Skip to content

GitLab

Commit 530e4ea0 authored by Peter Eastman's avatar Peter Eastman
Browse files

Cleaned up a few details of the Ewald code

parent 8382f88c
Hide whitespace changes
Inline Side-by-side
Showing with 9 additions and 6 deletions
platforms/cuda/src/kernels/gpu.cpp
View file @ 530e4ea0
...@@ -432,7 +432,7 @@ void gpuSetEwaldParameters(gpuContext gpu)//, float alphaEwald, int kmax ) ...@@ -432,7 +432,7 @@ void gpuSetEwaldParameters(gpuContext gpu)//, float alphaEwald, int kmax )
gpu->sim.alphaEwald = alpha; gpu->sim.alphaEwald = alpha;
gpu->sim.factorEwald = -1 / (4*alpha*alpha); gpu->sim.factorEwald = -1 / (4*alpha*alpha);
gpu->sim.kmax = 20+1; gpu->sim.kmax = 20+1;
gpu->psEwaldEikr = new CUDAStream<float2>(gpu->sim.paddedNumberOfAtoms*gpu->sim.kmax, 1, "EwaldEikr"); gpu->psEwaldEikr = new CUDAStream<float2>(gpu->sim.atoms*gpu->sim.kmax, 1, "EwaldEikr");
gpu->sim.pEwaldEikr = gpu->psEwaldEikr->_pDevStream[0]; gpu->sim.pEwaldEikr = gpu->psEwaldEikr->_pDevStream[0];
gpu->psEwaldCosSinSum = new CUDAStream<float2>((gpu->sim.kmax*2-1) * (gpu->sim.kmax*2-1) * (gpu->sim.kmax*2-1), 1, "EwaldCosSinSum"); gpu->psEwaldCosSinSum = new CUDAStream<float2>((gpu->sim.kmax*2-1) * (gpu->sim.kmax*2-1) * (gpu->sim.kmax*2-1), 1, "EwaldCosSinSum");
gpu->sim.pEwaldCosSinSum = gpu->psEwaldCosSinSum->_pDevStream[0]; gpu->sim.pEwaldCosSinSum = gpu->psEwaldCosSinSum->_pDevStream[0];
......
platforms/cuda/src/kernels/kCalculateCDLJForces.cu
View file @ 530e4ea0
...@@ -198,18 +198,21 @@ void kCalculateCDLJForces(gpuContext gpu) ...@@ -198,18 +198,21 @@ void kCalculateCDLJForces(gpuContext gpu)
kFindInteractionsWithinBlocksEwaldDirect_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block, kFindInteractionsWithinBlocksEwaldDirect_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
sizeof(unsigned int)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit); sizeof(unsigned int)*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit);
if (gpu->bOutputBufferPerWarp) if (gpu->bOutputBufferPerWarp)
{
// Vanilla (N2) Ewald
kCalculateCDLJEwaldDirectByWarpForces_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block, kCalculateCDLJEwaldDirectByWarpForces_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
(sizeof(Atom)+sizeof(float3))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit); (sizeof(Atom)+sizeof(float3))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit);
else
kCalculateCDLJEwaldDirectForces_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block,
(sizeof(Atom)+sizeof(float3))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit);
LAUNCHERROR("kCalculateCDLJEwaldDirectForces");
if (false)
{
// O(N2) Ewald summation
kCalculateCDLJEwaldReciprocalForces_kernel<<<gpu->sim.blocks, gpu->sim.update_threads_per_block>>>(); kCalculateCDLJEwaldReciprocalForces_kernel<<<gpu->sim.blocks, gpu->sim.update_threads_per_block>>>();
LAUNCHERROR("kCalculateCDLJEwaldReciprocalForces"); LAUNCHERROR("kCalculateCDLJEwaldReciprocalForces");
} }
else else
{ {
kCalculateCDLJEwaldDirectForces_kernel<<<gpu->sim.nonbond_blocks, gpu->sim.nonbond_threads_per_block, // O(N3/2) Ewald summation
(sizeof(Atom)+sizeof(float3))*gpu->sim.nonbond_threads_per_block>>>(gpu->sim.pInteractingWorkUnit);
LAUNCHERROR("kCalculateCDLJEwaldDirectForces");
kCalculateEwaldFastEikr_kernel<<<gpu->sim.blocks, gpu->sim.update_threads_per_block>>>(); kCalculateEwaldFastEikr_kernel<<<gpu->sim.blocks, gpu->sim.update_threads_per_block>>>();
LAUNCHERROR("kCalculateEwaldFastEikr"); LAUNCHERROR("kCalculateEwaldFastEikr");
kCalculateEwaldFastCosSinSums_kernel<<<gpu->sim.blocks, gpu->sim.update_threads_per_block>>>(); kCalculateEwaldFastCosSinSums_kernel<<<gpu->sim.blocks, gpu->sim.update_threads_per_block>>>();
......
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