Skip to content

GitLab

Commit 9a3877b9 authored by peastman's avatar peastman
Browse files

Merge pull request #1128 from peastman/cufft 

Reenabled cuFFT on CUDA 7.5 or later
parents 12e8a362 ec9cbc4b
Show whitespace changes
Inline Side-by-side
Showing with 3 additions and 1 deletion
platforms/cuda/src/CudaKernels.cpp
View file @ 9a3877b9
...@@ -1682,7 +1682,9 @@ void CudaCalcNonbondedForceKernel::initialize(const System& system, const Nonbon ...@@ -1682,7 +1682,9 @@ void CudaCalcNonbondedForceKernel::initialize(const System& system, const Nonbon
pmeAtomRange = CudaArray::create<int>(cu, gridSizeX*gridSizeY*gridSizeZ+1, "pmeAtomRange"); pmeAtomRange = CudaArray::create<int>(cu, gridSizeX*gridSizeY*gridSizeZ+1, "pmeAtomRange");
pmeAtomGridIndex = CudaArray::create<int2>(cu, numParticles, "pmeAtomGridIndex"); pmeAtomGridIndex = CudaArray::create<int2>(cu, numParticles, "pmeAtomGridIndex");
sort = new CudaSort(cu, new SortTrait(), cu.getNumAtoms()); sort = new CudaSort(cu, new SortTrait(), cu.getNumAtoms());
useCudaFFT = false; // We might switch back in the future, once Nvidia has all their bugs worked out int cufftVersion;
cufftGetVersion(&cufftVersion);
useCudaFFT = (cufftVersion >= 7050); // There was a critical bug in version 7.0
if (useCudaFFT) { if (useCudaFFT) {
cufftResult result = cufftPlan3d(&fftForward, gridSizeX, gridSizeY, gridSizeZ, cu.getUseDoublePrecision() ? CUFFT_D2Z : CUFFT_R2C); cufftResult result = cufftPlan3d(&fftForward, gridSizeX, gridSizeY, gridSizeZ, cu.getUseDoublePrecision() ? CUFFT_D2Z : CUFFT_R2C);
if (result != CUFFT_SUCCESS) if (result != CUFFT_SUCCESS)
......
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