internal CudaFractalEngine(Device device)
{
if (device == null)
{
throw new ArgumentException("Invalid device passed to CudaFractalEngine.", "device");
}
this.device = device;
context = device.CreateContext();
iterBlockCount = Util.Clamp(device.MultiprocessorCount * 2, 2, 64);
System.IO.MemoryStream ptxStream = new System.IO.MemoryStream(Kernels.KernelResources.kernels_ptx);
module = context.LoadModule(ptxStream);
initIteratorsKernel = module.GetKernel("init_iterators_kernel");
resetIteratorsKernel = module.GetKernel("reset_iterators_kernel");
iterateKernel = module.GetKernel("iterate_kernel");
updateStatsKernel = module.GetKernel("update_stats_kernel");
resetOutputKernel = module.GetKernel("reset_output_kernel");
updateOutputKernel = module.GetKernel("update_output_kernel");
glOutputBufferID = 0;
mainStream = new Cuda.Stream();
iterPosStateBuffer = DeviceBuffer.Alloc(8, IteratorCount);
iterColorStateBuffer = DeviceBuffer.Alloc(8, IteratorCount);
iterStatBuffer = DeviceBuffer.Alloc(Marshal.SizeOf(typeof(NativeIterStatEntry)), IteratorCount);
globalStatBuffer = DeviceBuffer.Alloc(Marshal.SizeOf(typeof(NativeGlobalStatEntry)), 1);
entropyXBuffer = DeviceBuffer.Alloc(16, IteratorCount);
entropyCBuffer = DeviceBuffer.Alloc(4, IteratorCount);
entropySeedBuffer = DeviceBuffer.Alloc(4, IteratorCount);
uint[] seeds = new uint[IteratorCount];
for (int i = 0; i < IteratorCount; i++)
{
seeds[i] = (uint)rand.Next(65536);
}
CudaMem.Copy(seeds, entropySeedBuffer);
paletteImage = CudaArray.Null;
paletteTex = module.GetTexRef("paletteTex");
resetBeginEvt = new Event();
resetEndEvt = new Event();
cycleIterEvt = new Event();
cycleStatEvt = new Event();
cycleEndEvt = new Event();
toneBeginEvt = new Event();
toneEndEvt = new Event();
initIteratorsKernel.SetBlockShape(IterBlockSize, 1, 1);
initIteratorsKernel.SetGridDim(IterBlockCount, 1);
initIteratorsKernel.SetSharedSize(0);
resetIteratorsKernel.SetBlockShape(IterBlockSize, 1, 1);
resetIteratorsKernel.SetGridDim(IterBlockCount, 1);
resetIteratorsKernel.SetSharedSize(0);
iterateKernel.SetBlockShape(IterBlockSize, 1, 1);
iterateKernel.SetGridDim(IterBlockCount, 1);
iterateKernel.SetSharedSize(0);
updateStatsKernel.SetBlockShape(1, 1, 1);
updateStatsKernel.SetGridDim(1, 1);
updateStatsKernel.SetSharedSize(0);
initIteratorsKernel.Launch(entropyXBuffer.Ptr.RawPtr, entropyCBuffer.Ptr.RawPtr, entropySeedBuffer.Ptr.RawPtr);
context.Synchronize();
}
public CudaFractalEngine()
{
device = Device.Devices[0];
context = device.CreateContext();
iterBlockCount = Util.Clamp(device.MultiprocessorCount * 2, 2, 64);
//System.Reflection.Assembly loadedAssembly = typeof(CudaFractalEngine).Assembly;
//System.IO.Stream stream = loadedAssembly.GetManifestResourceStream(typeof(CudaFractalEngine), "kernels.ptx");
System.IO.MemoryStream stream = new System.IO.MemoryStream(CudaResources.kernels_ptx);
module = context.LoadModule(stream);
initIteratorsKernel = module.GetKernel("init_iterators_kernel");
resetIteratorsKernel = module.GetKernel("reset_iterators_kernel");
iterateKernel = module.GetKernel("iterate_kernel");
updateStatsKernel = module.GetKernel("update_stats_kernel");
resetOutputKernel = module.GetKernel("reset_output_kernel");
updateOutputKernel = module.GetKernel("update_output_kernel");
glOutputBufferID = 0;
mainStream = new Cuda.Stream();
iterPosStateBuffer = DeviceBuffer2D.Alloc(8, IterBlockSize, IterBlockCount);
module.WriteConstant("iterPosStateBuffer", iterPosStateBuffer);
iterColorStateBuffer = DeviceBuffer2D.Alloc(16, IterBlockSize, IterBlockCount);
module.WriteConstant("iterColorStateBuffer", iterColorStateBuffer);
entropyXBuffer = DeviceBuffer2D.Alloc(16, IterBlockSize, IterBlockCount);
module.WriteConstant("entropyXBuffer", entropyXBuffer);
entropyCBuffer = DeviceBuffer2D.Alloc(4, IterBlockSize, IterBlockCount);
module.WriteConstant("entropyCBuffer", entropyCBuffer);
entropySeedBuffer = DeviceBuffer2D.Alloc(4, IterBlockSize, IterBlockCount);
module.WriteConstant("entropySeedBuffer", entropySeedBuffer);
HostBuffer2D <uint> hostEntropySeedBuffer = HostBuffer2D <uint> .Alloc(IterBlockSize, IterBlockCount);
uint rnd;
for (int y = 0; y < IterBlockCount; y++)
{
for (int x = 0; x < IterBlockSize; x++)
{
rnd = (uint)rand.Next(65536);
hostEntropySeedBuffer[y, x] = rnd;
}
}
CudaMem.Copy(hostEntropySeedBuffer, entropySeedBuffer);
hostEntropySeedBuffer.Free();
dotCountBuffer = DeviceBuffer2D.Alloc(8, IterBlockSize, IterBlockCount);
module.WriteConstant("dotCountBuffer", dotCountBuffer);
peakDensityBuffer = DeviceBuffer2D.Alloc(4, IterBlockSize, IterBlockCount);
module.WriteConstant("peakDensityBuffer", peakDensityBuffer);
totalIterCountMem = DevicePtr.AllocRaw(8);
module.WriteConstant("totalIterCountMem", totalIterCountMem);
totalDotCountMem = DevicePtr.AllocRaw(8);
module.WriteConstant("totalDotCountMem", totalDotCountMem);
densityMem = DevicePtr.AllocRaw(4);
module.WriteConstant("densityMem", densityMem);
peakDensityMem = DevicePtr.AllocRaw(4);
module.WriteConstant("peakDensityMem", peakDensityMem);
scaleConstantMem = DevicePtr.AllocRaw(4);
module.WriteConstant("scaleConstantMem", scaleConstantMem);
paletteImage = CudaArray.Null;
paletteTex = module.GetTexRef("paletteTex");
resetBeginEvt = new Event();
resetEndEvt = new Event();
cycleIterEvt = new Event();
cycleStatEvt = new Event();
cycleEndEvt = new Event();
toneBeginEvt = new Event();
toneEndEvt = new Event();
initIteratorsKernel.SetBlockShape(IterBlockSize, 1, 1);
initIteratorsKernel.SetGridDim(IterBlockCount, 1);
initIteratorsKernel.SetSharedSize(0);
resetIteratorsKernel.SetBlockShape(IterBlockSize, 1, 1);
resetIteratorsKernel.SetGridDim(IterBlockCount, 1);
resetIteratorsKernel.SetSharedSize(0);
iterateKernel.SetBlockShape(IterBlockSize, 1, 1);
iterateKernel.SetGridDim(IterBlockCount, 1);
iterateKernel.SetSharedSize(0);
updateStatsKernel.SetBlockShape(1, 1, 1);
updateStatsKernel.SetGridDim(1, 1);
updateStatsKernel.SetSharedSize(0);
initIteratorsKernel.Launch();
context.Synchronize();
}