public static void FloatKernel(GpuArrayWrite <float> output, d3 gridDim,
d3 blockDim, d3 blockId, d3 threadId)
{
var index = blockId.d3Dex(gridDim) * blockDim.Volume() + threadId.d3Dex(blockDim);
var totalVolume = blockDim.Volume() * gridDim.Volume();
output.AddGpuData(
val: (2 * index - totalVolume) / (float)totalVolume,
index: index, block: blockId, thread: threadId);
}
public static void RunGpuKernel(d3 blocks, d3 threads, Action <d3, d3, d3, d3> kernel)
{
foreach (var blockId in blocks.Raster())
{
foreach (var threadId in threads.Raster())
{
kernel(blocks, threads, blockId, threadId);
}
}
}
public static IEnumerable <d3> Raster(this d3 bounds)
{
for (var iBx = 0; iBx < bounds.X; iBx++)
{
for (var iBy = 0; iBy < bounds.Y; iBy++)
{
for (var iBz = 0; iBz < bounds.Z; iBz++)
{
yield return(new d3(x: iBx, y: iBy, z: iBz));
}
}
}
}
public void AddGpuData(T val, int index, d3 block, d3 thread)
{
_array[index].Add(new GpuData <T>(index: index, order: _array[index].Count(), val: val, block: block, thread: thread));
}
public static int d3Dex(this d3 d3Id, d3 d3Dim)
{
return(d3Id.X + d3Id.Y * d3Dim.X + d3Id.Z * d3Dim.X * d3Dim.Y);
}
public static Action <d3, d3, d3, d3> FloatKernelWrap(GpuArrayWrite <float> gpuArrayWrite,
d3 gridDim, d3 blockDim)
{
return((d3 gd, d3 bd, d3 blockId, d3 threadId) =>
FloatKernel(gpuArrayWrite, gd, bd, blockId, threadId));
}
public static void IntKernel(GpuArrayWrite <int> output, d3 gridDim,
d3 blockDim, d3 blockId, d3 threadId)
{
output.AddGpuData(1, blockId.X * blockDim.X + threadId.X, blockId, threadId);
}
public static int Volume(this d3 what)
{
return(what.X * what.Y * what.Z);
}