public CUdeviceptr Allocate(uint bytes)
{
CUdeviceptr dptr = new CUdeviceptr();
this.LastError = CUDADriver.cuMemAlloc(ref dptr, bytes);
return(dptr);
}
public override void DoLayout()
{
CUdeviceptr p1 = new CUdeviceptr();
CUDADriver.cuMemAlloc(ref p1, 1 << 10);
byte[] b = new byte[1 << 10];
CUDADriver.cuMemcpyHtoD(p1, b, (uint)b.Length);
CUfunction func = new CUfunction();
CUResult res;
int nnodes = (int)Network.VertexCount * 2;
int blocks = 32;
if (nnodes < 1024 * blocks)
{
nnodes = 1024 * blocks;
}
while ((nnodes & (prop.SIMDWidth - 1)) != 0)
{
nnodes++;
}
nnodes--;
//float dtime = 0.025f; float dthf = dtime * 0.5f;
//float epssq = 0.05f * 0.05f;
//float itolsq = 1.0f / (0.5f * 0.5f);
CUDADriver.cuModuleGetFunction(ref func, mod, "dummy");
// Float4[] data = new Float4[100];
CUdeviceptr ptr = new CUdeviceptr();
//CUDADriver.cuMemAlloc(ref ptr, (uint) 100 * System.Runtime.InteropServices.Marshal.SizeOf(Float4));
CUDADriver.cuParamSeti(func, 0, (uint)ptr.Pointer);
CUDADriver.cuParamSetSize(func, 4);
res = CUDADriver.cuLaunch(func);
if (res != CUResult.Success)
{
Logger.AddMessage(LogEntryType.Warning, "CUDA Error in dummy function: " + res.ToString());
}
// InitializationKernel<<<1, 1>>>();
CUDADriver.cuModuleGetFunction(ref func, mod, "InitializationKernel");
res = CUDADriver.cuLaunch(func);
if (res != CUResult.Success)
{
Logger.AddMessage(LogEntryType.Warning, "CUDA Error in InitializationKernel: " + res.ToString());
}
// BoundingBoxKernel<<<blocks * FACTOR1, THREADS1>>>();
CUDADriver.cuModuleGetFunction(ref func, mod, "BoundingBoxKernel: " + res.ToString());
CUDADriver.cuLaunch(func);
if (res != CUResult.Success)
{
Logger.AddMessage(LogEntryType.Warning, "CUDA Error in BoundingBoxKernel: " + res.ToString());
}
// TreeBuildingKernel<<<blocks * FACTOR2, THREADS2>>>();
CUDADriver.cuModuleGetFunction(ref func, mod, "TreeBuildingKernel: " + res.ToString());
CUDADriver.cuLaunch(func);
if (res != CUResult.Success)
{
Logger.AddMessage(LogEntryType.Warning, "CUDA Error in TreeBuildingKernel: " + res.ToString());
}
// SummarizationKernel<<<blocks * FACTOR3, THREADS3>>>();
CUDADriver.cuModuleGetFunction(ref func, mod, "SummarizationKernel: " + res.ToString());
CUDADriver.cuLaunch(func);
if (res != CUResult.Success)
{
Logger.AddMessage(LogEntryType.Warning, "CUDA Error in SummarizationKernel: " + res.ToString());
}
// ForceCalculationKernel<<<blocks * FACTOR5, THREADS5>>>();
CUDADriver.cuModuleGetFunction(ref func, mod, "ForceCalculationKernel: " + res.ToString());
CUDADriver.cuLaunch(func);
if (res != CUResult.Success)
{
Logger.AddMessage(LogEntryType.Warning, "CUDA Error in ForceCalculationKernel: " + res.ToString());
}
// IntegrationKernel<<<blocks * FACTOR6, THREADS6>>>();
CUDADriver.cuModuleGetFunction(ref func, mod, "IntegrationKernel");
CUDADriver.cuLaunch(func);
if (res != CUResult.Success)
{
Logger.AddMessage(LogEntryType.Warning, "CUDA Error in IntegrationKernel: " + res.ToString());
}
}
