private void InitGPUSupport()
{
// Make device
device = new Device(DriverType.Hardware, DeviceCreationFlags.Debug, FeatureLevel.Level_11_0);
// Compile compute shader
computeShader = GPGPUHelper.LoadComputeShader(device, "GPGPU/FilterObjects.compute", "Filter", out shaderByteCode);
}
public void testgpu()
{
// Make device
Device device = new Device(DriverType.Hardware, DeviceCreationFlags.None, FeatureLevel.Level_11_0);
ShaderBytecode byteCode;
ComputeShader compute = GPGPUHelper.LoadComputeShader(device, "GPGPU/Simple.compute", "main", out byteCode);
Texture2D uavTexture;
UnorderedAccessView computeResult = GPGPUHelper.CreateUnorderedAccessView(device, 1024, 1024, SlimDX.DXGI.Format.R8G8B8A8_UNorm, out uavTexture);
device.ImmediateContext.ComputeShader.Set(compute);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(computeResult, 0);
device.ImmediateContext.Dispatch(32, 32, 1);
Texture2D.ToFile(device.ImmediateContext, uavTexture, ImageFileFormat.Png, "uav.png");
}
public byte[] FilterGPU(byte[] bgra, ushort[] depth, DepthSpacePoint[] depthSpaceData,
int nearThresh, int farThresh, int haloSize)
{
if (computeShader == null)
{
return(new byte[0]);
}
// Initialize last frame with current color frame, if it was reset
if (bLastFrameReset)
{
lastFramePixels = bgra;
bLastFrameReset = false;
}
// -- Create halo array --
List <int2> halo = new List <int2>();
int s = haloSize;
int xd = s;
int yd = s / 2;
int S = (xd + yd) / 2;
int x0 = -xd;
int x1 = +xd;
int y0 = -yd;
int y1 = +yd;
int actualHaloSize = 0;
for (int y = y0; y < y1; ++y)
{
for (int x = x0; x < x1; ++x)
{
if (Math.Abs(x) + Math.Abs(y) <= S)
{
halo.Add(new int2(x, y));
++actualHaloSize;
}
}
}
// --
// -- Perform data transformations so the arrays can be passed to the GPU --
var bgraDataTransformed = new int4[1920 * 1080];
for (int i = 0, j = 0; i < bgra.Length; i += 4, ++j)
{
bgraDataTransformed[j] = new int4(bgra[i], bgra[i + 1], bgra[i + 2], bgra[i + 3]);
}
var lastFrameDataTransformed = new int4[1920 * 1080];
for (int i = 0, j = 0; i < bgra.Length; i += 4, ++j)
{
lastFrameDataTransformed[j] = new int4(lastFramePixels[i], lastFramePixels[i + 1], lastFramePixels[i + 2], lastFramePixels[i + 3]);
}
// --
//var sw = Stopwatch.StartNew();
// Create a constant buffer to pass the filter configuration
var cbuffer = GPGPUHelper.CreateConstantBuffer(device, new int[] { nearThresh, farThresh, haloSize });
// -- Create GPULists using the immediate context and pass the data --
GPUList <int4> bgraData = new GPUList <int4>(device.ImmediateContext);
bgraData.AddRange(bgraDataTransformed);
GPUList <uint> depthData = new GPUList <uint>(device.ImmediateContext);
depthData.AddRange(depth.Select(d => (uint)d));
GPUList <DepthSpacePoint> depthSpacePointData = new GPUList <DepthSpacePoint>(device.ImmediateContext, depthSpaceData);
//depthSpacePointData.AddRange(depthSpaceData.Select(dsp => {
// if (dsp.X == float.NegativeInfinity || dsp.Y == -float.NegativeInfinity)
// {
// return new DepthSpacePoint() { X = -1, Y = -1 };
// }
// else
// {
// return dsp;
// }
//}));
GPUList <int4> lastFrameData = new GPUList <int4>(device.ImmediateContext);
lastFrameData.AddRange(lastFrameDataTransformed);
var resultArray = new int4[1920 * 1080];
GPUList <int4> resultData = new GPUList <int4>(device.ImmediateContext, resultArray);
GPUList <int2> haloData = new GPUList <int2>(device.ImmediateContext, halo);
// --
var sw = Stopwatch.StartNew();
// Set the buffers and uavs
device.ImmediateContext.ComputeShader.Set(computeShader);
device.ImmediateContext.ComputeShader.SetConstantBuffer(cbuffer, 0);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(bgraData.UnorderedAccess, 0);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(depthData.UnorderedAccess, 1);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(depthSpacePointData.UnorderedAccess, 2);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(lastFrameData.UnorderedAccess, 3);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(resultData.UnorderedAccess, 4);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(haloData.UnorderedAccess, 5);
// Run the compute shader
device.ImmediateContext.Dispatch(1920 * 1080 / 256, 1, 1);
// Get result. This call blocks, until the result was calculated
// because the MapSubresource call waits.
var result = resultData.ToArray();
sw.Stop();
// -- Clean up --
device.ImmediateContext.ComputeShader.SetConstantBuffer(null, 0);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(null, 0);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(null, 1);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(null, 2);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(null, 3);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(null, 4);
device.ImmediateContext.ComputeShader.SetUnorderedAccessView(null, 5);
cbuffer.Dispose();
bgraData.Dispose();
depthData.Dispose();
depthSpacePointData.Dispose();
lastFrameData.Dispose();
resultData.Dispose();
haloData.Dispose();
// --
Debug.WriteLine($"Filtering took {sw.ElapsedMilliseconds} ms");
var resultBytes = new byte[1920 * 1080 * 4];
for (int i = 0, j = 0; i < resultBytes.Length; i += 4, ++j)
{
resultBytes[i] = (byte)result[j].x;
resultBytes[i + 1] = (byte)result[j].y;
resultBytes[i + 2] = (byte)result[j].z;
resultBytes[i + 3] = (byte)result[j].a;
}
lastFramePixels = resultBytes;
return(resultBytes);
}
