Skip to content

GitLab

Commit 38a60fbf authored by Peter Eastman's avatar Peter Eastman
Browse files

Created GBSA kernels optimized for CPU

parent d28df828
Hide whitespace changes
Inline Side-by-side
Showing with 379 additions and 9 deletions
platforms/opencl/src/OpenCLKernels.cpp
View file @ 38a60fbf
......@@ -1683,6 +1683,7 @@ void OpenCLCalcGBSAOBCForceKernel::initialize(const System& system, const GBSAOB
double OpenCLCalcGBSAOBCForceKernel::execute(ContextImpl& context, bool includeForces, bool includeEnergy) {
OpenCLNonbondedUtilities& nb = cl.getNonbondedUtilities();
bool deviceIsCpu = (cl.getDevice().getInfo<CL_DEVICE_TYPE>() == CL_DEVICE_TYPE_CPU);
if (!hasCreatedKernels) {
// These Kernels cannot be created in initialize(), because the OpenCLNonbondedUtilities has not been initialized yet then.
......@@ -1700,21 +1701,27 @@ double OpenCLCalcGBSAOBCForceKernel::execute(ContextImpl& context, bool includeF
defines["NUM_ATOMS"] = intToString(cl.getNumAtoms());
defines["PADDED_NUM_ATOMS"] = intToString(cl.getPaddedNumAtoms());
defines["NUM_BLOCKS"] = OpenCLExpressionUtilities::intToString(cl.getNumAtomBlocks());
string file = (cl.getSIMDWidth() == 32 ? OpenCLKernelSources::gbsaObc_nvidia : OpenCLKernelSources::gbsaObc_default);
string file;
if (deviceIsCpu)
file = OpenCLKernelSources::gbsaObc_cpu;
else if (cl.getSIMDWidth() == 32)
file = OpenCLKernelSources::gbsaObc_nvidia;
else
file = OpenCLKernelSources::gbsaObc_default;
cl::Program program = cl.createProgram(file, defines);
int index = 0;
computeBornSumKernel = cl::Kernel(program, "computeBornSum");
computeBornSumKernel.setArg<cl::Buffer>(index++, bornSum->getDeviceBuffer());
computeBornSumKernel.setArg<cl::Buffer>(index++, cl.getPosq().getDeviceBuffer());
computeBornSumKernel.setArg<cl::Buffer>(index++, params->getDeviceBuffer());
computeBornSumKernel.setArg(index++, OpenCLContext::ThreadBlockSize*13*sizeof(cl_float), NULL);
computeBornSumKernel.setArg(index++, (deviceIsCpu ? OpenCLContext::TileSize : OpenCLContext::ThreadBlockSize)*13*sizeof(cl_float), NULL);
computeBornSumKernel.setArg(index++, OpenCLContext::ThreadBlockSize*sizeof(cl_float), NULL);
if (nb.getUseCutoff()) {
computeBornSumKernel.setArg<cl::Buffer>(index++, nb.getInteractingTiles().getDeviceBuffer());
computeBornSumKernel.setArg<cl::Buffer>(index++, nb.getInteractionCount().getDeviceBuffer());
index += 2; // The periodic box size arguments are set when the kernel is executed.
computeBornSumKernel.setArg<cl_uint>(index++, maxTiles);
if (cl.getSIMDWidth() == 32)
if (cl.getSIMDWidth() == 32 || deviceIsCpu)
computeBornSumKernel.setArg<cl::Buffer>(index++, nb.getInteractionFlags().getDeviceBuffer());
}
else
......@@ -1726,14 +1733,14 @@ double OpenCLCalcGBSAOBCForceKernel::execute(ContextImpl& context, bool includeF
force1Kernel.setArg<cl::Buffer>(index++, cl.getPosq().getDeviceBuffer());
force1Kernel.setArg<cl::Buffer>(index++, bornRadii->getDeviceBuffer());
force1Kernel.setArg<cl::Buffer>(index++, bornForce->getDeviceBuffer());
force1Kernel.setArg(index++, OpenCLContext::ThreadBlockSize*13*sizeof(cl_float), NULL);
force1Kernel.setArg(index++, (deviceIsCpu ? OpenCLContext::TileSize : OpenCLContext::ThreadBlockSize)*13*sizeof(cl_float), NULL);
force1Kernel.setArg(index++, OpenCLContext::ThreadBlockSize*sizeof(mm_float4), NULL);
if (nb.getUseCutoff()) {
force1Kernel.setArg<cl::Buffer>(index++, nb.getInteractingTiles().getDeviceBuffer());
force1Kernel.setArg<cl::Buffer>(index++, nb.getInteractionCount().getDeviceBuffer());
index += 2; // The periodic box size arguments are set when the kernel is executed.
force1Kernel.setArg<cl_uint>(index++, maxTiles);
if (cl.getSIMDWidth() == 32)
if (cl.getSIMDWidth() == 32 || deviceIsCpu)
force1Kernel.setArg<cl::Buffer>(index++, nb.getInteractionFlags().getDeviceBuffer());
}
else
......@@ -1770,9 +1777,9 @@ double OpenCLCalcGBSAOBCForceKernel::execute(ContextImpl& context, bool includeF
}
}
int numTiles = cl.getNumAtomBlocks()*(cl.getNumAtomBlocks()+1)/2;
cl.executeKernel(computeBornSumKernel, numTiles*OpenCLContext::TileSize);
cl.executeKernel(computeBornSumKernel, numTiles*OpenCLContext::TileSize, deviceIsCpu ? 1 : -1);
cl.executeKernel(reduceBornSumKernel, cl.getPaddedNumAtoms());
cl.executeKernel(force1Kernel, numTiles*OpenCLContext::TileSize);
cl.executeKernel(force1Kernel, numTiles*OpenCLContext::TileSize, deviceIsCpu ? 1 : -1);
cl.executeKernel(reduceBornForceKernel, cl.getPaddedNumAtoms());
return 0.0;
}
......
platforms/opencl/src/kernels/gbsaObc_cpu.cl 0 → 100644
View file @ 38a60fbf
#define TILE_SIZE 32
typedef struct {
float x, y, z;
float q;
float fx, fy, fz, fw;
float radius, scaledRadius;
float bornSum;
float bornRadius;
float bornForce;
} AtomData;
/**
* Compute the Born sum.
*/
__kernel __attribute__((reqd_work_group_size(WORK_GROUP_SIZE, 1, 1)))
void computeBornSum(__global float* global_bornSum, __global float4* posq, __global float2* global_params, __local AtomData* localData, __local float* tempBuffer,
#ifdef USE_CUTOFF
__global ushort2* tiles, __global unsigned int* interactionCount, float4 periodicBoxSize, float4 invPeriodicBoxSize, unsigned int maxTiles, __global unsigned int* interactionFlags) {
#else
unsigned int numTiles) {
#endif
#ifdef USE_CUTOFF
unsigned int numTiles = interactionCount[0];
unsigned int pos = get_group_id(0)*(numTiles > maxTiles ? NUM_BLOCKS*(NUM_BLOCKS+1)/2 : numTiles)/get_num_groups(0);
unsigned int end = (get_group_id(0)+1)*(numTiles > maxTiles ? NUM_BLOCKS*(NUM_BLOCKS+1)/2 : numTiles)/get_num_groups(0);
#else
unsigned int pos = get_group_id(0)*numTiles/get_num_groups(0);
unsigned int end = (get_group_id(0)+1)*numTiles/get_num_groups(0);
#endif
unsigned int lasty = 0xFFFFFFFF;
while (pos < end) {
// Extract the coordinates of this tile
unsigned int x, y;
#ifdef USE_CUTOFF
if (numTiles <= maxTiles) {
ushort2 tileIndices = tiles[pos];
x = tileIndices.x;
y = tileIndices.y;
}
else
#endif
{
y = (unsigned int) floor(NUM_BLOCKS+0.5f-sqrt((NUM_BLOCKS+0.5f)*(NUM_BLOCKS+0.5f)-2*pos));
x = (pos-y*NUM_BLOCKS+y*(y+1)/2);
if (x >= NUM_BLOCKS) { // Occasionally happens due to roundoff error.
y++;
x = (pos-y*NUM_BLOCKS+y*(y+1)/2);
}
}
// Load the data for this tile if we don't already have it cached.
if (lasty != y) {
for (int localAtomIndex = 0; localAtomIndex < TILE_SIZE; localAtomIndex++) {
unsigned int j = y*TILE_SIZE + localAtomIndex;
float4 tempPosq = posq[j];
localData[localAtomIndex].x = tempPosq.x;
localData[localAtomIndex].y = tempPosq.y;
localData[localAtomIndex].z = tempPosq.z;
localData[localAtomIndex].q = tempPosq.w;
float2 tempParams = global_params[j];
localData[localAtomIndex].radius = tempParams.x;
localData[localAtomIndex].scaledRadius = tempParams.y;
}
}
if (x == y) {
// This tile is on the diagonal.
for (unsigned int tgx = 0; tgx < TILE_SIZE; tgx++) {
unsigned int atom1 = x*TILE_SIZE+tgx;
float bornSum = 0.0f;
float4 posq1 = posq[atom1];
float2 params1 = global_params[atom1];
for (unsigned int j = 0; j < TILE_SIZE; j++) {
float4 posq2 = (float4) (localData[j].x, localData[j].y, localData[j].z, localData[j].q);
float4 delta = (float4) (posq2.xyz - posq1.xyz, 0.0f);
#ifdef USE_PERIODIC
delta.xyz -= floor(delta.xyz*invPeriodicBoxSize.xyz+0.5f)*periodicBoxSize.xyz;
#endif
float r2 = dot(delta.xyz, delta.xyz);
#ifdef USE_CUTOFF
if (atom1 < NUM_ATOMS && y*TILE_SIZE+j < NUM_ATOMS && r2 < CUTOFF_SQUARED) {
#else
if (atom1 < NUM_ATOMS && y*TILE_SIZE+j < NUM_ATOMS) {
#endif
float invR = RSQRT(r2);
float r = RECIP(invR);
float2 params2 = (float2) (localData[j].radius, localData[j].scaledRadius);
float rScaledRadiusJ = r+params2.y;
if ((j != tgx) && (params1.x < rScaledRadiusJ)) {
float l_ij = RECIP(max(params1.x, fabs(r-params2.y)));
float u_ij = RECIP(rScaledRadiusJ);
float l_ij2 = l_ij*l_ij;
float u_ij2 = u_ij*u_ij;
float ratio = LOG(u_ij * RECIP(l_ij));
bornSum += l_ij - u_ij + 0.25f*r*(u_ij2-l_ij2) + (0.50f*invR*ratio) +
(0.25f*params2.y*params2.y*invR)*(l_ij2-u_ij2);
if (params1.x < params2.x-r)
bornSum += 2.0f*(RECIP(params1.x)-l_ij);
}
}
}
// Write results.
unsigned int offset = x*TILE_SIZE + tgx + get_group_id(0)*PADDED_NUM_ATOMS;
global_bornSum[offset] += bornSum;
}
}
else {
// This is an off-diagonal tile.
for (int tgx = 0; tgx < TILE_SIZE; tgx++)
localData[tgx].bornSum = 0.0f;
// Compute the full set of interactions in this tile.
for (unsigned int tgx = 0; tgx < TILE_SIZE; tgx++) {
unsigned int atom1 = x*TILE_SIZE+tgx;
float bornSum = 0.0f;
float4 posq1 = posq[atom1];
float2 params1 = global_params[atom1];
for (unsigned int j = 0; j < TILE_SIZE; j++) {
float4 posq2 = (float4) (localData[j].x, localData[j].y, localData[j].z, localData[j].q);
float4 delta = (float4) (posq2.xyz - posq1.xyz, 0.0f);
#ifdef USE_PERIODIC
delta.xyz -= floor(delta.xyz*invPeriodicBoxSize.xyz+0.5f)*periodicBoxSize.xyz;
#endif
float r2 = dot(delta.xyz, delta.xyz);
#ifdef USE_CUTOFF
if (atom1 < NUM_ATOMS && y*TILE_SIZE+j < NUM_ATOMS && r2 < CUTOFF_SQUARED) {
#else
if (atom1 < NUM_ATOMS && y*TILE_SIZE+j < NUM_ATOMS) {
#endif
float invR = RSQRT(r2);
float r = RECIP(invR);
float2 params2 = (float2) (localData[j].radius, localData[j].scaledRadius);
float rScaledRadiusJ = r+params2.y;
if (params1.x < rScaledRadiusJ) {
float l_ij = RECIP(max(params1.x, fabs(r-params2.y)));
float u_ij = RECIP(rScaledRadiusJ);
float l_ij2 = l_ij*l_ij;
float u_ij2 = u_ij*u_ij;
float ratio = LOG(u_ij * RECIP(l_ij));
bornSum += l_ij - u_ij + 0.25f*r*(u_ij2-l_ij2) + (0.50f*invR*ratio) +
(0.25f*params2.y*params2.y*invR)*(l_ij2-u_ij2);
if (params1.x < params2.x-r)
bornSum += 2.0f*(RECIP(params1.x)-l_ij);
}
float rScaledRadiusI = r+params1.y;
if (params2.x < rScaledRadiusI) {
float l_ij = RECIP(max(params2.x, fabs(r-params1.y)));
float u_ij = RECIP(rScaledRadiusI);
float l_ij2 = l_ij*l_ij;
float u_ij2 = u_ij*u_ij;
float ratio = LOG(u_ij * RECIP(l_ij));
float term = l_ij - u_ij + 0.25f*r*(u_ij2-l_ij2) + (0.50f*invR*ratio) +
(0.25f*params1.y*params1.y*invR)*(l_ij2-u_ij2);
if (params2.x < params1.x-r)
term += 2.0f*(RECIP(params2.x)-l_ij);
localData[j].bornSum += term;
}
}
}
// Write results for atom1.
unsigned int offset = atom1 + get_group_id(0)*PADDED_NUM_ATOMS;
global_bornSum[offset] += localData[tgx].bornSum;
}
}
// Write results
for (int tgx = 0; tgx < TILE_SIZE; tgx++) {
unsigned int offset = y*TILE_SIZE+tgx + get_group_id(0)*PADDED_NUM_ATOMS;
global_bornSum[offset] += localData[tgx].bornSum;
}
lasty = y;
pos++;
}
}
/**
* First part of computing the GBSA interaction.
*/
__kernel __attribute__((reqd_work_group_size(WORK_GROUP_SIZE, 1, 1)))
void computeGBSAForce1(__global float4* forceBuffers, __global float* energyBuffer,
__global float4* posq, __global float* global_bornRadii,
__global float* global_bornForce, __local AtomData* localData, __local float4* tempBuffer,
#ifdef USE_CUTOFF
__global ushort2* tiles, __global unsigned int* interactionCount, float4 periodicBoxSize, float4 invPeriodicBoxSize, unsigned int maxTiles, __global unsigned int* interactionFlags) {
#else
unsigned int numTiles) {
#endif
#ifdef USE_CUTOFF
unsigned int numTiles = interactionCount[0];
unsigned int pos = get_group_id(0)*(numTiles > maxTiles ? NUM_BLOCKS*(NUM_BLOCKS+1)/2 : numTiles)/get_num_groups(0);
unsigned int end = (get_group_id(0)+1)*(numTiles > maxTiles ? NUM_BLOCKS*(NUM_BLOCKS+1)/2 : numTiles)/get_num_groups(0);
#else
unsigned int pos = get_group_id(0)*numTiles/get_num_groups(0);
unsigned int end = (get_group_id(0)+1)*numTiles/get_num_groups(0);
#endif
float energy = 0.0f;
unsigned int lasty = 0xFFFFFFFF;
while (pos < end) {
// Extract the coordinates of this tile
unsigned int x, y;
#ifdef USE_CUTOFF
if (numTiles <= maxTiles) {
ushort2 tileIndices = tiles[pos];
x = tileIndices.x;
y = tileIndices.y;
}
else
#endif
{
y = (unsigned int) floor(NUM_BLOCKS+0.5f-sqrt((NUM_BLOCKS+0.5f)*(NUM_BLOCKS+0.5f)-2*pos));
x = (pos-y*NUM_BLOCKS+y*(y+1)/2);
if (x >= NUM_BLOCKS) { // Occasionally happens due to roundoff error.
y++;
x = (pos-y*NUM_BLOCKS+y*(y+1)/2);
}
}
// Load the data for this tile if we don't already have it cached.
if (lasty != y) {
for (int localAtomIndex = 0; localAtomIndex < TILE_SIZE; localAtomIndex++) {
unsigned int j = y*TILE_SIZE + localAtomIndex;
float4 tempPosq = posq[j];
localData[localAtomIndex].x = tempPosq.x;
localData[localAtomIndex].y = tempPosq.y;
localData[localAtomIndex].z = tempPosq.z;
localData[localAtomIndex].q = tempPosq.w;
localData[localAtomIndex].bornRadius = global_bornRadii[j];
}
}
if (x == y) {
// This tile is on the diagonal.
for (unsigned int tgx = 0; tgx < TILE_SIZE; tgx++) {
unsigned int atom1 = x*TILE_SIZE+tgx;
float4 force = 0.0f;
float4 posq1 = posq[atom1];
float bornRadius1 = global_bornRadii[atom1];
for (unsigned int j = 0; j < TILE_SIZE; j++) {
float4 posq2 = (float4) (localData[j].x, localData[j].y, localData[j].z, localData[j].q);
float4 delta = (float4) (posq2.xyz - posq1.xyz, 0.0f);
#ifdef USE_PERIODIC
delta.xyz -= floor(delta.xyz*invPeriodicBoxSize.xyz+0.5f)*periodicBoxSize.xyz;
#endif
float r2 = dot(delta.xyz, delta.xyz);
#ifdef USE_CUTOFF
if (atom1 < NUM_ATOMS && y*TILE_SIZE+j < NUM_ATOMS && r2 < CUTOFF_SQUARED) {
#else
if (atom1 < NUM_ATOMS && y*TILE_SIZE+j < NUM_ATOMS) {
#endif
float invR = RSQRT(r2);
float r = RECIP(invR);
float bornRadius2 = localData[j].bornRadius;
float alpha2_ij = bornRadius1*bornRadius2;
float D_ij = r2*RECIP(4.0f*alpha2_ij);
float expTerm = EXP(-D_ij);
float denominator2 = r2 + alpha2_ij*expTerm;
float denominator = SQRT(denominator2);
float tempEnergy = (PREFACTOR*posq1.w*posq2.w)*RECIP(denominator);
float Gpol = tempEnergy*RECIP(denominator2);
float dGpol_dalpha2_ij = -0.5f*Gpol*expTerm*(1.0f+D_ij);
force.w += dGpol_dalpha2_ij*bornRadius2;
float dEdR = Gpol*(1.0f - 0.25f*expTerm);
energy += 0.5f*tempEnergy;
force.xyz -= delta.xyz*dEdR;
}
}
// Write results.
unsigned int offset = x*TILE_SIZE + tgx + get_group_id(0)*PADDED_NUM_ATOMS;
forceBuffers[offset].xyz = forceBuffers[offset].xyz+force.xyz;
global_bornForce[offset] += force.w;
}
}
else {
// This is an off-diagonal tile.
for (int tgx = 0; tgx < TILE_SIZE; tgx++) {
localData[tgx].fx = 0.0f;
localData[tgx].fy = 0.0f;
localData[tgx].fz = 0.0f;
localData[tgx].fw = 0.0f;
}
// Compute the full set of interactions in this tile.
for (unsigned int tgx = 0; tgx < TILE_SIZE; tgx++) {
unsigned int atom1 = x*TILE_SIZE+tgx;
float4 force = 0.0f;
float4 posq1 = posq[atom1];
float bornRadius1 = global_bornRadii[atom1];
for (unsigned int j = 0; j < TILE_SIZE; j++) {
float4 posq2 = (float4) (localData[j].x, localData[j].y, localData[j].z, localData[j].q);
float4 delta = (float4) (posq2.xyz - posq1.xyz, 0.0f);
#ifdef USE_PERIODIC
delta.xyz -= floor(delta.xyz*invPeriodicBoxSize.xyz+0.5f)*periodicBoxSize.xyz;
#endif
float r2 = dot(delta.xyz, delta.xyz);
#ifdef USE_CUTOFF
if (atom1 < NUM_ATOMS && y*TILE_SIZE+j < NUM_ATOMS && r2 < CUTOFF_SQUARED) {
#else
if (atom1 < NUM_ATOMS && y*TILE_SIZE+j < NUM_ATOMS) {
#endif
float invR = RSQRT(r2);
float r = RECIP(invR);
float bornRadius2 = localData[j].bornRadius;
float alpha2_ij = bornRadius1*bornRadius2;
float D_ij = r2*RECIP(4.0f*alpha2_ij);
float expTerm = EXP(-D_ij);
float denominator2 = r2 + alpha2_ij*expTerm;
float denominator = SQRT(denominator2);
float tempEnergy = (PREFACTOR*posq1.w*posq2.w)*RECIP(denominator);
float Gpol = tempEnergy*RECIP(denominator2);
float dGpol_dalpha2_ij = -0.5f*Gpol*expTerm*(1.0f+D_ij);
force.w += dGpol_dalpha2_ij*bornRadius2;
float dEdR = Gpol*(1.0f - 0.25f*expTerm);
energy += tempEnergy;
delta.xyz *= dEdR;
force.xyz -= delta.xyz;
localData[j].fx += delta.x;
localData[j].fy += delta.y;
localData[j].fz += delta.z;
localData[j].fw += dGpol_dalpha2_ij*bornRadius1;
}
}
// Write results for atom1.
unsigned int offset = atom1 + get_group_id(0)*PADDED_NUM_ATOMS;
forceBuffers[offset].xyz = forceBuffers[offset].xyz+force.xyz;
}
}
// Write results
for (int tgx = 0; tgx < TILE_SIZE; tgx++) {
unsigned int offset = y*TILE_SIZE+tgx + get_group_id(0)*PADDED_NUM_ATOMS;
float4 f = forceBuffers[offset];
f.x += localData[tgx].fx;
f.y += localData[tgx].fy;
f.z += localData[tgx].fz;
forceBuffers[offset] = f;
global_bornForce[offset] += localData[tgx].fw;
}
lasty = y;
pos++;
}
energyBuffer[get_global_id(0)] += energy;
}
platforms/opencl/src/kernels/gbsaObc_default.cl
View file @ 38a60fbf
......@@ -29,7 +29,6 @@ void computeBornSum(__global float* global_bornSum, __global float4* posq, __glo
unsigned int pos = get_group_id(0)*numTiles/get_num_groups(0);
unsigned int end = (get_group_id(0)+1)*numTiles/get_num_groups(0);
#endif
float energy = 0.0f;
unsigned int lasty = 0xFFFFFFFF;
while (pos < end) {
......
platforms/opencl/src/kernels/gbsaObc_nvidia.cl
View file @ 38a60fbf
......@@ -31,7 +31,6 @@ void computeBornSum(__global float* global_bornSum, __global float4* posq, __glo
unsigned int pos = warp*numTiles/totalWarps;
unsigned int end = (warp+1)*numTiles/totalWarps;
#endif
float energy = 0.0f;
unsigned int lasty = 0xFFFFFFFF;
while (pos < end) {
......
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