public void SetUp()
{
_gpu = CudafyHost.CreateDevice(CudafyModes.Target);
Console.WriteLine("CUDA driver version={0}", _gpu.GetDriverVersion());
_fft = GPGPUFFT.Create(_gpu);
_hostInput = new float[N * BATCH];
_hostInputCplx = new ComplexF[N * BATCH];
_hostOutput = new float[N * BATCH];
_hostOutputCplx = new ComplexF[N * BATCH];
_devInput = _gpu.Allocate(_hostInput);
_devInputCplx = _gpu.Allocate(_hostInputCplx);
_devInter = _gpu.Allocate<float>(N * 2 * BATCH);
_devInterCplx = _gpu.Allocate<ComplexF>(N * BATCH);
_devOutput = _gpu.Allocate(_hostOutput);
_devOutputCplx = _gpu.Allocate(_hostOutputCplx);
Console.WriteLine("CUFFT version={0}", _fft.GetVersion());
for (int b = 0; b < BATCH; b++)
{
for (int i = 0; i < N; i++)
{
ComplexF cf = new ComplexF();
cf.x = (float)((10.0F * Math.Sin(100 * 2 * Math.PI * i / N * Math.PI / 180)));
cf.y = (float)((10.0F * Math.Sin(200 * 2 * Math.PI * i / N * Math.PI / 180)));
_hostInput[i + b * N] = cf.x;
_hostInputCplx[i + b * N] = cf;
}
}
}
public void SetUp()
{
_gpu = CudafyHost.CreateDevice(CudafyModes.Target);
_hostInput = new double[N * BATCH];
_hostInputCplx = new ComplexD[N * BATCH];
_hostOutput = new double[N * BATCH];
_hostOutputCplx = new ComplexD[N * BATCH];
_devInput = _gpu.Allocate(_hostInput);
_devInputCplx = _gpu.Allocate(_hostInputCplx);
_devInter = _gpu.Allocate <double>(N * 2 * BATCH);
_devInterCplx = _gpu.Allocate <ComplexD>(N * BATCH);
_devOutput = _gpu.Allocate(_hostOutput);
_devOutputCplx = _gpu.Allocate(_hostOutputCplx);
_fft = GPGPUFFT.Create(_gpu);
for (int b = 0; b < BATCH; b++)
{
for (int i = 0; i < N; i++)
{
ComplexD cf = new ComplexD();
cf.x = (double)((10.0F * Math.Sin(100 * 2 * Math.PI * i / N * Math.PI / 180)));
cf.y = (double)((10.0F * Math.Sin(200 * 2 * Math.PI * i / N * Math.PI / 180)));
_hostInput[i + b * N] = cf.x;
_hostInputCplx[i + b * N] = cf;
}
}
}
public void SetUp()
{
_gpu = CudafyHost.CreateDevice(CudafyModes.Target);
Console.WriteLine(_gpu.GetDriverVersion());
_fft = GPGPUFFT.Create(_gpu);
_hostInput = new float[N * BATCH];
_hostInputCplx = new ComplexF[N * BATCH];
_hostOutput = new float[N * BATCH];
_hostOutputCplx = new ComplexF[N * BATCH];
_devInput = _gpu.Allocate(_hostInput);
_devInputCplx = _gpu.Allocate(_hostInputCplx);
_devInter = _gpu.Allocate <float>(N * 2 * BATCH);
_devInterCplx = _gpu.Allocate <ComplexF>(N * BATCH);
_devOutput = _gpu.Allocate(_hostOutput);
_devOutputCplx = _gpu.Allocate(_hostOutputCplx);
Console.WriteLine(_fft.GetVersion());
for (int b = 0; b < BATCH; b++)
{
for (int i = 0; i < N; i++)
{
ComplexF cf = new ComplexF();
cf.x = (float)((10.0F * Math.Sin(100 * 2 * Math.PI * i / N * Math.PI / 180)));
cf.y = (float)((10.0F * Math.Sin(200 * 2 * Math.PI * i / N * Math.PI / 180)));
_hostInput[i + b * N] = cf.x;
_hostInputCplx[i + b * N] = cf;
}
}
}
public static IEnumerable <string> TestCUDASDK()
{
StringBuilder sb = new StringBuilder();
NvccCompilerOptions nvcc = null;
if (IntPtr.Size == 8)
{
nvcc = NvccCompilerOptions.Createx64();
}
else
{
nvcc = NvccCompilerOptions.Createx86();
}
yield return(string.Format("Platform={0}", nvcc.Platform));
yield return("Checking for CUDA SDK at " + nvcc.CompilerPath);
if (!nvcc.TryTest())
{
yield return("Could not locate CUDA Include directory.");
}
else
{
yield return(string.Format("CUDA SDK Version={0}", nvcc.Version));
yield return("Attempting to cudafy a kernel function.");
var mod = CudafyTranslator.Cudafy(nvcc.Platform, eArchitecture.sm_11, nvcc.Version, false, typeof(CUDACheck));
yield return("Successfully translated to CUDA C.");
yield return("Attempting to compile CUDA C code.");
string s = mod.Compile(eGPUCompiler.CudaNvcc, true);
yield return("Successfully compiled CUDA C into a module.");
if (CudafyHost.GetDeviceCount(eGPUType.Cuda) > 0)
{
yield return("Attempting to instantiate CUDA device object (GPGPU).");
var gpu = CudafyHost.GetDevice(eGPUType.Cuda, 0);
yield return("Successfully got CUDA device 0.");
yield return("Attempting to load module.");
gpu.LoadModule(mod);
yield return("Successfully loaded module.");
yield return("Attempting to transfer data to GPU.");
int[] a = new int[1024];
int[] b = new int[1024];
int[] c = new int[1024];
Random rand = new Random();
for (int i = 0; i < 1024; i++)
{
a[i] = rand.Next(16384);
b[i] = rand.Next(16384);
}
int[] dev_a = gpu.CopyToDevice(a);
int[] dev_b = gpu.CopyToDevice(b);
int[] dev_c = gpu.Allocate(c);
yield return("Successfully transferred data to GPU.");
yield return("Attempting to launch function on GPU.");
gpu.Launch(1, 1024).TestKernelFunction(dev_a, dev_b, dev_c);
yield return("Successfully launched function on GPU.");
yield return("Attempting to transfer results back from GPU.");
gpu.CopyFromDevice(dev_c, c);
yield return("Successfully transferred results from GPU.");
yield return("Testing results.");
int errors = 0;
for (int i = 0; i < 1024; i++)
{
if (a[i] + b[i] != c[i])
{
errors++;
}
}
if (errors == 0)
{
yield return("Successfully tested results.");
}
else
{
yield return("Test failed - results not as expected.");
}
yield return("Checking for math libraries (FFT, BLAS, SPARSE, RAND).");
var fft = GPGPUFFT.Create(gpu);
int version = fft.GetVersion();
if (version > 0)
{
yield return("Successfully detected.");
}
}
}
}
public void SetUp()
{
_gpu = CudafyHost.CreateDevice(CudafyModes.Target);
_hostInput = new double[N * BATCH];
_hostInputCplx = new ComplexD[N * BATCH];
_hostOutput = new double[N * BATCH];
_hostOutputCplx = new ComplexD[N * BATCH];
_devInput = _gpu.Allocate(_hostInput);
_devInputCplx = _gpu.Allocate(_hostInputCplx);
_devInter = _gpu.Allocate<double>(N * 2 * BATCH);
_devInterCplx = _gpu.Allocate<ComplexD>(N * BATCH);
_devOutput = _gpu.Allocate(_hostOutput);
_devOutputCplx = _gpu.Allocate(_hostOutputCplx);
_fft = GPGPUFFT.Create(_gpu);
for (int b = 0; b < BATCH; b++)
{
for (int i = 0; i < N; i++)
{
ComplexD cf = new ComplexD();
cf.x = (double)((10.0F * Math.Sin(100 * 2 * Math.PI * i / N * Math.PI / 180)));
cf.y = (double)((10.0F * Math.Sin(200 * 2 * Math.PI * i / N * Math.PI / 180)));
_hostInput[i + b * N] = cf.x;
_hostInputCplx[i + b * N] = cf;
}
}
}
