BLOCK_SIZE ){ if( amoebaGpu->maxMapTorqueDifference > BLOCK_SIZE ){
(void) fprintf( amoebaGpu->log, "block size (%d) in amoebaMapTorqueReduce_kernel is too small ( > %d)! -- aborting.\n", (void) fprintf( amoebaGpu->log, "cudaComputeAmoebaMapTorquesAndAddTotalForce2: block size (%d) in amoebaMapTorqueReduce_kernel is too small ( > %d)! -- aborting.\n",
BLOCK_SIZE, amoebaGpu->maxMapTorqueDifference ); BLOCK_SIZE, amoebaGpu->maxMapTorqueDifference );
exit(-1); exit(-1);
} }
......
plugins/amoeba/platforms/cuda/src/kernels/kCalculateAmoebaCudaRotateFrame.cu
View file @ 9298d000
...@@ -53,10 +53,12 @@ __device__ static float normVector3( float* vector ) ...@@ -53,10 +53,12 @@ __device__ static float normVector3( float* vector )
#undef AMOEBA_DEBUG #undef AMOEBA_DEBUG
// ZThenX == 0 // ZThenX == 0
// Bisector == 1 // Bisector == 1
// ZBisect == 2 // ZBisect == 2
// ThreeFold == 3 // ThreeFold == 3
// ZOnly == 4
// NoAxisType == 5
__global__ __global__
#if (__CUDA_ARCH__ >= 200) #if (__CUDA_ARCH__ >= 200)
...@@ -293,7 +295,7 @@ void kCudaComputeLabFrameMoments_kernel( void ) ...@@ -293,7 +295,7 @@ void kCudaComputeLabFrameMoments_kernel( void )
sum = normVector3( vectorZ ); sum = normVector3( vectorZ );
} }
} else if( axisType == 4 ){ } else if( axisType >= 4 ){
vectorX[0] = 0.1f; vectorX[0] = 0.1f;
vectorX[1] = 0.1f; vectorX[1] = 0.1f;
......
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