public override void ApplyPalette(Palette palette)
{
context.Synchronize();
mainStream.Synchronize();
if (palette.Width <= 0 || palette.Height <= 0)
{
throw new ArgumentException("palette may not be empty.");
}
paletteImage.Free();
paletteImage = CudaArray.Allocate(palette.Width, palette.Height, CudaArrayFormat.Float, 4);
HostBuffer2D <Vec4> hostPaletteBuffer = HostBuffer2D <Vec4> .Alloc(palette.Width, palette.Height);
Color col;
Vec4 colorVec;
for (int y = 0; y < palette.Height; y++)
{
for (int x = 0; x < palette.Width; x++)
{
col = palette.GetPixel(x, y);
colorVec = new Vec4(
(float)col.R / 255.0f,
(float)col.G / 255.0f,
(float)col.B / 255.0f,
1.0f);
hostPaletteBuffer[y, x] = colorVec;
}
}
CudaMem.Copy(hostPaletteBuffer, paletteImage);
hostPaletteBuffer.Free();
paletteTex.Array = paletteImage;
paletteTex.SetFormat(CudaArrayFormat.Float, 4);
paletteTex.AddressModeX = TexAddressMode.Clamp;
paletteTex.AddressModeY = TexAddressMode.Clamp;
paletteTex.FilterMode = TexFilterMode.Linear;
paletteTex.Flags = TexFlags.NormalizedCoordinates;
iterateKernel.SetTexRef(paletteTex);
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();
}