public void SetUp()
{
_gpu = CudafyHost.GetDevice();
_sparse = GPGPUSPARSE.Create(_gpu);
_blas = GPGPUBLAS.Create(_gpu);
_solver = new Solver(_gpu, _blas, _sparse);
}
public void SetUp()
{
_gpu = CudafyHost.CreateDevice(CudafyModes.Target);
_sparse = GPGPUSPARSE.Create(_gpu);
_hiMatrixMN = new double[M * N];
_hiMatrixMN2 = new double[M * N];
_hoMatrixMN = new double[M * N];
_hoPerVector = new int[M];
_hoPerVector2 = new int[N];
_diPerVector2 = _gpu.Allocate(_hoPerVector2);
_diMatrixMN = _gpu.Allocate(_hiMatrixMN);
_diMatrixMN2 = _gpu.Allocate(_hiMatrixMN2);
_diPerVector = _gpu.Allocate(_hoPerVector);
}
public Solver(GPGPU gpu, GPGPUBLAS blas, GPGPUSPARSE sparse)
{
this.gpu = gpu;
this.blas = blas;
this.sparse = sparse;
var km = CudafyModule.TryDeserialize();
if (km == null || !km.TryVerifyChecksums())
{
km = CudafyTranslator.Cudafy();
km.TrySerialize();
}
gpu.LoadModule(km);
}
public void SetUp()
{
_gpu = CudafyHost.GetDevice();
_sparse = GPGPUSPARSE.Create(_gpu);
hiMatrixMN = new double[M * N];
hiMatrixMK = new double[M * K];
hiMatrixKM = new double[K * M];
hiMatrixKN = new double[K * N];
hiMatrixNN = new double[N * N];
hiVectorXM = new double[M];
hiVectorXN = new double[N];
hiVectorYM = new double[M];
hiVectorYN = new double[N];
gpuResultM = new double[M];
gpuResultN = new double[N];
gpuResultMN = new double[M * N];
}
public void SetUp()
{
_gpu = CudafyHost.GetDevice();
_sparse = GPGPUSPARSE.Create(_gpu);
_hiVectorX = new float[N];
_hiVectorY = new float[N];
_hoVectorY = new float[N];
FillBufferSparse(_hiVectorX, out NNZ);
FillBuffer(_hiVectorY);
_hiIndicesX = new int[NNZ];
_hoValsX = new float[NNZ];
_hiValsX = new float[NNZ];
GetSparseIndex(_hiVectorX, _hiValsX, _hiIndicesX);
_diValsX = _gpu.Allocate(_hiValsX);
_diIndicesX = _gpu.Allocate(_hiIndicesX);
_diVectorY = _gpu.Allocate(_hiVectorY);
}
public SharpBLAS(GPGPU gpu)
{
Gpu = gpu;
Blas = GPGPUBLAS.Create(gpu);
Sparse = GPGPUSPARSE.Create(gpu);
}
