Skip to content

GitLab

Commit 3b6925ae authored by Andy Simmonett's avatar Andy Simmonett Committed by GitHub
Browse files

Merge pull request #1 from peastman/ljpme 

Cleanup to LJ PME code
parents 5a8a8aa9 f7a102fb
Hide whitespace changes
Inline Side-by-side
Showing with 143 additions and 77 deletions
platforms/opencl/src/OpenCLSort.cpp
View file @ 3b6925ae
......@@ -124,8 +124,9 @@ void OpenCLSort::sort(OpenCLArray& data) {
computeRangeKernel.setArg<cl_uint>(1, data.getSize());
computeRangeKernel.setArg<cl::Buffer>(2, dataRange->getDeviceBuffer());
computeRangeKernel.setArg(3, rangeKernelSize*trait->getKeySize(), NULL);
computeRangeKernel.setArg<cl_int>(4, numBuckets);
computeRangeKernel.setArg<cl::Buffer>(5, bucketOffset->getDeviceBuffer());
computeRangeKernel.setArg(4, rangeKernelSize*trait->getKeySize(), NULL);
computeRangeKernel.setArg<cl_int>(5, numBuckets);
computeRangeKernel.setArg<cl::Buffer>(6, bucketOffset->getDeviceBuffer());
context.executeKernel(computeRangeKernel, rangeKernelSize, rangeKernelSize);
// Assign array elements to buckets.
......
platforms/opencl/src/kernels/customGBEnergyN2.cl
View file @ 3b6925ae
......@@ -18,7 +18,7 @@ __kernel void computeN2Energy(
#endif
__global mixed* restrict energyBuffer, __local real4* restrict local_force,
__global const real4* restrict posq, __local real4* restrict local_posq, __global const unsigned int* restrict exclusions,
__global const ushort2* exclusionTiles,
__global const ushort2* exclusionTiles, int needEnergy,
#ifdef USE_CUTOFF
__global const int* restrict tiles, __global const unsigned int* restrict interactionCount, real4 periodicBoxSize, real4 invPeriodicBoxSize,
real4 periodicBoxVecX, real4 periodicBoxVecY, real4 periodicBoxVecZ, unsigned int maxTiles, __global const real4* restrict blockCenter,
......@@ -82,7 +82,8 @@ __kernel void computeN2Energy(
COMPUTE_INTERACTION
dEdR /= -r;
}
energy += 0.5f*tempEnergy;
if (needEnergy)
energy += 0.5f*tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
#ifdef USE_CUTOFF
......@@ -133,7 +134,8 @@ __kernel void computeN2Energy(
COMPUTE_INTERACTION
dEdR /= -r;
}
energy += tempEnergy;
if (needEnergy)
energy += tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
atom2 = tbx+tj;
......@@ -250,7 +252,7 @@ __kernel void computeN2Energy(
LOAD_ATOM1_PARAMETERS
const unsigned int localAtomIndex = get_local_id(0);
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+tgx] : y*TILE_SIZE + tgx);
unsigned int j = interactingAtoms[pos*TILE_SIZE+tgx];
#else
unsigned int j = y*TILE_SIZE + tgx;
#endif
......@@ -289,7 +291,8 @@ __kernel void computeN2Energy(
COMPUTE_INTERACTION
dEdR /= -r;
}
energy += tempEnergy;
if (needEnergy)
energy += tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
atom2 = tbx+tj;
......@@ -328,7 +331,8 @@ __kernel void computeN2Energy(
COMPUTE_INTERACTION
dEdR /= -r;
}
energy += tempEnergy;
if (needEnergy)
energy += tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
atom2 = tbx+tj;
......
platforms/opencl/src/kernels/customGBEnergyN2_cpu.cl
View file @ 3b6925ae
......@@ -18,7 +18,7 @@ __kernel void computeN2Energy(
#endif
__global mixed* restrict energyBuffer, __local real4* restrict local_force,
__global const real4* restrict posq, __local real4* restrict local_posq, __global const unsigned int* restrict exclusions,
__global const ushort2* exclusionTiles,
__global const ushort2* exclusionTiles, int needEnergy,
#ifdef USE_CUTOFF
__global const int* restrict tiles, __global const unsigned int* restrict interactionCount, real4 periodicBoxSize, real4 invPeriodicBoxSize,
real4 periodicBoxVecX, real4 periodicBoxVecY, real4 periodicBoxVecZ, unsigned int maxTiles, __global const real4* restrict blockCenter,
......@@ -255,7 +255,7 @@ __kernel void computeN2Energy(
for (int localAtomIndex = 0; localAtomIndex < TILE_SIZE; localAtomIndex++) {
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+localAtomIndex] : y*TILE_SIZE+localAtomIndex);
unsigned int j = interactingAtoms[pos*TILE_SIZE+localAtomIndex];
#else
unsigned int j = y*TILE_SIZE+localAtomIndex;
#endif
......
platforms/opencl/src/kernels/customGBValueN2.cl
View file @ 3b6925ae
......@@ -229,7 +229,7 @@ __kernel void computeN2Value(__global const real4* restrict posq, __local real4*
LOAD_ATOM1_PARAMETERS
const unsigned int localAtomIndex = get_local_id(0);
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+tgx] : y*TILE_SIZE + tgx);
unsigned int j = interactingAtoms[pos*TILE_SIZE+tgx];
#else
unsigned int j = y*TILE_SIZE + tgx;
#endif
......
platforms/opencl/src/kernels/customGBValueN2_cpu.cl
View file @ 3b6925ae
......@@ -228,7 +228,7 @@ __kernel void computeN2Value(__global const real4* restrict posq, __local real4*
for (int localAtomIndex = 0; localAtomIndex < TILE_SIZE; localAtomIndex++) {
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+localAtomIndex] : y*TILE_SIZE+localAtomIndex);
unsigned int j = interactingAtoms[pos*TILE_SIZE+localAtomIndex];
#else
unsigned int j = y*TILE_SIZE+localAtomIndex;
#endif
......
platforms/opencl/src/kernels/customIntegratorPerDof.cl
View file @ 3b6925ae
......@@ -26,7 +26,7 @@ void storePos(__global real4* restrict posq, __global real4* restrict posqCorrec
__kernel void computePerDof(__global real4* restrict posq, __global real4* restrict posqCorrection, __global mixed4* restrict posDelta,
__global mixed4* restrict velm, __global const real4* restrict force, __global const mixed2* restrict dt, __global const mixed* restrict globals,
__global mixed* restrict sum, __global const float4* restrict gaussianValues, unsigned int gaussianBaseIndex, __global const float4* restrict uniformValues,
const real energy, __global mixed* restrict energyParamDerivs
const mixed energy, __global mixed* restrict energyParamDerivs
PARAMETER_ARGUMENTS) {
mixed stepSize = dt[0].y;
int index = get_global_id(0);
......
platforms/opencl/src/kernels/customNonbonded.cl
View file @ 3b6925ae
......@@ -14,8 +14,10 @@ if (!isExcluded) {
#endif
COMPUTE_FORCE
#if USE_SWITCH
tempForce = tempForce*switchValue - tempEnergy*switchDeriv;
tempEnergy *= switchValue;
tempForce = tempForce*switchValue - customEnergy*switchDeriv;
tempEnergy += customEnergy*switchValue;
#else
tempEnergy += customEnergy;
#endif
dEdR += tempForce*invR;
}
platforms/opencl/src/kernels/findInteractingBlocks.cl
View file @ 3b6925ae
......@@ -27,8 +27,19 @@ __kernel void findBlockBounds(int numAtoms, real4 periodicBoxSize, real4 invPeri
maxPos = max(maxPos, pos);
}
real4 blockSize = 0.5f*(maxPos-minPos);
real4 center = 0.5f*(maxPos+minPos);
center.w = 0;
for (int i = base; i < last; i++) {
pos = posq[i];
real4 delta = posq[i]-center;
#ifdef USE_PERIODIC
APPLY_PERIODIC_TO_DELTA(delta)
#endif
center.w = max(center.w, delta.x*delta.x+delta.y*delta.y+delta.z*delta.z);
}
center.w = sqrt(center.w);
blockBoundingBox[index] = blockSize;
blockCenter[index] = 0.5f*(maxPos+minPos);
blockCenter[index] = center;
sortedBlocks[index] = (real2) (blockSize.x+blockSize.y+blockSize.z, index);
index += get_global_size(0);
base = index*TILE_SIZE;
......@@ -88,13 +99,13 @@ __kernel void findBlocksWithInteractions(real4 periodicBoxSize, real4 invPeriodi
__local real3 posBuffer[GROUP_SIZE];
__local volatile int workgroupTileIndex[GROUP_SIZE/32];
__local bool includeBlockFlags[GROUP_SIZE];
__local short2 atomCountBuffer[GROUP_SIZE];
__local volatile short2 atomCountBuffer[GROUP_SIZE];
__local int* buffer = workgroupBuffer+BUFFER_SIZE*(warpStart/32);
__local int* exclusionsForX = warpExclusions+MAX_EXCLUSIONS*(warpStart/32);
__local volatile int* tileStartIndex = workgroupTileIndex+(warpStart/32);
// Loop over blocks.
for (int block1 = startBlockIndex+warpIndex; block1 < startBlockIndex+numBlocks; block1 += totalWarps) {
// Load data for this block. Note that all threads in a warp are processing the same block.
......@@ -142,10 +153,18 @@ __kernel void findBlocksWithInteractions(real4 periodicBoxSize, real4 invPeriodi
#ifdef USE_PERIODIC
APPLY_PERIODIC_TO_DELTA(blockDelta)
#endif
includeBlock2 &= (blockDelta.x*blockDelta.x+blockDelta.y*blockDelta.y+blockDelta.z*blockDelta.z < (PADDED_CUTOFF+blockCenterX.w+blockCenterY.w)*(PADDED_CUTOFF+blockCenterX.w+blockCenterY.w));
blockDelta.x = max((real) 0, fabs(blockDelta.x)-blockSizeX.x-blockSizeY.x);
blockDelta.y = max((real) 0, fabs(blockDelta.y)-blockSizeX.y-blockSizeY.y);
blockDelta.z = max((real) 0, fabs(blockDelta.z)-blockSizeX.z-blockSizeY.z);
includeBlock2 &= (blockDelta.x*blockDelta.x+blockDelta.y*blockDelta.y+blockDelta.z*blockDelta.z < PADDED_CUTOFF_SQUARED);
#ifdef TRICLINIC
// The calculation to find the nearest periodic copy is only guaranteed to work if the nearest copy is less than half a box width away.
// If there's any possibility we might have missed it, do a detailed check.
if (periodicBoxSize.z/2-blockSizeX.z-blockSizeY.z < PADDED_CUTOFF || periodicBoxSize.y/2-blockSizeX.y-blockSizeY.y < PADDED_CUTOFF)
includeBlock2 = true;
#endif
if (includeBlock2) {
unsigned short y = (unsigned short) sortedBlocks[block2].y;
for (int k = 0; k < numExclusions; k++)
......@@ -165,8 +184,7 @@ __kernel void findBlocksWithInteractions(real4 periodicBoxSize, real4 invPeriodi
// Check each atom in block Y for interactions.
int start = y*TILE_SIZE;
int atom2 = start+indexInWarp;
int atom2 = y*TILE_SIZE+indexInWarp;
real3 pos2 = posq[atom2].xyz;
#ifdef USE_PERIODIC
if (singlePeriodicCopy)
......
platforms/opencl/src/kernels/gbsaObc.cl
View file @ 3b6925ae
......@@ -232,7 +232,7 @@ __kernel void computeBornSum(
real4 posq1 = posq[atom1];
float2 params1 = global_params[atom1];
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+tgx] : y*TILE_SIZE + tgx);
unsigned int j = interactingAtoms[pos*TILE_SIZE+tgx];
#else
unsigned int j = y*TILE_SIZE + tgx;
#endif
......@@ -385,7 +385,7 @@ __kernel void computeGBSAForce1(
#else
__global real4* restrict forceBuffers, __global real* restrict global_bornForce,
#endif
__global mixed* restrict energyBuffer, __global const real4* restrict posq, __global const real* restrict global_bornRadii,
__global mixed* restrict energyBuffer, __global const real4* restrict posq, __global const real* restrict global_bornRadii, int needEnergy,
#ifdef USE_CUTOFF
__global const int* restrict tiles, __global const unsigned int* restrict interactionCount, real4 periodicBoxSize, real4 invPeriodicBoxSize,
real4 periodicBoxVecX, real4 periodicBoxVecY, real4 periodicBoxVecZ, unsigned int maxTiles, __global const real4* restrict blockCenter,
......@@ -452,7 +452,8 @@ __kernel void computeGBSAForce1(
if (atom1 != y*TILE_SIZE+j)
tempEnergy -= scaledChargeProduct/CUTOFF;
#endif
energy += 0.5f*tempEnergy;
if (needEnergy)
energy += 0.5f*tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
#ifdef USE_CUTOFF
......@@ -506,7 +507,8 @@ __kernel void computeGBSAForce1(
#ifdef USE_CUTOFF
tempEnergy -= scaledChargeProduct/CUTOFF;
#endif
energy += tempEnergy;
if (needEnergy)
energy += tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
localData[tbx+tj].fx += delta.x;
......@@ -617,7 +619,7 @@ __kernel void computeGBSAForce1(
real4 posq1 = posq[atom1];
real bornRadius1 = global_bornRadii[atom1];
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+tgx] : y*TILE_SIZE + tgx);
unsigned int j = interactingAtoms[pos*TILE_SIZE+tgx];
#else
unsigned int j = y*TILE_SIZE + tgx;
#endif
......@@ -669,7 +671,8 @@ __kernel void computeGBSAForce1(
#ifdef USE_CUTOFF
tempEnergy -= scaledChargeProduct/CUTOFF;
#endif
energy += tempEnergy;
if (needEnergy)
energy += tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
localData[tbx+tj].fx += delta.x;
......@@ -717,7 +720,8 @@ __kernel void computeGBSAForce1(
#ifdef USE_CUTOFF
tempEnergy -= scaledChargeProduct/CUTOFF;
#endif
energy += tempEnergy;
if (needEnergy)
energy += tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
localData[tbx+tj].fx += delta.x;
......
platforms/opencl/src/kernels/gbsaObc_cpu.cl
View file @ 3b6925ae
......@@ -228,7 +228,7 @@ __kernel void computeBornSum(
for (int localAtomIndex = 0; localAtomIndex < TILE_SIZE; localAtomIndex++) {
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+localAtomIndex] : y*TILE_SIZE+localAtomIndex);
unsigned int j = interactingAtoms[pos*TILE_SIZE+localAtomIndex];
#else
unsigned int j = y*TILE_SIZE+localAtomIndex;
#endif
......@@ -407,7 +407,7 @@ __kernel void computeGBSAForce1(
#else
__global real4* restrict forceBuffers, __global real* restrict global_bornForce,
#endif
__global mixed* restrict energyBuffer, __global const real4* restrict posq, __global const real* restrict global_bornRadii,
__global mixed* restrict energyBuffer, __global const real4* restrict posq, __global const real* restrict global_bornRadii, int needEnergy,
#ifdef USE_CUTOFF
__global const int* restrict tiles, __global const unsigned int* restrict interactionCount, real4 periodicBoxSize, real4 invPeriodicBoxSize,
real4 periodicBoxVecX, real4 periodicBoxVecY, real4 periodicBoxVecZ, unsigned int maxTiles, __global const real4* restrict blockCenter,
......@@ -641,7 +641,7 @@ __kernel void computeGBSAForce1(
for (int localAtomIndex = 0; localAtomIndex < TILE_SIZE; localAtomIndex++) {
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+localAtomIndex] : y*TILE_SIZE+localAtomIndex);
unsigned int j = interactingAtoms[pos*TILE_SIZE+localAtomIndex];
#else
unsigned int j = y*TILE_SIZE+localAtomIndex;
#endif
......
platforms/opencl/src/kernels/langevin.cl
View file @ 3b6925ae
......@@ -72,7 +72,7 @@ __kernel void integrateLangevinPart2(__global real4* restrict posq, __global rea
* Select the step size to use for the next step.
*/
__kernel void selectLangevinStepSize(mixed maxStepSize, mixed errorTol, mixed tau, mixed kT, __global mixed2* restrict dt,
__kernel void selectLangevinStepSize(mixed maxStepSize, mixed errorTol, mixed friction, mixed kT, __global mixed2* restrict dt,
__global const mixed4* restrict velm, __global const real4* restrict force, __global mixed* restrict paramBuffer, __local mixed* restrict params, __local mixed* restrict error) {
// Calculate the error.
......@@ -110,9 +110,9 @@ __kernel void selectLangevinStepSize(mixed maxStepSize, mixed errorTol, mixed ta
// Recalculate the integration parameters.
mixed vscale = exp(-newStepSize/tau);
mixed fscale = (1-vscale)*tau;
mixed noisescale = sqrt(2*kT/tau)*sqrt(0.5f*(1-vscale*vscale)*tau);
mixed vscale = exp(-newStepSize*friction);
mixed fscale = (friction == 0 ? newStepSize : (1-vscale)/friction);
mixed noisescale = sqrt(kT*(1-vscale*vscale));
params[VelScale] = vscale;
params[ForceScale] = fscale;
params[NoiseScale] = noisescale;
......
platforms/opencl/src/kernels/nonbonded.cl
View file @ 3b6925ae
......@@ -269,7 +269,7 @@ __kernel void computeNonbonded(
LOAD_ATOM1_PARAMETERS
const unsigned int localAtomIndex = get_local_id(0);
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+tgx] : y*TILE_SIZE + tgx);
unsigned int j = interactingAtoms[pos*TILE_SIZE+tgx];
#else
unsigned int j = y*TILE_SIZE + tgx;
#endif
......
platforms/opencl/src/kernels/nonbonded_cpu.cl
View file @ 3b6925ae
......@@ -269,7 +269,7 @@ __kernel void computeNonbonded(
for (int localAtomIndex = 0; localAtomIndex < TILE_SIZE; localAtomIndex++) {
#ifdef USE_CUTOFF
unsigned int j = (numTiles <= maxTiles ? interactingAtoms[pos*TILE_SIZE+localAtomIndex] : y*TILE_SIZE+localAtomIndex);
unsigned int j = interactingAtoms[pos*TILE_SIZE+localAtomIndex];
#else
unsigned int j = y*TILE_SIZE+localAtomIndex;
#endif
......
platforms/opencl/src/kernels/sort.cl
View file @ 3b6925ae
......@@ -8,7 +8,7 @@ KEY_TYPE getValue(DATA_TYPE value) {
* Sort a list that is short enough to entirely fit in local memory. This is executed as
* a single thread block.
*/
__kernel void sortShortList(__global DATA_TYPE* __restrict__ data, uint length, __local DATA_TYPE* dataBuffer) {
__kernel void sortShortList(__global DATA_TYPE* restrict data, uint length, __local DATA_TYPE* dataBuffer) {
// Load the data into local memory.
for (int index = get_local_id(0); index < length; index += get_local_size(0))
......@@ -49,8 +49,8 @@ __kernel void sortShortList(__global DATA_TYPE* __restrict__ data, uint length,
* Calculate the minimum and maximum value in the array to be sorted. This kernel
* is executed as a single work group.
*/
__kernel void computeRange(__global const DATA_TYPE* restrict data, uint length, __global KEY_TYPE* restrict range, __local KEY_TYPE* restrict buffer,
uint numBuckets, __global uint* restrict bucketOffset) {
__kernel void computeRange(__global const DATA_TYPE* restrict data, uint length, __global KEY_TYPE* restrict range, __local KEY_TYPE* restrict minBuffer,
__local KEY_TYPE* restrict maxBuffer, uint numBuckets, __global uint* restrict bucketOffset) {
KEY_TYPE minimum = MAX_KEY;
KEY_TYPE maximum = MIN_KEY;
......@@ -64,23 +64,18 @@ __kernel void computeRange(__global const DATA_TYPE* restrict data, uint length,
// Now reduce them.
buffer[get_local_id(0)] = minimum;
minBuffer[get_local_id(0)] = minimum;
maxBuffer[get_local_id(0)] = maximum;
barrier(CLK_LOCAL_MEM_FENCE);
for (uint step = 1; step < get_local_size(0); step *= 2) {
if (get_local_id(0)+step < get_local_size(0) && get_local_id(0)%(2*step) == 0)
buffer[get_local_id(0)] = min(buffer[get_local_id(0)], buffer[get_local_id(0)+step]);
barrier(CLK_LOCAL_MEM_FENCE);
}
minimum = buffer[0];
barrier(CLK_LOCAL_MEM_FENCE);
buffer[get_local_id(0)] = maximum;
barrier(CLK_LOCAL_MEM_FENCE);
for (uint step = 1; step < get_local_size(0); step *= 2) {
if (get_local_id(0)+step < get_local_size(0) && get_local_id(0)%(2*step) == 0)
buffer[get_local_id(0)] = max(buffer[get_local_id(0)], buffer[get_local_id(0)+step]);
if (get_local_id(0)+step < get_local_size(0) && get_local_id(0)%(2*step) == 0) {
minBuffer[get_local_id(0)] = min(minBuffer[get_local_id(0)], minBuffer[get_local_id(0)+step]);
maxBuffer[get_local_id(0)] = max(maxBuffer[get_local_id(0)], maxBuffer[get_local_id(0)+step]);
}
barrier(CLK_LOCAL_MEM_FENCE);
}
maximum = buffer[0];
minimum = minBuffer[0];
maximum = maxBuffer[0];
if (get_local_id(0) == 0) {
range[0] = minimum;
range[1] = maximum;
......@@ -96,7 +91,7 @@ __kernel void computeRange(__global const DATA_TYPE* restrict data, uint length,
* Assign elements to buckets.
*/
__kernel void assignElementsToBuckets(__global const DATA_TYPE* restrict data, uint length, uint numBuckets, __global const KEY_TYPE* restrict range,
__global uint* bucketOffset, __global uint* restrict bucketOfElement, __global uint* restrict offsetInBucket) {
__global uint* restrict bucketOffset, __global uint* restrict bucketOfElement, __global uint* restrict offsetInBucket) {
#ifdef AMD_ATOMIC_WORK_AROUND
// Do a byte write to force all memory accesses to interactionCount to use the complete path.
// This avoids the atomic access from causing all word accesses to other buffers from using the slow complete path.
......
platforms/reference/include/ReferenceCustomDynamics.h
View file @ 3b6925ae
......@@ -51,13 +51,14 @@ private:
std::vector<CustomIntegratorUtilities::Comparison> comparisons;
std::vector<bool> invalidatesForces, needsForces, needsEnergy, computeBothForceAndEnergy;
std::vector<int> forceGroupFlags, blockEnd;
RealOpenMM energy;
std::map<std::string, double> energyParamDerivs;
Lepton::CompiledExpression kineticEnergyExpression;
bool kineticEnergyNeedsForce;
CompiledExpressionSet expressionSet;
int xIndex, vIndex, mIndex, fIndex, energyIndex, gaussianIndex, uniformIndex;
std::vector<int> forceVariableIndex, energyVariableIndex, perDofVariableIndex, stepVariableIndex;
double x, v, m, f, energy, gaussian, uniform;
int xIndex, vIndex;
std::vector<int> perDofVariableIndex, stepVariableIndex;
std::vector<double> perDofVariable;
void initialize(OpenMM::ContextImpl& context, std::vector<RealOpenMM>& masses, std::map<std::string, RealOpenMM>& globals);
......@@ -65,7 +66,7 @@ private:
void computePerDof(int numberOfAtoms, std::vector<OpenMM::RealVec>& results, const std::vector<OpenMM::RealVec>& atomCoordinates,
const std::vector<OpenMM::RealVec>& velocities, const std::vector<OpenMM::RealVec>& forces, const std::vector<RealOpenMM>& masses,
const std::vector<std::vector<OpenMM::RealVec> >& perDof, const Lepton::CompiledExpression& expression, int forceIndex);
const std::vector<std::vector<OpenMM::RealVec> >& perDof, const Lepton::CompiledExpression& expression);
void recordChangedParameters(OpenMM::ContextImpl& context, std::map<std::string, RealOpenMM>& globals);
......
platforms/reference/include/ReferenceKernels.h
View file @ 3b6925ae
......@@ -604,12 +604,21 @@ public:
* @param nz the number of grid points along the Z axis
*/
void getPMEParameters(double& alpha, int& nx, int& ny, int& nz) const;
/**
* Get the dispersion parameters being used for the dispersion term in LJPME.
*
* @param alpha the separation parameter
* @param nx the number of grid points along the X axis
* @param ny the number of grid points along the Y axis
* @param nz the number of grid points along the Z axis
*/
void getLJPMEParameters(double& alpha, int& nx, int& ny, int& nz) const;
private:
int numParticles, num14;
int **bonded14IndexArray;
RealOpenMM **particleParamArray, **bonded14ParamArray;
RealOpenMM nonbondedCutoff, switchingDistance, rfDielectric, ewaldAlpha, dispersionCoefficient;
int kmax[3], gridSize[3];
RealOpenMM nonbondedCutoff, switchingDistance, rfDielectric, ewaldAlpha, ewaldDispersionAlpha, dispersionCoefficient;
int kmax[3], gridSize[3], dispersionGridSize[3];
bool useSwitchingFunction;
std::vector<std::set<int> > exclusions;
NonbondedMethod nonbondedMethod;
......
platforms/reference/include/ReferenceLJCoulombIxn.h
View file @ 3b6925ae
......@@ -38,14 +38,14 @@ class ReferenceLJCoulombIxn {
bool useSwitch;
bool periodic;
bool ewald;
bool pme;
bool pme, ljpme;
const OpenMM::NeighborList* neighborList;
OpenMM::RealVec periodicBoxVectors[3];
RealOpenMM cutoffDistance, switchingDistance;
RealOpenMM krf, crf;
RealOpenMM alphaEwald;
RealOpenMM alphaEwald, alphaDispersionEwald;
int numRx, numRy, numRz;
int meshDim[3];
int meshDim[3], dispersionMeshDim[3];
// parameter indices
......@@ -139,16 +139,28 @@ class ReferenceLJCoulombIxn {
/**---------------------------------------------------------------------------------------
Set the force to use Particle-Mesh Ewald (PME) summation.
@param alpha the Ewald separation parameter
@param gridSize the dimensions of the mesh
--------------------------------------------------------------------------------------- */
void setUsePME(RealOpenMM alpha, int meshSize[3]);
/**---------------------------------------------------------------------------------------
Set the force to use Particle-Mesh Ewald (PME) summation for dispersion.
@param dalpha the dispersion Ewald separation parameter
@param dgridSize the dimensions of the dispersion mesh
--------------------------------------------------------------------------------------- */
void setUseLJPME(RealOpenMM dalpha, int dmeshSize[3]);
/**---------------------------------------------------------------------------------------
Calculate LJ Coulomb pair ixn
......
platforms/reference/include/ReferencePME.h
View file @ 3b6925ae
......@@ -87,6 +87,28 @@ pme_exec(pme_t pme,
RealOpenMM * energy);
/*
* Evaluate reciprocal space PME dispersion energy and forces.
*
* Args:
*
* pme Opaque pme_t object, must have been initialized with pme_init()
* x Pointer to coordinate data array (nm)
* f Pointer to force data array (will be written as kJ/mol/nm)
* c6s Array of c6 coefficients (units of sqrt(kJ/mol).nm^3 )
* box Simulation cell dimensions (nm)
* energy Total energy (will be written in units of kJ/mol)
*/
int OPENMM_EXPORT
pme_exec_dpme(pme_t pme,
const std::vector<OpenMM::RealVec>& atomCoordinates,
std::vector<OpenMM::RealVec>& forces,
const std::vector<RealOpenMM>& c6s,
const OpenMM::RealVec periodicBoxVectors[3],
RealOpenMM * energy);
/* Release all memory in pme structure */
int OPENMM_EXPORT
......@@ -94,4 +116,4 @@ pme_destroy(pme_t pme);
} // namespace OpenMM
#endif // __ReferencePME_H__
\ No newline at end of file
#endif // __ReferencePME_H__
platforms/reference/include/ReferencePlatform.h
View file @ 3b6925ae
......@@ -56,7 +56,7 @@ public:
void contextDestroyed(ContextImpl& context) const;
};
class ReferencePlatform::PlatformData {
class OPENMM_EXPORT ReferencePlatform::PlatformData {
public:
PlatformData(const System& system);
~PlatformData();
......
platforms/reference/include/ReferenceStochasticDynamics.h
View file @ 3b6925ae
/* Portions copyright (c) 2006-2012 Stanford University and Simbios.
/* Portions copyright (c) 2006-2016 Stanford University and Simbios.
* Contributors: Pande Group
*
* Permission is hereby granted, free of charge, to any person obtaining
......@@ -36,7 +36,7 @@ class OPENMM_EXPORT ReferenceStochasticDynamics : public ReferenceDynamics {
std::vector<OpenMM::RealVec> xPrime;
std::vector<RealOpenMM> inverseMasses;
RealOpenMM _tau;
RealOpenMM friction;
public:
......@@ -46,12 +46,12 @@ class OPENMM_EXPORT ReferenceStochasticDynamics : public ReferenceDynamics {
@param numberOfAtoms number of atoms
@param deltaT delta t for dynamics
@param tau viscosity
@param friction friction coefficient
@param temperature temperature
--------------------------------------------------------------------------------------- */
ReferenceStochasticDynamics(int numberOfAtoms, RealOpenMM deltaT, RealOpenMM tau, RealOpenMM temperature);
ReferenceStochasticDynamics(int numberOfAtoms, RealOpenMM deltaT, RealOpenMM friction, RealOpenMM temperature);
/**---------------------------------------------------------------------------------------
......@@ -63,13 +63,11 @@ class OPENMM_EXPORT ReferenceStochasticDynamics : public ReferenceDynamics {
/**---------------------------------------------------------------------------------------
Get tau
@return tau
Get friction coefficient
--------------------------------------------------------------------------------------- */
RealOpenMM getTau() const;
RealOpenMM getFriction() const;
/**---------------------------------------------------------------------------------------
......
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