Skip to content

GitLab

"platforms/cuda/tests/TestCudaCompoundIntegrator.cpp" did not exist on "1de2e2a018e190da664165dc0ac343fc16143c66"
Commit e04283fe authored by Peter Eastman's avatar Peter Eastman
Browse files

Implemented finer grained neighbor list. Also various minor optimizations.

parent 56d7ed02
Hide whitespace changes
Inline Side-by-side
Showing with 816 additions and 221 deletions
platforms/cuda/src/CudaKernels.cpp
View file @ e04283fe
......@@ -47,7 +47,7 @@ using namespace std;
static void calcForces(OpenMMContextImpl& context, CudaPlatform::PlatformData& data) {
_gpuContext* gpu = data.gpu;
if (data.stepCount%100 == 0)
if (data.nonbondedMethod != NO_CUTOFF && data.stepCount%100 == 0)
gpuReorderAtoms(gpu);
data.stepCount++;
kClearForces(gpu);
......
platforms/cuda/src/kernels/gpu.cpp
View file @ e04283fe
......@@ -3127,10 +3127,10 @@ void gpuReorderAtoms(gpuContext gpu)
// Select a bin for each molecule, then sort them by bin.
bool useHilbert = (numMolecules > 5000); // For small systems, a simple zigzag curve works better than a Hilbert curve.
bool useHilbert = (numMolecules > 5000 || atoms.size() > 8); // For small systems, a simple zigzag curve works better than a Hilbert curve.
float binWidth;
if (useHilbert)
binWidth = max(max(maxx-minx, maxy-miny), maxz-minz)/255.0;
binWidth = max(max(maxx-minx, maxy-miny), maxz-minz)/255.0;
else
binWidth = 0.2*sqrt(gpu->sim.nonbondedCutoffSqr);
int xbins = 1 + (int) ((maxx-minx)/binWidth);
......
platforms/cuda/src/kernels/kCalculateCDLJForces.cu
View file @ e04283fe
......@@ -140,12 +140,14 @@ void kCalculateCDLJForces(gpuContext gpu)
}
gpu->psInteractionCount->Download();
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
kFindInteractionsWithinBlocksCutoff_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(unsigned int)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
if (gpu->bOutputBufferPerWarp)
kCalculateCDLJCutoffByWarpForces_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float3))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
else
kCalculateCDLJCutoffForces_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float3))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
LAUNCHERROR("kCalculateCDLJCutoffForces");
break;
case PERIODIC:
......@@ -162,12 +164,14 @@ void kCalculateCDLJForces(gpuContext gpu)
}
gpu->psInteractionCount->Download();
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
kFindInteractionsWithinBlocksPeriodic_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(unsigned int)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
if (gpu->bOutputBufferPerWarp)
kCalculateCDLJPeriodicByWarpForces_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float3))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
else
kCalculateCDLJPeriodicForces_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float3))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
LAUNCHERROR("kCalculateCDLJPeriodicForces");
}
}
\ No newline at end of file
platforms/cuda/src/kernels/kCalculateCDLJForces.h
View file @ e04283fe
......@@ -42,6 +42,9 @@ __global__ void METHOD_NAME(kCalculateCDLJ, Forces_kernel)(unsigned int* workUni
unsigned int warp = (blockIdx.x*blockDim.x+threadIdx.x)/GRID;
int pos = warp*numWorkUnits/totalWarps;
int end = (warp+1)*numWorkUnits/totalWarps;
#ifdef USE_CUTOFF
float3* tempBuffer = (float3*) &sA[cSim.nonbond_threads_per_block];
#endif
int lasty = -1;
while (pos < end)
......@@ -210,47 +213,142 @@ __global__ void METHOD_NAME(kCalculateCDLJ, Forces_kernel)(unsigned int* workUni
apos.w *= cSim.epsfac;
if (!bExclusionFlag)
{
for (unsigned int j = 0; j < GRID; j++)
#ifdef USE_CUTOFF
unsigned int flags = cSim.pInteractionFlag[pos];
if (flags == 0)
{
dx = psA[tj].x - apos.x;
dy = psA[tj].y - apos.y;
dz = psA[tj].z - apos.z;
// No interactions in this block.
}
else if (flags == 0xFFFFFFFF)
#endif
{
// Compute all interactions within this block.
for (unsigned int j = 0; j < GRID; j++)
{
dx = psA[tj].x - apos.x;
dy = psA[tj].y - apos.y;
dz = psA[tj].z - apos.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
r2 = dx * dx + dy * dy + dz * dz;
invR = 1.0f / sqrt(r2);
sig = a.x + psA[tj].sig;
sig2 = invR * sig;
sig2 *= sig2;
sig6 = sig2 * sig2 * sig2;
eps = a.y * psA[tj].eps;
dEdR = eps * (12.0f * sig6 - 6.0f) * sig6;
r2 = dx * dx + dy * dy + dz * dz;
invR = 1.0f / sqrt(r2);
sig = a.x + psA[tj].sig;
sig2 = invR * sig;
sig2 *= sig2;
sig6 = sig2 * sig2 * sig2;
eps = a.y * psA[tj].eps;
dEdR = eps * (12.0f * sig6 - 6.0f) * sig6;
#ifdef USE_CUTOFF
dEdR += apos.w * psA[tj].q * (invR - 2.0f * cSim.reactionFieldK * r2);
dEdR += apos.w * psA[tj].q * (invR - 2.0f * cSim.reactionFieldK * r2);
#else
dEdR += apos.w * psA[tj].q * invR;
dEdR += apos.w * psA[tj].q * invR;
#endif
dEdR *= invR * invR;
dEdR *= invR * invR;
#ifdef USE_CUTOFF
if (r2 > cSim.nonbondedCutoffSqr)
{
dEdR = 0.0f;
if (r2 > cSim.nonbondedCutoffSqr)
{
dEdR = 0.0f;
}
#endif
dx *= dEdR;
dy *= dEdR;
dz *= dEdR;
af.x -= dx;
af.y -= dy;
af.z -= dz;
psA[tj].fx += dx;
psA[tj].fy += dy;
psA[tj].fz += dz;
tj = (tj + 1) & (GRID - 1);
}
}
#ifdef USE_CUTOFF
else
{
// Compute only a subset of the interactions in this block.
for (unsigned int j = 0; j < GRID; j++)
{
if ((flags&(1<<j)) != 0)
{
dx = psA[j].x - apos.x;
dy = psA[j].y - apos.y;
dz = psA[j].z - apos.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
dx *= dEdR;
dy *= dEdR;
dz *= dEdR;
af.x -= dx;
af.y -= dy;
af.z -= dz;
psA[tj].fx += dx;
psA[tj].fy += dy;
psA[tj].fz += dz;
tj = (tj + 1) & (GRID - 1);
r2 = dx * dx + dy * dy + dz * dz;
invR = 1.0f / sqrt(r2);
sig = a.x + psA[j].sig;
sig2 = invR * sig;
sig2 *= sig2;
sig6 = sig2 * sig2 * sig2;
eps = a.y * psA[j].eps;
dEdR = eps * (12.0f * sig6 - 6.0f) * sig6;
#ifdef USE_CUTOFF
dEdR += apos.w * psA[j].q * (invR - 2.0f * cSim.reactionFieldK * r2);
#else
dEdR += apos.w * psA[j].q * invR;
#endif
dEdR *= invR * invR;
#ifdef USE_CUTOFF
if (r2 > cSim.nonbondedCutoffSqr)
{
dEdR = 0.0f;
}
#endif
dx *= dEdR;
dy *= dEdR;
dz *= dEdR;
af.x -= dx;
af.y -= dy;
af.z -= dz;
tempBuffer[threadIdx.x].x = dx;
tempBuffer[threadIdx.x].y = dy;
tempBuffer[threadIdx.x].z = dz;
// Sum the forces on atom j.
if (tgx % 2 == 0)
{
tempBuffer[threadIdx.x].x += tempBuffer[threadIdx.x+1].x;
tempBuffer[threadIdx.x].y += tempBuffer[threadIdx.x+1].y;
tempBuffer[threadIdx.x].z += tempBuffer[threadIdx.x+1].z;
}
if (tgx % 4 == 0)
{
tempBuffer[threadIdx.x].x += tempBuffer[threadIdx.x+2].x;
tempBuffer[threadIdx.x].y += tempBuffer[threadIdx.x+2].y;
tempBuffer[threadIdx.x].z += tempBuffer[threadIdx.x+2].z;
}
if (tgx % 8 == 0)
{
tempBuffer[threadIdx.x].x += tempBuffer[threadIdx.x+4].x;
tempBuffer[threadIdx.x].y += tempBuffer[threadIdx.x+4].y;
tempBuffer[threadIdx.x].z += tempBuffer[threadIdx.x+4].z;
}
if (tgx % 16 == 0)
{
tempBuffer[threadIdx.x].x += tempBuffer[threadIdx.x+8].x;
tempBuffer[threadIdx.x].y += tempBuffer[threadIdx.x+8].y;
tempBuffer[threadIdx.x].z += tempBuffer[threadIdx.x+8].z;
}
if (tgx == 0)
{
psA[j].fx += tempBuffer[threadIdx.x].x + tempBuffer[threadIdx.x+16].x;
psA[j].fy += tempBuffer[threadIdx.x].y + tempBuffer[threadIdx.x+16].y;
psA[j].fz += tempBuffer[threadIdx.x].z + tempBuffer[threadIdx.x+16].z;
}
}
}
}
#endif
}
else // bExclusion
{
......
platforms/cuda/src/kernels/kCalculateCDLJObcGbsaForces1.cu
View file @ e04283fe
......@@ -106,6 +106,8 @@ extern __global__ void kFindBlockBoundsCutoff_kernel();
extern __global__ void kFindBlockBoundsPeriodic_kernel();
extern __global__ void kFindBlocksWithInteractionsCutoff_kernel();
extern __global__ void kFindBlocksWithInteractionsPeriodic_kernel();
extern __global__ void kFindInteractionsWithinBlocksCutoff_kernel(unsigned int*, int);
extern __global__ void kFindInteractionsWithinBlocksPeriodic_kernel(unsigned int*, int);
void kCalculateCDLJObcGbsaForces1(gpuContext gpu)
{
......@@ -145,19 +147,21 @@ void kCalculateCDLJObcGbsaForces1(gpuContext gpu)
exit(-1);
}
gpu->psInteractionCount->Download();
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
kFindInteractionsWithinBlocksCutoff_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(unsigned int)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
if (gpu->bRecalculateBornRadii)
{
kCalculateObcGbsaBornSum(gpu);
kReduceObcGbsaBornSum(gpu);
}
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
if (gpu->bOutputBufferPerWarp)
kCalculateCDLJObcGbsaCutoffByWarpForces1_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
else
kCalculateCDLJObcGbsaCutoffForces1_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
LAUNCHERROR("kCalculateCDLJCutoffForces");
(sizeof(Atom)+sizeof(float))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
LAUNCHERROR("kCalculateCDLJObcGbsaCutoffForces1");
break;
case PERIODIC:
kFindBlockBoundsPeriodic_kernel<<<(gpu->psGridBoundingBox->_length+63)/64, 64>>>();
......@@ -172,19 +176,21 @@ void kCalculateCDLJObcGbsaForces1(gpuContext gpu)
exit(-1);
}
gpu->psInteractionCount->Download();
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
kFindInteractionsWithinBlocksPeriodic_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(unsigned int)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
if (gpu->bRecalculateBornRadii)
{
kCalculateObcGbsaBornSum(gpu);
kReduceObcGbsaBornSum(gpu);
}
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
if (gpu->bOutputBufferPerWarp)
kCalculateCDLJObcGbsaPeriodicByWarpForces1_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
else
kCalculateCDLJObcGbsaPeriodicForces1_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
LAUNCHERROR("kCalculateCDLJPeriodicForces");
(sizeof(Atom)+sizeof(float))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
LAUNCHERROR("kCalculateCDLJObcGbsaPeriodicForces1");
break;
}
}
platforms/cuda/src/kernels/kCalculateCDLJObcGbsaForces1.h
View file @ e04283fe
......@@ -42,6 +42,9 @@ __global__ void METHOD_NAME(kCalculateCDLJObcGbsa, Forces1_kernel)(unsigned int*
unsigned int warp = (blockIdx.x*blockDim.x+threadIdx.x)/GRID;
int pos = warp*numWorkUnits/totalWarps;
int end = (warp+1)*numWorkUnits/totalWarps;
#ifdef USE_CUTOFF
float* tempBuffer = (float*) &sA[cSim.nonbond_threads_per_block];
#endif
int lasty = -1;
while (pos < end)
......@@ -241,63 +244,184 @@ __global__ void METHOD_NAME(kCalculateCDLJObcGbsa, Forces1_kernel)(unsigned int*
apos.w *= cSim.epsfac;
if (!bExclusionFlag)
{
for (int j = 0; j < GRID; j++)
#ifdef USE_CUTOFF
unsigned int flags = cSim.pInteractionFlag[pos];
if (flags == 0)
{
float dx = psA[tj].x - apos.x;
float dy = psA[tj].y - apos.y;
float dz = psA[tj].z - apos.z;
// No interactions in this block.
}
else if (flags == 0xFFFFFFFF)
#endif
{
// Compute all interactions within this block.
for (int j = 0; j < GRID; j++)
{
float dx = psA[tj].x - apos.x;
float dy = psA[tj].y - apos.y;
float dz = psA[tj].z - apos.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
float r2 = dx * dx + dy * dy + dz * dz;
float r2 = dx * dx + dy * dy + dz * dz;
// CDLJ part
float invR = 1.0f / sqrt(r2);
float sig = a.x + psA[tj].sig;
float sig2 = invR * sig;
sig2 *= sig2;
float sig6 = sig2 * sig2 * sig2;
float eps = a.y * psA[tj].eps;
float dEdR = eps * (12.0f * sig6 - 6.0f) * sig6;
// CDLJ part
float invR = 1.0f / sqrt(r2);
float sig = a.x + psA[tj].sig;
float sig2 = invR * sig;
sig2 *= sig2;
float sig6 = sig2 * sig2 * sig2;
float eps = a.y * psA[tj].eps;
float dEdR = eps * (12.0f * sig6 - 6.0f) * sig6;
#ifdef USE_CUTOFF
dEdR += apos.w * psA[tj].q * (invR - 2.0f * cSim.reactionFieldK * r2);
dEdR += apos.w * psA[tj].q * (invR - 2.0f * cSim.reactionFieldK * r2);
#else
dEdR += apos.w * psA[tj].q * invR;
dEdR += apos.w * psA[tj].q * invR;
#endif
dEdR *= invR * invR;
dEdR *= invR * invR;
// ObcGbsaForce1 part
float alpha2_ij = br * psA[tj].br;
float D_ij = r2 / (4.0f * alpha2_ij);
float expTerm = exp(-D_ij);
float denominator2 = r2 + alpha2_ij * expTerm;
float denominator = sqrt(denominator2);
float Gpol = (q2 * psA[tj].q) / (denominator * denominator2);
float dGpol_dalpha2_ij = -0.5f * Gpol * expTerm * (1.0f + D_ij);
af.w += dGpol_dalpha2_ij * psA[tj].br;
psA[tj].fb += dGpol_dalpha2_ij * br;
dEdR += Gpol * (1.0f - 0.25f * expTerm);
// ObcGbsaForce1 part
float alpha2_ij = br * psA[tj].br;
float D_ij = r2 / (4.0f * alpha2_ij);
float expTerm = exp(-D_ij);
float denominator2 = r2 + alpha2_ij * expTerm;
float denominator = sqrt(denominator2);
float Gpol = (q2 * psA[tj].q) / (denominator * denominator2);
float dGpol_dalpha2_ij = -0.5f * Gpol * expTerm * (1.0f + D_ij);
af.w += dGpol_dalpha2_ij * psA[tj].br;
psA[tj].fb += dGpol_dalpha2_ij * br;
dEdR += Gpol * (1.0f - 0.25f * expTerm);
#ifdef USE_CUTOFF
if (r2 > cSim.nonbondedCutoffSqr)
{
dEdR = 0.0f;
if (r2 > cSim.nonbondedCutoffSqr)
{
dEdR = 0.0f;
}
#endif
// Add forces
dx *= dEdR;
dy *= dEdR;
dz *= dEdR;
af.x -= dx;
af.y -= dy;
af.z -= dz;
psA[tj].fx += dx;
psA[tj].fy += dy;
psA[tj].fz += dz;
tj = (tj + 1) & (GRID - 1);
}
}
#ifdef USE_CUTOFF
else
{
// Compute only a subset of the interactions in this block.
for (int j = 0; j < GRID; j++)
{
if ((flags&(1<<j)) != 0)
{
float dx = psA[j].x - apos.x;
float dy = psA[j].y - apos.y;
float dz = psA[j].z - apos.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
float r2 = dx * dx + dy * dy + dz * dz;
// Add forces
dx *= dEdR;
dy *= dEdR;
dz *= dEdR;
af.x -= dx;
af.y -= dy;
af.z -= dz;
psA[tj].fx += dx;
psA[tj].fy += dy;
psA[tj].fz += dz;
tj = (tj + 1) & (GRID - 1);
// CDLJ part
float invR = 1.0f / sqrt(r2);
float sig = a.x + psA[j].sig;
float sig2 = invR * sig;
sig2 *= sig2;
float sig6 = sig2 * sig2 * sig2;
float eps = a.y * psA[j].eps;
float dEdR = eps * (12.0f * sig6 - 6.0f) * sig6;
#ifdef USE_CUTOFF
dEdR += apos.w * psA[j].q * (invR - 2.0f * cSim.reactionFieldK * r2);
#else
dEdR += apos.w * psA[j].q * invR;
#endif
dEdR *= invR * invR;
// ObcGbsaForce1 part
float alpha2_ij = br * psA[j].br;
float D_ij = r2 / (4.0f * alpha2_ij);
float expTerm = exp(-D_ij);
float denominator2 = r2 + alpha2_ij * expTerm;
float denominator = sqrt(denominator2);
float Gpol = (q2 * psA[j].q) / (denominator * denominator2);
float dGpol_dalpha2_ij = -0.5f * Gpol * expTerm * (1.0f + D_ij);
af.w += dGpol_dalpha2_ij * psA[j].br;
dEdR += Gpol * (1.0f - 0.25f * expTerm);
// Sum the Born forces.
tempBuffer[threadIdx.x] = dGpol_dalpha2_ij * br;
if (tgx % 2 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+1];
if (tgx % 4 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+2];
if (tgx % 8 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+4];
if (tgx % 16 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+8];
if (tgx == 0)
psA[j].fb += tempBuffer[threadIdx.x] + tempBuffer[threadIdx.x+16];
#ifdef USE_CUTOFF
if (r2 > cSim.nonbondedCutoffSqr)
{
dEdR = 0.0f;
}
#endif
// Add forces
dx *= dEdR;
dy *= dEdR;
dz *= dEdR;
af.x -= dx;
af.y -= dy;
af.z -= dz;
tempBuffer[threadIdx.x] = dx;
if (tgx % 2 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+1];
if (tgx % 4 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+2];
if (tgx % 8 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+4];
if (tgx % 16 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+8];
if (tgx == 0)
psA[j].fx += tempBuffer[threadIdx.x] + tempBuffer[threadIdx.x+16];
tempBuffer[threadIdx.x] = dy;
if (tgx % 2 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+1];
if (tgx % 4 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+2];
if (tgx % 8 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+4];
if (tgx % 16 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+8];
if (tgx == 0)
psA[j].fy += tempBuffer[threadIdx.x] + tempBuffer[threadIdx.x+16];
tempBuffer[threadIdx.x] = dz;
if (tgx % 2 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+1];
if (tgx % 4 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+2];
if (tgx % 8 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+4];
if (tgx % 16 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+8];
if (tgx == 0)
psA[j].fz += tempBuffer[threadIdx.x] + tempBuffer[threadIdx.x+16];
}
}
}
#endif
}
else // bExclusion
{
......
platforms/cuda/src/kernels/kCalculateObcGbsaBornSum.cu
View file @ e04283fe
......@@ -187,19 +187,19 @@ void kCalculateObcGbsaBornSum(gpuContext gpu)
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
if (gpu->bOutputBufferPerWarp)
kCalculateObcGbsaCutoffByWarpBornSum_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
else
kCalculateObcGbsaCutoffBornSum_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
break;
case PERIODIC:
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
if (gpu->bOutputBufferPerWarp)
kCalculateObcGbsaPeriodicByWarpBornSum_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
else
kCalculateObcGbsaPeriodicBornSum_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(Atom)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
break;
}
LAUNCHERROR("kCalculateBornSum");
......
platforms/cuda/src/kernels/kCalculateObcGbsaBornSum.h
View file @ e04283fe
......@@ -43,6 +43,9 @@ __global__ void METHOD_NAME(kCalculateObcGbsa, BornSum_kernel)(unsigned int* wor
#ifdef USE_OUTPUT_BUFFER_PER_WARP
unsigned int warp = (blockIdx.x*blockDim.x+threadIdx.x)/GRID;
#endif
#ifdef USE_CUTOFF
float* tempBuffer = (float*) &sA[cSim.nonbond_threads_per_block];
#endif
while (pos >= 0)
{
......@@ -107,8 +110,8 @@ __global__ void METHOD_NAME(kCalculateObcGbsa, BornSum_kernel)(unsigned int* wor
(0.50f * rInverse * ratio) +
(0.25f * psA[j].sr * psA[j].sr * rInverse) *
(l_ij2 - u_ij2);
if (ar.x < (psA[j].r - r))
float rj = psA[j].r;
if (ar.x < (rj - r))
{
apos.w += 2.0f * ((1.0f / ar.x) - l_ij);
}
......@@ -142,69 +145,169 @@ __global__ void METHOD_NAME(kCalculateObcGbsa, BornSum_kernel)(unsigned int* wor
sA[threadIdx.x].sr = temp1.y;
sA[threadIdx.x].sum = apos.w = 0.0f;
for (unsigned int j = 0; j < GRID; j++)
#ifdef USE_CUTOFF
unsigned int flags = cSim.pInteractionFlag[pos + (blockIdx.x*workUnits)/gridDim.x];
if (flags == 0)
{
// No interactions in this block.
}
else if (flags == 0xFFFFFFFF)
#endif
{
dx = psA[tj].x - apos.x;
dy = psA[tj].y - apos.y;
dz = psA[tj].z - apos.z;
// Compute all interactions within this block.
for (unsigned int j = 0; j < GRID; j++)
{
dx = psA[tj].x - apos.x;
dy = psA[tj].y - apos.y;
dz = psA[tj].z - apos.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
r2 = dx * dx + dy * dy + dz * dz;
r2 = dx * dx + dy * dy + dz * dz;
#ifdef USE_PERIODIC
if (i < cSim.atoms && y+tj < cSim.atoms && r2 < cSim.nonbondedCutoffSqr)
if (i < cSim.atoms && y+tj < cSim.atoms && r2 < cSim.nonbondedCutoffSqr)
#elif defined USE_CUTOFF
if (r2 < cSim.nonbondedCutoffSqr)
if (r2 < cSim.nonbondedCutoffSqr)
#endif
{
r = sqrt(r2);
float rInverse = 1.0f / r;
float rScaledRadiusJ = r + psA[tj].sr;
if (ar.x < rScaledRadiusJ)
{
float l_ij = 1.0f / max(ar.x, fabs(r - psA[tj].sr));
float u_ij = 1.0f / rScaledRadiusJ;
float l_ij2 = l_ij * l_ij;
float u_ij2 = u_ij * u_ij;
float ratio = log(u_ij / l_ij);
float term = l_ij -
u_ij +
0.25f * r * (u_ij2 - l_ij2) +
(0.50f * rInverse * ratio) +
(0.25f * psA[tj].sr * psA[tj].sr * rInverse) *
(l_ij2 - u_ij2);
if (ar.x < (psA[tj].sr - r))
r = sqrt(r2);
float rInverse = 1.0f / r;
float rScaledRadiusJ = r + psA[tj].sr;
if (ar.x < rScaledRadiusJ)
{
term += 2.0f * ((1.0f / ar.x) - l_ij);
float l_ij = 1.0f / max(ar.x, fabs(r - psA[tj].sr));
float u_ij = 1.0f / rScaledRadiusJ;
float l_ij2 = l_ij * l_ij;
float u_ij2 = u_ij * u_ij;
float ratio = log(u_ij / l_ij);
float term = l_ij -
u_ij +
0.25f * r * (u_ij2 - l_ij2) +
(0.50f * rInverse * ratio) +
(0.25f * psA[tj].sr * psA[tj].sr * rInverse) *
(l_ij2 - u_ij2);
float srj = psA[tj].sr;
if (ar.x < (srj - r))
{
term += 2.0f * ((1.0f / ar.x) - l_ij);
}
apos.w += term;
}
float rScaledRadiusI = r + ar.y;
if (psA[tj].r < rScaledRadiusI)
{
float l_ij = 1.0f / max(psA[tj].r, fabs(r - ar.y));
float u_ij = 1.0f / rScaledRadiusI;
float l_ij2 = l_ij * l_ij;
float u_ij2 = u_ij * u_ij;
float ratio = log(u_ij / l_ij);
float term = l_ij -
u_ij +
0.25f * r * (u_ij2 - l_ij2) +
(0.50f * rInverse * ratio) +
(0.25f * ar.y * ar.y * rInverse) *
(l_ij2 - u_ij2);
float rj = psA[tj].r;
if (rj < (ar.y - r))
{
term += 2.0f * ((1.0f / psA[tj].r) - l_ij);
}
psA[tj].sum += term;
}
apos.w += term;
}
float rScaledRadiusI = r + ar.y;
if (psA[tj].r < rScaledRadiusI)
{
float l_ij = 1.0f / max(psA[tj].r, fabs(r - ar.y));
float u_ij = 1.0f / rScaledRadiusI;
float l_ij2 = l_ij * l_ij;
float u_ij2 = u_ij * u_ij;
float ratio = log(u_ij / l_ij);
float term = l_ij -
u_ij +
0.25f * r * (u_ij2 - l_ij2) +
(0.50f * rInverse * ratio) +
(0.25f * ar.y * ar.y * rInverse) *
(l_ij2 - u_ij2);
tj = (tj - 1) & (GRID - 1);
}
}
#ifdef USE_CUTOFF
else
{
// Compute only a subset of the interactions in this block.
if (psA[tj].r < (ar.y - r))
for (unsigned int j = 0; j < GRID; j++)
{
if ((flags&(1<<j)) != 0)
{
tempBuffer[threadIdx.x] = 0.0f;
dx = psA[j].x - apos.x;
dy = psA[j].y - apos.y;
dz = psA[j].z - apos.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
r2 = dx * dx + dy * dy + dz * dz;
#ifdef USE_PERIODIC
if (i < cSim.atoms && y+j < cSim.atoms && r2 < cSim.nonbondedCutoffSqr)
#elif defined USE_CUTOFF
if (r2 < cSim.nonbondedCutoffSqr)
#endif
{
term += 2.0f * ((1.0f / psA[tj].r) - l_ij);
r = sqrt(r2);
float rInverse = 1.0f / r;
float rScaledRadiusJ = r + psA[j].sr;
if (ar.x < rScaledRadiusJ)
{
float l_ij = 1.0f / max(ar.x, fabs(r - psA[j].sr));
float u_ij = 1.0f / rScaledRadiusJ;
float l_ij2 = l_ij * l_ij;
float u_ij2 = u_ij * u_ij;
float ratio = log(u_ij / l_ij);
float term = l_ij -
u_ij +
0.25f * r * (u_ij2 - l_ij2) +
(0.50f * rInverse * ratio) +
(0.25f * psA[j].sr * psA[j].sr * rInverse) *
(l_ij2 - u_ij2);
float srj = psA[j].sr;
if (ar.x < (srj - r))
{
term += 2.0f * ((1.0f / ar.x) - l_ij);
}
apos.w += term;
}
float rScaledRadiusI = r + ar.y;
if (psA[j].r < rScaledRadiusI)
{
float l_ij = 1.0f / max(psA[j].r, fabs(r - ar.y));
float u_ij = 1.0f / rScaledRadiusI;
float l_ij2 = l_ij * l_ij;
float u_ij2 = u_ij * u_ij;
float ratio = log(u_ij / l_ij);
float term = l_ij -
u_ij +
0.25f * r * (u_ij2 - l_ij2) +
(0.50f * rInverse * ratio) +
(0.25f * ar.y * ar.y * rInverse) *
(l_ij2 - u_ij2);
float rj = psA[j].r;
if (rj < (ar.y - r))
{
term += 2.0f * ((1.0f / psA[j].r) - l_ij);
}
tempBuffer[threadIdx.x] = term;
}
}
psA[tj].sum += term;
// Sum the terms.
if (tgx % 2 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+1];
if (tgx % 4 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+2];
if (tgx % 8 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+4];
if (tgx % 16 == 0)
tempBuffer[threadIdx.x] += tempBuffer[threadIdx.x+8];
if (tgx == 0)
psA[j].sum += tempBuffer[threadIdx.x] + tempBuffer[threadIdx.x+16];
}
}
tj = (tj - 1) & (GRID - 1);
}
#endif
// Write results
#ifdef USE_OUTPUT_BUFFER_PER_WARP
......
platforms/cuda/src/kernels/kCalculateObcGbsaForces2.cu
View file @ e04283fe
......@@ -47,7 +47,6 @@ struct Atom {
float z;
float r;
float sr;
float sr2;
float fx;
float fy;
float fz;
......@@ -123,19 +122,19 @@ void kCalculateObcGbsaForces2(gpuContext gpu)
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
if (gpu->bOutputBufferPerWarp)
kCalculateObcGbsaCutoffByWarpForces2_kernel<<<gpu->sim.bornForce2_blocks, gpu->sim.bornForce2_threads_per_block,
sizeof(Atom)*gpu->sim.bornForce2_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float3))*gpu->sim.bornForce2_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
else
kCalculateObcGbsaCutoffForces2_kernel<<<gpu->sim.bornForce2_blocks, gpu->sim.bornForce2_threads_per_block,
sizeof(Atom)*gpu->sim.bornForce2_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float3))*gpu->sim.bornForce2_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
break;
case PERIODIC:
numWithInteractions = gpu->psInteractionCount->_pSysData[0];
if (gpu->bOutputBufferPerWarp)
kCalculateObcGbsaPeriodicByWarpForces2_kernel<<<gpu->sim.bornForce2_blocks, gpu->sim.bornForce2_threads_per_block,
sizeof(Atom)*gpu->sim.bornForce2_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float3))*gpu->sim.bornForce2_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
else
kCalculateObcGbsaPeriodicForces2_kernel<<<gpu->sim.bornForce2_blocks, gpu->sim.bornForce2_threads_per_block,
sizeof(Atom)*gpu->sim.bornForce2_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
(sizeof(Atom)+sizeof(float3))*gpu->sim.bornForce2_threads_per_block>>>(gpu->sim.pInteractingWorkUnit, numWithInteractions);
break;
}
LAUNCHERROR("kCalculateObcGbsaForces2");
......
platforms/cuda/src/kernels/kCalculateObcGbsaForces2.h
View file @ e04283fe
......@@ -38,10 +38,13 @@
__global__ void METHOD_NAME(kCalculateObcGbsa, Forces2_kernel)(unsigned int* workUnit, int numWorkUnits)
{
extern __shared__ Atom sA[];
unsigned int totalWarps = cSim.nonbond_blocks*cSim.nonbond_threads_per_block/GRID;
unsigned int totalWarps = cSim.bornForce2_blocks*cSim.bornForce2_threads_per_block/GRID;
unsigned int warp = (blockIdx.x*blockDim.x+threadIdx.x)/GRID;
int pos = warp*numWorkUnits/totalWarps;
int end = (warp+1)*numWorkUnits/totalWarps;
#ifdef USE_CUTOFF
float3* tempBuffer = (float3*) &sA[cSim.bornForce2_threads_per_block];
#endif
int lasty = -1;
while (pos < end)
......@@ -72,7 +75,6 @@ __global__ void METHOD_NAME(kCalculateObcGbsa, Forces2_kernel)(unsigned int* wor
sA[threadIdx.x].z = apos.z;
sA[threadIdx.x].r = a.x;
sA[threadIdx.x].sr = a.y;
sA[threadIdx.x].sr2 = a.y * a.y;
sA[threadIdx.x].fb = fb;
for (unsigned int j = (tgx+1)&(GRID-1); j != tgx; j = (j+1)&(GRID-1))
......@@ -105,7 +107,7 @@ __global__ void METHOD_NAME(kCalculateObcGbsa, Forces2_kernel)(unsigned int* wor
// Born Forces term
float term = 0.125f *
(1.000f + psA[j].sr2 * r2Inverse) * t3 +
(1.000f + psA[j].sr * psA[j].sr * r2Inverse) * t3 +
0.250f * t1 * r2Inverse;
float dE = fb * term;
......@@ -162,98 +164,242 @@ __global__ void METHOD_NAME(kCalculateObcGbsa, Forces2_kernel)(unsigned int* wor
sA[threadIdx.x].z = temp.z;
sA[threadIdx.x].r = temp1.x;
sA[threadIdx.x].sr = temp1.y;
sA[threadIdx.x].sr2 = temp1.y * temp1.y;
}
float sr2 = a.y * a.y;
for (int j = 0; j < GRID; j++)
#ifdef USE_CUTOFF
unsigned int flags = cSim.pInteractionFlag[pos];
if (flags == 0)
{
// No interactions in this block.
}
else if (flags == 0xFFFFFFFF)
#endif
{
float dx = psA[tj].x - apos.x;
float dy = psA[tj].y - apos.y;
float dz = psA[tj].z - apos.z;
// Compute all interactions within this block.
for (int j = 0; j < GRID; j++)
{
float dx = psA[tj].x - apos.x;
float dy = psA[tj].y - apos.y;
float dz = psA[tj].z - apos.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
float r2 = dx * dx + dy * dy + dz * dz;
float r = sqrt(r2);
float r2 = dx * dx + dy * dy + dz * dz;
float r = sqrt(r2);
// Interleaved Atom I and J Born Forces and sum components
float r2Inverse = 1.0f / r2;
float rScaledRadiusJ = r + psA[tj].sr;
float rScaledRadiusI = r + a.y;
float rInverse = 1.0f / r;
float l_ijJ = 1.0f / max(a.x, fabs(r - psA[tj].sr));
float l_ijI = 1.0f / max(psA[tj].r, fabs(r - a.y));
float u_ijJ = 1.0f / rScaledRadiusJ;
float u_ijI = 1.0f / rScaledRadiusI;
float l_ij2J = l_ijJ * l_ijJ;
float l_ij2I = l_ijI * l_ijI;
float u_ij2J = u_ijJ * u_ijJ;
float u_ij2I = u_ijI * u_ijI;
float t1J = log (u_ijJ / l_ijJ);
float t1I = log (u_ijI / l_ijI);
float t2J = (l_ij2J - u_ij2J);
float t2I = (l_ij2I - u_ij2I);
float t3J = t2J * rInverse;
float t3I = t2I * rInverse;
t1J *= rInverse;
t1I *= rInverse;
// Interleaved Atom I and J Born Forces and sum components
float r2Inverse = 1.0f / r2;
float rScaledRadiusJ = r + psA[tj].sr;
float rScaledRadiusI = r + a.y;
float rInverse = 1.0f / r;
float l_ijJ = 1.0f / max(a.x, fabs(r - psA[tj].sr));
float l_ijI = 1.0f / max(psA[tj].r, fabs(r - a.y));
float u_ijJ = 1.0f / rScaledRadiusJ;
float u_ijI = 1.0f / rScaledRadiusI;
float l_ij2J = l_ijJ * l_ijJ;
float l_ij2I = l_ijI * l_ijI;
float u_ij2J = u_ijJ * u_ijJ;
float u_ij2I = u_ijI * u_ijI;
float t1J = log (u_ijJ / l_ijJ);
float t1I = log (u_ijI / l_ijI);
float t2J = (l_ij2J - u_ij2J);
float t2I = (l_ij2I - u_ij2I);
float t3J = t2J * rInverse;
float t3I = t2I * rInverse;
t1J *= rInverse;
t1I *= rInverse;
// Born Forces term
float term = 0.125f *
(1.000f + psA[tj].sr2 * r2Inverse) * t3J +
0.250f * t1J * r2Inverse;
float dE = fb * term;
// Born Forces term
float term = 0.125f *
(1.000f + psA[tj].sr * psA[tj].sr * r2Inverse) * t3J +
0.250f * t1J * r2Inverse;
float dE = fb * term;
#if defined USE_PERIODIC
if (a.x >= rScaledRadiusJ || i >= cSim.atoms || y+tj >= cSim.atoms || r2 > cSim.nonbondedCutoffSqr)
if (a.x >= rScaledRadiusJ || i >= cSim.atoms || y+tj >= cSim.atoms || r2 > cSim.nonbondedCutoffSqr)
#elif defined USE_CUTOFF
if (a.x >= rScaledRadiusJ || r2 > cSim.nonbondedCutoffSqr)
if (a.x >= rScaledRadiusJ || r2 > cSim.nonbondedCutoffSqr)
#else
if (a.x >= rScaledRadiusJ)
if (a.x >= rScaledRadiusJ)
#endif
{
dE = 0.0f;
}
{
dE = 0.0f;
}
float d = dx * dE;
af.x -= d;
psA[tj].fx += d;
d = dy * dE;
af.y -= d;
psA[tj].fy += d;
d = dz * dE;
af.z -= d;
psA[tj].fz += d;
float d = dx * dE;
af.x -= d;
psA[tj].fx += d;
d = dy * dE;
af.y -= d;
psA[tj].fy += d;
d = dz * dE;
af.z -= d;
psA[tj].fz += d;
// Atom J Born sum term
term = 0.125f *
(1.000f + sr2 * r2Inverse) * t3I +
0.250f * t1I * r2Inverse;
dE = psA[tj].fb * term;
// Atom J Born sum term
term = 0.125f *
(1.000f + sr2 * r2Inverse) * t3I +
0.250f * t1I * r2Inverse;
dE = psA[tj].fb * term;
float rj = psA[tj].r;
#ifdef USE_PERIODIC
if (psA[tj].r >= rScaledRadiusI || i >= cSim.atoms || y+tj >= cSim.atoms || r2 > cSim.nonbondedCutoffSqr)
if (rj >= rScaledRadiusI || i >= cSim.atoms || y+tj >= cSim.atoms || r2 > cSim.nonbondedCutoffSqr)
#elif defined USE_CUTOFF
if (psA[tj].r >= rScaledRadiusI || r2 > cSim.nonbondedCutoffSqr)
if (rj >= rScaledRadiusI || r2 > cSim.nonbondedCutoffSqr)
#else
if (psA[tj].r >= rScaledRadiusI)
if (rj >= rScaledRadiusI)
#endif
{
dE = 0.0f;
}
dx *= dE;
dy *= dE;
dz *= dE;
psA[tj].fx += dx;
psA[tj].fy += dy;
psA[tj].fz += dz;
af.x -= dx;
af.y -= dy;
af.z -= dz;
tj = (tj + 1) & (GRID - 1);
}
}
#ifdef USE_CUTOFF
else
{
// Compute only a subset of the interactions in this block.
for (int j = 0; j < GRID; j++)
{
dE = 0.0f;
if ((flags&(1<<j)) != 0)
{
float dx = psA[j].x - apos.x;
float dy = psA[j].y - apos.y;
float dz = psA[j].z - apos.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
float r2 = dx * dx + dy * dy + dz * dz;
float r = sqrt(r2);
// Interleaved Atom I and J Born Forces and sum components
float r2Inverse = 1.0f / r2;
float rScaledRadiusJ = r + psA[j].sr;
float rScaledRadiusI = r + a.y;
float rInverse = 1.0f / r;
float l_ijJ = 1.0f / max(a.x, fabs(r - psA[j].sr));
float l_ijI = 1.0f / max(psA[j].r, fabs(r - a.y));
float u_ijJ = 1.0f / rScaledRadiusJ;
float u_ijI = 1.0f / rScaledRadiusI;
float l_ij2J = l_ijJ * l_ijJ;
float l_ij2I = l_ijI * l_ijI;
float u_ij2J = u_ijJ * u_ijJ;
float u_ij2I = u_ijI * u_ijI;
float t1J = log (u_ijJ / l_ijJ);
float t1I = log (u_ijI / l_ijI);
float t2J = (l_ij2J - u_ij2J);
float t2I = (l_ij2I - u_ij2I);
float t3J = t2J * rInverse;
float t3I = t2I * rInverse;
t1J *= rInverse;
t1I *= rInverse;
// Born Forces term
float term = 0.125f *
(1.000f + psA[j].sr * psA[j].sr * r2Inverse) * t3J +
0.250f * t1J * r2Inverse;
float dE = fb * term;
#if defined USE_PERIODIC
if (a.x >= rScaledRadiusJ || i >= cSim.atoms || y+j >= cSim.atoms || r2 > cSim.nonbondedCutoffSqr)
#elif defined USE_CUTOFF
if (a.x >= rScaledRadiusJ || r2 > cSim.nonbondedCutoffSqr)
#else
if (a.x >= rScaledRadiusJ)
#endif
{
dE = 0.0f;
}
float d = dx * dE;
af.x -= d;
tempBuffer[threadIdx.x].x = d;
d = dy * dE;
af.y -= d;
tempBuffer[threadIdx.x].y = d;
d = dz * dE;
af.z -= d;
tempBuffer[threadIdx.x].z = d;
// Atom J Born sum term
term = 0.125f *
(1.000f + sr2 * r2Inverse) * t3I +
0.250f * t1I * r2Inverse;
dE = psA[j].fb * term;
float rj = psA[j].r;
#ifdef USE_PERIODIC
if (rj >= rScaledRadiusI || i >= cSim.atoms || y+j >= cSim.atoms || r2 > cSim.nonbondedCutoffSqr)
#elif defined USE_CUTOFF
if (rj >= rScaledRadiusI || r2 > cSim.nonbondedCutoffSqr)
#else
if (rj >= rScaledRadiusI)
#endif
{
dE = 0.0f;
}
dx *= dE;
dy *= dE;
dz *= dE;
tempBuffer[threadIdx.x].x += dx;
tempBuffer[threadIdx.x].y += dy;
tempBuffer[threadIdx.x].z += dz;
af.x -= dx;
af.y -= dy;
af.z -= dz;
// Sum the forces on atom j.
if (tgx % 2 == 0)
{
tempBuffer[threadIdx.x].x += tempBuffer[threadIdx.x+1].x;
tempBuffer[threadIdx.x].y += tempBuffer[threadIdx.x+1].y;
tempBuffer[threadIdx.x].z += tempBuffer[threadIdx.x+1].z;
}
if (tgx % 4 == 0)
{
tempBuffer[threadIdx.x].x += tempBuffer[threadIdx.x+2].x;
tempBuffer[threadIdx.x].y += tempBuffer[threadIdx.x+2].y;
tempBuffer[threadIdx.x].z += tempBuffer[threadIdx.x+2].z;
}
if (tgx % 8 == 0)
{
tempBuffer[threadIdx.x].x += tempBuffer[threadIdx.x+4].x;
tempBuffer[threadIdx.x].y += tempBuffer[threadIdx.x+4].y;
tempBuffer[threadIdx.x].z += tempBuffer[threadIdx.x+4].z;
}
if (tgx % 16 == 0)
{
tempBuffer[threadIdx.x].x += tempBuffer[threadIdx.x+8].x;
tempBuffer[threadIdx.x].y += tempBuffer[threadIdx.x+8].y;
tempBuffer[threadIdx.x].z += tempBuffer[threadIdx.x+8].z;
}
if (tgx == 0)
{
psA[j].fx += tempBuffer[threadIdx.x].x + tempBuffer[threadIdx.x+16].x;
psA[j].fy += tempBuffer[threadIdx.x].y + tempBuffer[threadIdx.x+16].y;
psA[j].fz += tempBuffer[threadIdx.x].z + tempBuffer[threadIdx.x+16].z;
}
}
}
dx *= dE;
dy *= dE;
dz *= dE;
psA[tj].fx += dx;
psA[tj].fy += dy;
psA[tj].fz += dz;
af.x -= dx;
af.y -= dy;
af.z -= dz;
tj = (tj + 1) & (GRID - 1);
}
#endif
// Write results
float4 of;
......
platforms/cuda/src/kernels/kFindInteractingBlocks.h
View file @ e04283fe
......@@ -112,3 +112,88 @@ __global__ void METHOD_NAME(kFindBlocksWithInteractions, _kernel)()
cSim.pInteractionFlag[pos] = (dx*dx+dy*dy+dz*dz > cSim.nonbondedCutoffSqr ? 0 : 1);
}
}
/**
* Compare each atom in one block to the bounding box of another block, and set
* flags for which ones are interacting.
*/
__global__ void METHOD_NAME(kFindInteractionsWithinBlocks, _kernel)(unsigned int* workUnit, int numWorkUnits)
{
extern __shared__ unsigned int flags[];
unsigned int totalWarps = cSim.nonbond_blocks*cSim.nonbond_threads_per_block/GRID;
unsigned int warp = (blockIdx.x*blockDim.x+threadIdx.x)/GRID;
int pos = warp*numWorkUnits/totalWarps;
int end = (warp+1)*numWorkUnits/totalWarps;
unsigned int index = threadIdx.x & (GRID - 1);
int lasty = -1;
float4 apos;
while (pos < end)
{
// Extract cell coordinates from appropriate work unit
unsigned int x = workUnit[pos];
unsigned int y = ((x >> 2) & 0x7fff);
bool bExclusionFlag = (x & 0x1);
x = (x >> 17);
if (x == y || bExclusionFlag)
{
// Assume this block will be dense.
if (index == 0)
cSim.pInteractionFlag[pos] = 0xFFFFFFFF;
}
else
{
// Load the bounding box for x and the atom positions for y.
float4 center = cSim.pGridCenter[x];
float4 boxSize = cSim.pGridBoundingBox[x];
if (y != lasty)
{
apos = cSim.pPosq[(y<<GRIDBITS)+index];
}
// Find the distance of the atom from the bounding box.
float dx = apos.x-center.x;
float dy = apos.y-center.y;
float dz = apos.z-center.z;
#ifdef USE_PERIODIC
dx -= floor(dx/cSim.periodicBoxSizeX+0.5f)*cSim.periodicBoxSizeX;
dy -= floor(dy/cSim.periodicBoxSizeY+0.5f)*cSim.periodicBoxSizeY;
dz -= floor(dz/cSim.periodicBoxSizeZ+0.5f)*cSim.periodicBoxSizeZ;
#endif
dx = max(0.0f, abs(dx)-boxSize.x);
dy = max(0.0f, abs(dy)-boxSize.y);
dz = max(0.0f, abs(dz)-boxSize.z);
flags[threadIdx.x] = (dx*dx+dy*dy+dz*dz > cSim.nonbondedCutoffSqr ? 0 : 1 << index);
// Sum the flags.
if (index % 2 == 0)
flags[threadIdx.x] += flags[threadIdx.x+1];
if (index % 4 == 0)
flags[threadIdx.x] += flags[threadIdx.x+2];
if (index % 8 == 0)
flags[threadIdx.x] += flags[threadIdx.x+4];
if (index % 16 == 0)
flags[threadIdx.x] += flags[threadIdx.x+8];
if (index == 0)
{
unsigned int allFlags = flags[threadIdx.x] + flags[threadIdx.x+16];
// Count how many flags are set, and based on that decide whether to compute all interactions
// or only a fraction of them.
unsigned int bits = (allFlags&0x55555555) + ((allFlags>>1)&0x55555555);
bits = (bits&0x33333333) + ((bits>>2)&0x33333333);
bits = (bits&0x0F0F0F0F) + ((bits>>4)&0x0F0F0F0F);
bits = (bits&0x00FF00FF) + ((bits>>8)&0x00FF00FF);
bits = (bits&0x0000FFFF) + ((bits>>16)&0x0000FFFF);
cSim.pInteractionFlag[pos] = (bits > 12 ? 0xFFFFFFFF : allFlags);
}
lasty = y;
}
pos++;
}
}
platforms/cuda/tests/TestCudaNonbondedForce.cpp
View file @ e04283fe
......@@ -497,6 +497,36 @@ void testBlockInteractions(bool periodic) {
dy = max(0.0f, abs(dy)-gridSize1.y-gridSize2.y);
dz = max(0.0f, abs(dz)-gridSize1.z-gridSize2.z);
ASSERT(sqrt(dx*dx+dy*dy+dz*dz) < cutoff+TOL);
// Check the interaction flags.
unsigned int flags = data.gpu->psInteractionFlag->_pSysData[i];
for (int atom2 = 0; atom2 < 32; atom2++) {
if ((flags & 1) == 0) {
float4 pos2 = data.gpu->psPosq4->_pSysData[y*blockSize+atom2];
for (int atom1 = 0; atom1 < blockSize; ++atom1) {
float4 pos1 = data.gpu->psPosq4->_pSysData[x*blockSize+atom1];
float dx = pos2.x-pos1.x;
float dy = pos2.y-pos1.y;
float dz = pos2.z-pos1.z;
if (periodic) {
dx -= floor(0.5+dx/boxSize)*boxSize;
dy -= floor(0.5+dy/boxSize)*boxSize;
dz -= floor(0.5+dz/boxSize)*boxSize;
}
if (dx*dx+dy*dy+dz*dz < cutoff*cutoff) {
dx = pos2.x-center1.x;
dy = pos2.y-center1.y;
dz = pos2.z-center1.z;
dx = max(0.0f, abs(dx)-gridSize1.x);
dy = max(0.0f, abs(dy)-gridSize1.y);
dz = max(0.0f, abs(dz)-gridSize1.z);
}
ASSERT(dx*dx+dy*dy+dz*dz > cutoff*cutoff);
}
}
flags >>= 1;
}
}
// Check the tiles that did not have interactions to make sure all atoms are beyond the cutoff.
......
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