public string vectorSum()
{
string vecSum = @"
__kernel void vectorSum(__global float *v1, __global float *v2, __global float *v3) {
int i = get_global_id(0);
v3[i] = v1[i] + v2[i];
}
";
int size = 100000;
float[] v1_ = new float[size];
float[] v2_ = new float[size];
float[] v3_ = new float[size];
for (var i = 0; i < size; i++)
{
v1_[i] = (float)i;
v2_[i] = (float).5f;
}
var platform_ = ComputePlatform.Platforms[0];
ComputeContextPropertyList properties = new ComputeContextPropertyList(platform_);
ComputeContext ctx = new ComputeContext(ComputeDeviceTypes.Gpu, properties, null, IntPtr.Zero);
ComputeCommandQueue commands = new ComputeCommandQueue(ctx, ctx.Devices[0], ComputeCommandQueueFlags.None);
ComputeProgram program = new ComputeProgram(ctx, vecSum);
try
{
program.Build(null, null, null, IntPtr.Zero);
Console.WriteLine("program build completed");
}
catch
{
string log = program.GetBuildLog(ctx.Devices[0]);
}
ComputeBuffer <float> v1, v2, v3;
v1 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.ReadOnly | ComputeMemoryFlags.CopyHostPointer, v1_);
v2 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.ReadOnly | ComputeMemoryFlags.CopyHostPointer, v2_);
v3 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.WriteOnly | ComputeMemoryFlags.CopyHostPointer, v3_);
long[] worker = { size };
commands.WriteToBuffer(v1_, v1, false, null);
commands.WriteToBuffer(v2_, v2, false, null);
ComputeKernel sumKernal = program.CreateKernel("vectorSum");
Console.WriteLine("kernal created");
sumKernal.SetMemoryArgument(0, v1);
sumKernal.SetMemoryArgument(1, v2);
sumKernal.SetMemoryArgument(2, v3);
commands.Execute(sumKernal, null, worker, null, null);
Console.WriteLine("Executed");
commands.ReadFromBuffer <float>(v3, ref v3_, false, null);
StringBuilder sb = new StringBuilder();
for (int i = 0; i < size; i++)
{
sb.AppendFormat("{0} + {1} = {2}<br>", v1_[i].ToString(), v2_[i].ToString(), v3_[i].ToString());
}
var sum_expression_result = sb.ToString();
return(sum_expression_result);
}
public static void RunKernels(int nofKernels)
{
try
{
// Create the event wait list. An event list is not really needed for this example but it is important to see how it works.
// Note that events (like everything else) consume OpenCL resources and creating a lot of them may slow down execution.
// For this reason their use should be avoided if possible.
ComputeEventList eventList = new ComputeEventList();
commandQueue1.WriteToBuffer(input, CB_input, false, eventList);
commandQueue1.WriteToBuffer(weightIDs, CB_networkIndex, false, eventList);
// Execute the kernel "count" times. After this call returns, "eventList" will contain an event associated with this command.
// If eventList == null or typeof(eventList) == ReadOnlyCollection<ComputeEventBase>, a new event will not be created.
commandQueue1.Execute(kernel, null, new long[] { nofKernels }, null, eventList);
// Read back the results. If the command-queue has out-of-order execution enabled (default is off), ReadFromBuffer
// will not execute until any previous events in eventList (in our case only eventList[0]) are marked as complete
// by OpenCL. By default the command-queue will execute the commands in the same order as they are issued from the host.
// eventList will contain two events after this method returns.
commandQueue1.ReadFromBuffer(CB_output, ref output, false, eventList); // , eventList
// A blocking "ReadFromBuffer" (if 3rd argument is true) will wait for itself and any previous commands
// in the command queue or eventList to finish execution. Otherwise an explicit wait for all the opencl commands
// to finish has to be issued before "arrC" can be used.
// This explicit synchronization can be achieved in two ways:
// 1) Wait for the events in the list to finish,
eventList.Wait();
// 2) Or simply use
commandQueue1.Finish();
}
catch (Exception e)
{
Console.WriteLine(e.ToString());
}
}
private static void ConductSearch(ComputeContext context, ComputeKernel kernel)
{
var todos = GetQueenTaskPartition(NumQueens, 4);
var done = new List <QueenTask>();
ComputeEventList eventList = new ComputeEventList();
var commands = new ComputeCommandQueue(context, context.Devices[1], ComputeCommandQueueFlags.None);
Console.WriteLine("Starting {0} tasks, and working {1} at a time.", todos.Count, Spread);
QueenTask[] inProgress = GetNextAssignment(new QueenTask[] {}, todos, done);
var sw = new Stopwatch();
sw.Start();
while (inProgress.Any())
{
var taskBuffer =
new ComputeBuffer <QueenTask>(context,
ComputeMemoryFlags.ReadWrite | ComputeMemoryFlags.CopyHostPointer,
inProgress);
kernel.SetMemoryArgument(0, taskBuffer);
commands.WriteToBuffer(inProgress, taskBuffer, false, null);
for (int i = 0; i < 12; i++)
{
commands.Execute(kernel, null, new long[] { inProgress.Length }, null, eventList);
}
commands.ReadFromBuffer(taskBuffer, ref inProgress, false, eventList);
commands.Finish();
inProgress = GetNextAssignment(inProgress, todos, done);
}
sw.Stop();
Console.WriteLine(sw.ElapsedMilliseconds / 1000.0);
ulong sum = done.Select(state => state.solutions)
.Aggregate((total, next) => total + next);
Console.WriteLine("Q({0})={1}", NumQueens, sum);
}
private static void ConductSearch(ComputeContext context, ComputeKernel kernel)
{
var todos = GetQueenTaskPartition(NumQueens, 4);
var done = new List<QueenTask>();
ComputeEventList eventList = new ComputeEventList();
var commands = new ComputeCommandQueue(context, context.Devices[1], ComputeCommandQueueFlags.None);
Console.WriteLine("Starting {0} tasks, and working {1} at a time.", todos.Count, Spread);
QueenTask[] inProgress = GetNextAssignment(new QueenTask[] {}, todos, done);
var sw = new Stopwatch();
sw.Start();
while (inProgress.Any())
{
var taskBuffer =
new ComputeBuffer<QueenTask>(context,
ComputeMemoryFlags.ReadWrite | ComputeMemoryFlags.CopyHostPointer,
inProgress);
kernel.SetMemoryArgument(0, taskBuffer);
commands.WriteToBuffer(inProgress, taskBuffer, false, null);
for (int i = 0; i < 12; i++)
commands.Execute(kernel, null, new long[] { inProgress.Length }, null, eventList);
commands.ReadFromBuffer(taskBuffer, ref inProgress, false, eventList);
commands.Finish();
inProgress = GetNextAssignment(inProgress, todos, done);
}
sw.Stop();
Console.WriteLine(sw.ElapsedMilliseconds / 1000.0);
ulong sum = done.Select(state => state.solutions)
.Aggregate((total, next) => total + next);
Console.WriteLine("Q({0})={1}", NumQueens, sum);
}
public TerrainGen()
{
#if CPU_DEBUG
var platform = ComputePlatform.Platforms[1];
#else
var platform = ComputePlatform.Platforms[0];
#endif
_devices = new List<ComputeDevice>();
_devices.Add(platform.Devices[0]);
_properties = new ComputeContextPropertyList(platform);
_context = new ComputeContext(_devices, _properties, null, IntPtr.Zero);
_cmdQueue = new ComputeCommandQueue(_context, _devices[0], ComputeCommandQueueFlags.None);
#region setup generator kernel
bool loadFromSource = Gbl.HasRawHashChanged[Gbl.RawDir.Scripts];
loadFromSource = true;
_chunkWidthInBlocks = Gbl.LoadContent<int>("TGen_ChunkWidthInBlocks");
_chunkWidthInVerts = _chunkWidthInBlocks + 1;
_blockWidth = Gbl.LoadContent<int>("TGen_BlockWidthInMeters");
float lacunarity = Gbl.LoadContent<float>("TGen_Lacunarity");
float gain = Gbl.LoadContent<float>("TGen_Gain");
int octaves = Gbl.LoadContent<int>("TGen_Octaves");
float offset = Gbl.LoadContent<float>("TGen_Offset");
float hScale = Gbl.LoadContent<float>("TGen_HScale");
float vScale = Gbl.LoadContent<float>("TGen_VScale");
_genConstants = new ComputeBuffer<float>(_context, ComputeMemoryFlags.ReadOnly, 8);
var genArr = new[]{
lacunarity,
gain,
offset,
octaves,
hScale,
vScale,
_blockWidth,
_chunkWidthInBlocks
};
_cmdQueue.WriteToBuffer(genArr, _genConstants, false, null);
if (loadFromSource){
_generationPrgm = new ComputeProgram(_context, Gbl.LoadScript("TGen_Generator"));
#if CPU_DEBUG
_generationPrgm.Build(null, @"-g -s D:\Projects\Gondola\Scripts\GenTerrain.cl", null, IntPtr.Zero); //use option -I + scriptDir for header search
#else
_generationPrgm.Build(null, "", null, IntPtr.Zero);//use option -I + scriptDir for header search
#endif
Gbl.SaveBinary(_generationPrgm.Binaries, "TGen_Generator");
}
else{
var binary = Gbl.LoadBinary("TGen_Generator");
_generationPrgm = new ComputeProgram(_context, binary, _devices);
_generationPrgm.Build(null, "", null, IntPtr.Zero);
}
//loadFromSource = false;
_terrainGenKernel = _generationPrgm.CreateKernel("GenTerrain");
_normalGenKernel = _generationPrgm.CreateKernel("GenNormals");
//despite the script using float3 for these fields, we need to consider it to be float4 because the
//implementation is basically a float4 wrapper that uses zero for the last variable
_geometry = new ComputeBuffer<float>(_context, ComputeMemoryFlags.None, _chunkWidthInVerts*_chunkWidthInVerts*4);
_normals = new ComputeBuffer<ushort>(_context, ComputeMemoryFlags.None, _chunkWidthInVerts * _chunkWidthInVerts * 4);
_binormals = new ComputeBuffer<byte>(_context, ComputeMemoryFlags.None, _chunkWidthInVerts*_chunkWidthInVerts*4);
_tangents = new ComputeBuffer<byte>(_context, ComputeMemoryFlags.None, _chunkWidthInVerts*_chunkWidthInVerts*4);
_uvCoords = new ComputeBuffer<float>(_context, ComputeMemoryFlags.None, _chunkWidthInVerts*_chunkWidthInVerts*2);
_terrainGenKernel.SetMemoryArgument(0, _genConstants);
_terrainGenKernel.SetMemoryArgument(3, _geometry);
_terrainGenKernel.SetMemoryArgument(4, _uvCoords);
_normalGenKernel.SetMemoryArgument(0, _genConstants);
_normalGenKernel.SetMemoryArgument(3, _geometry);
_normalGenKernel.SetMemoryArgument(4, _normals);
_normalGenKernel.SetMemoryArgument(5, _binormals);
_normalGenKernel.SetMemoryArgument(6, _tangents);
#endregion
#region setup quadtree kernel
if (loadFromSource){
_qTreePrgm = new ComputeProgram(_context, Gbl.LoadScript("TGen_QTree"));
#if CPU_DEBUG
_qTreePrgm.Build(null, @"-g -s D:\Projects\Gondola\Scripts\Quadtree.cl", null, IntPtr.Zero);
#else
_qTreePrgm.Build(null, "", null, IntPtr.Zero);
#endif
Gbl.SaveBinary(_qTreePrgm.Binaries, "TGen_QTree");
}
else{
var binary = Gbl.LoadBinary("TGen_QTree");
_qTreePrgm = new ComputeProgram(_context, binary, _devices);
_qTreePrgm.Build(null, "", null, IntPtr.Zero);
}
_qTreeKernel = _qTreePrgm.CreateKernel("QuadTree");
_crossCullKernel = _qTreePrgm.CreateKernel("CrossCull");
_activeVerts = new ComputeBuffer<byte>(_context, ComputeMemoryFlags.None, _chunkWidthInVerts*_chunkWidthInVerts);
_dummy = new ComputeBuffer<int>(_context, ComputeMemoryFlags.None, 50);
var rawNormals = new ushort[_chunkWidthInVerts * _chunkWidthInVerts * 4];
_emptyVerts = new byte[_chunkWidthInVerts*_chunkWidthInVerts];
for (int i = 0; i < _emptyVerts.Length; i++){
_emptyVerts[i] = 1;
}
_cmdQueue.WriteToBuffer(rawNormals, _normals, true, null);
_cmdQueue.WriteToBuffer(_emptyVerts, _activeVerts, true, null);
_qTreeKernel.SetValueArgument(1, _chunkWidthInBlocks);
_qTreeKernel.SetMemoryArgument(2, _normals);
_qTreeKernel.SetMemoryArgument(3, _activeVerts);
_qTreeKernel.SetMemoryArgument(4, _dummy);
_crossCullKernel.SetValueArgument(1, _chunkWidthInBlocks);
_crossCullKernel.SetMemoryArgument(2, _normals);
_crossCullKernel.SetMemoryArgument(3, _activeVerts);
_crossCullKernel.SetMemoryArgument(4, _dummy);
#endregion
#region setup winding kernel
if (loadFromSource){
_winderPrgm = new ComputeProgram(_context, Gbl.LoadScript("TGen_VertexWinder"));
#if CPU_DEBUG
_winderPrgm.Build(null, @"-g -s D:\Projects\Gondola\Scripts\VertexWinder.cl", null, IntPtr.Zero);
#else
_winderPrgm.Build(null, "", null, IntPtr.Zero);
#endif
Gbl.SaveBinary(_winderPrgm.Binaries, "TGen_VertexWinder");
}
else{
var binary = Gbl.LoadBinary("TGen_VertexWinder");
_winderPrgm = new ComputeProgram(_context, binary, _devices);
_winderPrgm.Build(null, "", null, IntPtr.Zero);
}
_winderKernel = _winderPrgm.CreateKernel("VertexWinder");
_indicies = new ComputeBuffer<int>(_context, ComputeMemoryFlags.None, (_chunkWidthInBlocks)*(_chunkWidthInBlocks)*8);
_winderKernel.SetMemoryArgument(0, _activeVerts);
_winderKernel.SetMemoryArgument(1, _indicies);
_emptyIndices = new int[(_chunkWidthInBlocks)*(_chunkWidthInBlocks)*8];
for (int i = 0; i < (_chunkWidthInBlocks)*(_chunkWidthInBlocks)*8; i++){
_emptyIndices[i] = 0;
}
_cmdQueue.WriteToBuffer(_emptyIndices, _indicies, true, null);
#endregion
if (loadFromSource){
Gbl.AllowMD5Refresh[Gbl.RawDir.Scripts] = true;
}
_cmdQueue.Finish();
}
public void CopyToDevice()
{
_queue.WriteToBuffer(_cpubuffer, _gpubuffer, true, null);
}
public void Run(IComputeContext context, TextWriter log)
{
try
{
log.Write("Creating command queue... ");
var commands = new ComputeCommandQueue(context, context.Devices[0], ComputeCommandQueueFlags.None);
log.WriteLine("done.");
log.Write("Generating data... ");
int linearSize = 24;
SysIntX2 rectSize = new SysIntX2(4, 6);
SysIntX3 cubicSize = new SysIntX3(2, 3, 4);
float[] linearIn = new float[linearSize];
float[] linearOut = new float[linearSize];
float[,] rectIn = new float[(int)rectSize.Y, (int)rectSize.X];
float[,] rectOut = new float[(int)rectSize.Y, (int)rectSize.X];
float[,,] cubicIn = new float[(int)cubicSize.Z, (int)cubicSize.Y, (int)cubicSize.X];
float[,,] cubicOut = new float[(int)cubicSize.Z, (int)cubicSize.Y, (int)cubicSize.X];
for (int i = 0; i < linearSize; i++)
{
linearIn[i] = i;
}
for (int i = 0; i < (int)rectSize.X; i++)
{
for (int j = 0; j < (int)rectSize.Y; j++)
{
rectIn[j, i] = (float)(rectSize.X.ToInt32() * j + i);
}
}
for (int i = 0; i < (int)cubicSize.X; i++)
{
for (int j = 0; j < (int)cubicSize.Y; j++)
{
for (int k = 0; k < (int)cubicSize.Z; k++)
{
cubicIn[k, j, i] = (float)(k * cubicSize.Y.ToInt32() * cubicSize.X.ToInt32() + cubicSize.X.ToInt32() * j + i);
}
}
}
log.WriteLine("done.");
log.Write("Creating buffer... ");
var buffer = new ComputeBuffer <float>(context, ComputeMemoryFlags.ReadWrite, linearSize);
log.WriteLine("done.");
GC.Collect();
log.Write("Writing to buffer (linear)... ");
commands.WriteToBuffer(linearIn, buffer, false, null);
log.WriteLine("done.");
log.Write("Reading from buffer (linear)... ");
commands.ReadFromBuffer(buffer, ref linearOut, false, null);
log.WriteLine("done.");
GC.Collect();
commands.Finish();
log.Write("Comparing data... ");
Compare(linearIn, linearOut);
log.WriteLine("passed.");
GC.Collect();
log.Write("Writing to buffer (rectangular)... ");
commands.WriteToBuffer(rectIn, buffer, false, new SysIntX2(), new SysIntX2(), rectSize, null);
log.WriteLine("done.");
GC.Collect();
log.Write("Reading from buffer (rectangular)... ");
commands.ReadFromBuffer(buffer, ref rectOut, false, new SysIntX2(), new SysIntX2(), rectSize, null);
log.WriteLine("done.");
GC.Collect();
commands.Finish();
log.Write("Comparing data... ");
Compare(rectIn, rectOut);
log.WriteLine("passed.");
GC.Collect();
log.Write("Writing to buffer (cubic)... ");
commands.WriteToBuffer(cubicIn, buffer, false, new SysIntX3(), new SysIntX3(), cubicSize, null);
log.WriteLine("done.");
GC.Collect();
log.Write("Reading from buffer (cubic)... ");
commands.ReadFromBuffer(buffer, ref cubicOut, false, new SysIntX3(), new SysIntX3(), cubicSize, null);
log.WriteLine("done.");
GC.Collect();
commands.Finish();
log.Write("Comparing data... ");
Compare(cubicIn, cubicOut);
log.WriteLine("passed.");
}
catch (Exception e)
{
log.WriteLine(e.ToString());
}
}
public unsafe void EndSend()
{
for (int i = 0; i < points.Count; i++)
{
inx[i].x = (float)points[i].Item3.Real;
inx[i].y = (float)points[i].Item3.Imaginary;
inc[i].x = (float)points[i].Item4.Real;
inc[i].y = (float)points[i].Item4.Imaginary;
}
_krnl.SetMemoryArgument(0, x);
_krnl.SetMemoryArgument(1, c);
for (int i = 0; i < _ld.Count; i++)
{
_krnl.SetMemoryArgument(2 + i, outp[i]);
}
ComputeCommandQueue command = new ComputeCommandQueue(_context, _context.Devices[0], ComputeCommandQueueFlags.None);
command.WriteToBuffer(inx, x, false, null);
command.WriteToBuffer(inc, c, false, null);
command.Execute(_krnl, null, new long[] { points.Count }, null, null);
for (int i = 0; i < _ld.Count; i++)
command.ReadFromBuffer(outp[i], ref opl[i], false, null);
command.Finish();
output = new Queue<Tuple<int, int, List<ProcessLayer>>>();
for (int i = 0; i < points.Count; i++)
{
List<ProcessLayer> pl = new List<ProcessLayer>();
for (int ii = 0; ii < _ld.Count; ii++)
{
ProcessLayer p = _ld[ii].Clone();
p.c_active = opl[ii][i].c_active != 0;
p.c_calc = opl[ii][i].c_calc;
p.c_cmean = opl[ii][i].c_cmean;
p.c_cvariance = opl[ii][i].c_cvariance;
p.c_cvarsx = opl[ii][i].c_cvarsx;
p.c_isin = opl[ii][i].c_isin != 0;
p.c_n = opl[ii][i].c_n;
p.c_old2x = new Complex(opl[ii][i].c_old2x.x,opl[ii][i].c_old2x.y);
p.c_oldx = new Complex(opl[ii][i].c_oldx.x,opl[ii][i].c_oldx.y);
p.c_resn = opl[ii][i].c_resn;
p.c_resx = new Complex(opl[ii][i].c_resx.x,opl[ii][i].c_resx.y);
p.c_x = new Complex(opl[ii][i].c_x.x,opl[ii][i].c_x.y);
pl.Add(p);
}
output.Enqueue(Tuple.Create(points[i].Item1, points[i].Item2, pl));
}
}
public void CopyToDevice()
{
queue.WriteToBuffer(cpuBuffer, gpuBuffer, true, null);
}
private void CalculateConvolution(ComputeContext computeContext)
{
Stopwatch stopwatch = new Stopwatch();
stopwatch.Start();
float dx;
bool shiftXParse = float.TryParse(textBoxShiftX.Text, NumberStyles.Float, CultureInfo.InvariantCulture.NumberFormat, out dx);
if (!shiftXParse)
{
throw new SyntaxErrorException(", needs to be .");
}
float dy;
bool shiftYParse = float.TryParse(textBoxShiftX.Text, NumberStyles.Float, CultureInfo.InvariantCulture.NumberFormat, out dy);
if (!shiftYParse)
{
throw new SyntaxErrorException(", needs to be .");
}
float dz;
bool shiftZParse = float.TryParse(textBoxShiftX.Text, NumberStyles.Float, CultureInfo.InvariantCulture.NumberFormat, out dz);
if (!shiftZParse)
{
throw new SyntaxErrorException(", needs to be .");
}
int pixelCount = _imageDimensionX * _imageDimensionY * _imageDimensionZ;
Console.WriteLine("Computing...");
Console.WriteLine("Reading kernel...");
String kernelPath = Directory.GetParent(Directory.GetCurrentDirectory()).Parent.Parent.FullName;
String kernelString;
using (var sr = new StreamReader(kernelPath + "\\convolution.cl"))
kernelString = sr.ReadToEnd();
Console.WriteLine("Reading kernel... done");
float[] selectedTransformation = Transformations.GetTransformation((TransformationType)comboBoxTransform.SelectedItem, 1.0f / float.Parse(textBoxPixelSize.Text), 1.0f / float.Parse(textBoxPixelSize.Text), 1.0f / float.Parse(textBoxPixelSize.Text), dx, dy, dz);
//create openCL program
ComputeProgram computeProgram = new ComputeProgram(computeContext, kernelString);
computeProgram.Build(computeContext.Devices, null, null, IntPtr.Zero);
ComputeProgramBuildStatus computeProgramBuildStatus = computeProgram.GetBuildStatus(_selectedComputeDevice);
Console.WriteLine("computeProgramBuildStatus\n\t" + computeProgramBuildStatus);
String buildLog = computeProgram.GetBuildLog(_selectedComputeDevice);
Console.WriteLine("buildLog");
if (buildLog.Equals("\n"))
{
Console.WriteLine("\tbuildLog is empty...");
}
else
{
Console.WriteLine("\t" + buildLog);
}
float[] fluorophores = CsvData.ReadFluorophores(_sourceFilename);
/////////////////////////////////////////////
// Create a Command Queue & Event List
/////////////////////////////////////////////
ComputeCommandQueue computeCommandQueue = new ComputeCommandQueue(computeContext, _selectedComputeDevice, ComputeCommandQueueFlags.None);
////////////////////////////////////////////////////////////////
// Create Buffers Transform
////////////////////////////////////////////////////////////////
ComputeBuffer <float> fluorophoresCoords = new ComputeBuffer <float>(computeContext, ComputeMemoryFlags.ReadWrite, fluorophores.LongLength);
ComputeBuffer <float> transformationMatrix = new ComputeBuffer <float>(computeContext, ComputeMemoryFlags.ReadOnly, selectedTransformation.LongLength);
/////////////////////////////////////////////
// Create the transformFluorophoresKernel
///////////////////////////////////////////////////////////
ComputeKernel transformFluorophoresKernel = computeProgram.CreateKernel("transform_fluorophores");
/////////////////////////////////////////////
// Set the transformFluorophoresKernel arguments
/////////////////////////////////////////////
transformFluorophoresKernel.SetMemoryArgument(0, fluorophoresCoords);
transformFluorophoresKernel.SetMemoryArgument(1, transformationMatrix);
/////////////////////////////////////////////
// Configure the work-item structure
/////////////////////////////////////////////
long[] globalWorkOffsetTransformFluorophoresKernel = null;
long[] globalWorkSizeTransformFluorophoresKernel = new long[] { fluorophores.Length / 4 };
long[] localWorkSizeTransformFluorophoresKernel = null;
////////////////////////////////////////////////////////
// Enqueue the transformFluorophoresKernel for execution
////////////////////////////////////////////////////////
computeCommandQueue.WriteToBuffer(fluorophores, fluorophoresCoords, true, null);
computeCommandQueue.WriteToBuffer(selectedTransformation, transformationMatrix, true, null);
computeCommandQueue.Execute(transformFluorophoresKernel, globalWorkOffsetTransformFluorophoresKernel, globalWorkSizeTransformFluorophoresKernel, localWorkSizeTransformFluorophoresKernel, null);
// computeCommandQueue.ExecuteTask(transformFluorophoresKernel, transformFluorophoresEvents);
float[] transformedFluorophores = new float[fluorophores.Length];
computeCommandQueue.ReadFromBuffer(fluorophoresCoords, ref transformedFluorophores, true, null);
computeCommandQueue.Finish();
//TODO remove, only for testing
// for (int i = 0; i < transformedFluorophores.Length; i++)
// {
// Console.WriteLine(transformedFluorophores[i]);
// }
// /TODO remove, only for testing
stopwatch.Stop();
Console.WriteLine("Transform fluophores duration:\n\t" + stopwatch.Elapsed);
stopwatch.Reset();
stopwatch.Start();
// fluorophoresCoords are now transformed (done in place)
////////////////////////////////////////////////////////////////
// Create Buffers Convolve Fluorophores
////////////////////////////////////////////////////////////////
const int convolve_kernel_lwgs = 16;
int totalBuffer = (int)Math.Ceiling(pixelCount / (float)convolve_kernel_lwgs) * convolve_kernel_lwgs;
ComputeBuffer <float> resultImage = new ComputeBuffer <float>(computeContext, ComputeMemoryFlags.WriteOnly, totalBuffer);
/////////////////////////////////////////////
// Create the transformFluorophoresKernel
/////////////////////////////////////////////
ComputeKernel convolveFluorophoresKernel = computeProgram.CreateKernel("convolve_fluorophores");
/////////////////////////////////////////////
// Set the convolveFluorophoresKernel arguments
/////////////////////////////////////////////
convolveFluorophoresKernel.SetMemoryArgument(0, resultImage);
convolveFluorophoresKernel.SetValueArgument(1, _imageDimensionX);
convolveFluorophoresKernel.SetValueArgument(2, _imageDimensionY);
convolveFluorophoresKernel.SetMemoryArgument(3, fluorophoresCoords);
convolveFluorophoresKernel.SetLocalArgument(4, convolve_kernel_lwgs);
convolveFluorophoresKernel.SetValueArgument(5, fluorophores.Length / 4);
/////////////////////////////////////////////
// Configure the work-item structure
/////////////////////////////////////////////
long[] globalWorkOffsetTransformConvolveFluorophoresKernel = null;
long[] globalWorkSizeTransformConvolveFluorophoresKernel = new long[] { pixelCount };
long[] localWorkSizeTransformConvolveFluorophoresKernel = new long[] { convolve_kernel_lwgs };
////////////////////////////////////////////////////////
// Enqueue the convolveFluorophoresKernel for execution
////////////////////////////////////////////////////////
computeCommandQueue.Execute(convolveFluorophoresKernel, globalWorkOffsetTransformConvolveFluorophoresKernel, globalWorkSizeTransformConvolveFluorophoresKernel, localWorkSizeTransformConvolveFluorophoresKernel, null);
float[] resultImageData = new float[totalBuffer];
computeCommandQueue.ReadFromBuffer(resultImage, ref resultImageData, true, null);
computeCommandQueue.Finish();
for (int i = 0; i < pixelCount; i++)
{
Console.WriteLine(resultImageData[i]);
}
Console.WriteLine("Writing data to file...");
// CsvData.WriteToDisk("..\\..\\..\\output.csv", resultImageData);
TiffData.WriteToDisk(resultImageData, _saveFilename, _imageDimensionX, _imageDimensionY);
Bitmap bitmap = new Bitmap(_imageDimensionX, _imageDimensionY);
float max = resultImageData.Max();
float scale = 255 / (float)max;
// for (int r = 0; r < _imageDimensionY; r++)
// {
// for (int c = 0; c < _imageDimensionX; c++)
// {
// float value = resultImageData[c*(r + 1)];
// Color newColor = Color.FromArgb((int)(value * scale), (int)(value * scale), (int)(value * scale));
// bitmap.SetPixel(c,r, newColor);
// }
// }
ushort[] ushortdata = new ushort[resultImageData.Length];
for (int i = 0; i < resultImageData.Length; i++)
{
ushortdata[i] = (ushort)resultImageData[i];
}
uint[] convertGray16ToRgb = ConvertGray16ToRGB(ushortdata, 16);
byte[] bytes = new byte[convertGray16ToRgb.Length * 4];
//
// int[] resultImageData2 = new int[resultImageData.Length];
//
for (int index = 0; index < convertGray16ToRgb.Length; index++)
{
// resultImageData2[index] = (int)(scale*resultImageData[index]);
byte[] bytes1 = BitConverter.GetBytes(convertGray16ToRgb[index]);
bytes[index] = bytes1[0];
bytes[4 * index + 1] = bytes1[1];
bytes[4 * index + 2] = bytes1[2];
bytes[4 * index + 3] = bytes1[3];
}
//
// for (int r = 0; r < _imageDimensionY; r++)
// {
// for (int c = 0; c < _imageDimensionX; c++)
// {
// float value = resultImageData2[c*(r + 1)];
// Color newColor = Color.FromArgb((int)(value), (int)(value), (int)(value));
// bitmap.SetPixel(c,r, newColor);
// }
// }
// bitmap.Save("c:\\temp.bmp");
using (MemoryStream ms = new MemoryStream(bytes))
{
Image image = Bitmap.FromStream(ms);
image.Save("c:\\temp.bmp");
}
Console.WriteLine("Writing data to file... done");
stopwatch.Stop();
Console.WriteLine("Convolve fluophores duration:\n\t" + stopwatch.Elapsed);
Console.WriteLine("Computing... done");
}
private void CalculateConvolution(ComputeContext computeContext)
{
Stopwatch stopwatch = new Stopwatch();
stopwatch.Start();
float dx;
bool shiftXParse = float.TryParse(textBoxShiftX.Text, NumberStyles.Float, CultureInfo.InvariantCulture.NumberFormat, out dx);
if (!shiftXParse)
throw new SyntaxErrorException(", needs to be .");
float dy;
bool shiftYParse = float.TryParse(textBoxShiftX.Text, NumberStyles.Float, CultureInfo.InvariantCulture.NumberFormat, out dy);
if (!shiftYParse)
throw new SyntaxErrorException(", needs to be .");
float dz;
bool shiftZParse = float.TryParse(textBoxShiftX.Text, NumberStyles.Float, CultureInfo.InvariantCulture.NumberFormat, out dz);
if (!shiftZParse)
throw new SyntaxErrorException(", needs to be .");
int pixelCount = _imageDimensionX*_imageDimensionY*_imageDimensionZ;
Console.WriteLine("Computing...");
Console.WriteLine("Reading kernel...");
String kernelPath = Directory.GetParent(Directory.GetCurrentDirectory()).Parent.Parent.FullName;
String kernelString;
using (var sr = new StreamReader(kernelPath + "\\convolution.cl"))
kernelString = sr.ReadToEnd();
Console.WriteLine("Reading kernel... done");
float[] selectedTransformation = Transformations.GetTransformation((TransformationType)comboBoxTransform.SelectedItem, 1.0f / float.Parse(textBoxPixelSize.Text), 1.0f / float.Parse(textBoxPixelSize.Text), 1.0f / float.Parse(textBoxPixelSize.Text), dx, dy, dz);
//create openCL program
ComputeProgram computeProgram = new ComputeProgram(computeContext, kernelString);
computeProgram.Build(computeContext.Devices, null, null, IntPtr.Zero);
ComputeProgramBuildStatus computeProgramBuildStatus = computeProgram.GetBuildStatus(_selectedComputeDevice);
Console.WriteLine("computeProgramBuildStatus\n\t"+computeProgramBuildStatus);
String buildLog = computeProgram.GetBuildLog(_selectedComputeDevice);
Console.WriteLine("buildLog");
if (buildLog.Equals("\n"))
Console.WriteLine("\tbuildLog is empty...");
else
Console.WriteLine("\t" + buildLog);
float[] fluorophores = CsvData.ReadFluorophores(_sourceFilename);
/////////////////////////////////////////////
// Create a Command Queue & Event List
/////////////////////////////////////////////
ComputeCommandQueue computeCommandQueue = new ComputeCommandQueue(computeContext, _selectedComputeDevice, ComputeCommandQueueFlags.None);
////////////////////////////////////////////////////////////////
// Create Buffers Transform
////////////////////////////////////////////////////////////////
ComputeBuffer<float> fluorophoresCoords = new ComputeBuffer<float>(computeContext, ComputeMemoryFlags.ReadWrite, fluorophores.LongLength);
ComputeBuffer<float> transformationMatrix = new ComputeBuffer<float>(computeContext, ComputeMemoryFlags.ReadOnly, selectedTransformation.LongLength);
/////////////////////////////////////////////
// Create the transformFluorophoresKernel
///////////////////////////////////////////////////////////
ComputeKernel transformFluorophoresKernel = computeProgram.CreateKernel("transform_fluorophores");
/////////////////////////////////////////////
// Set the transformFluorophoresKernel arguments
/////////////////////////////////////////////
transformFluorophoresKernel.SetMemoryArgument(0, fluorophoresCoords);
transformFluorophoresKernel.SetMemoryArgument(1, transformationMatrix);
/////////////////////////////////////////////
// Configure the work-item structure
/////////////////////////////////////////////
long[] globalWorkOffsetTransformFluorophoresKernel = null;
long[] globalWorkSizeTransformFluorophoresKernel = new long[] { fluorophores.Length / 4 };
long[] localWorkSizeTransformFluorophoresKernel = null;
////////////////////////////////////////////////////////
// Enqueue the transformFluorophoresKernel for execution
////////////////////////////////////////////////////////
computeCommandQueue.WriteToBuffer(fluorophores, fluorophoresCoords, true, null);
computeCommandQueue.WriteToBuffer(selectedTransformation, transformationMatrix, true, null);
computeCommandQueue.Execute(transformFluorophoresKernel, globalWorkOffsetTransformFluorophoresKernel, globalWorkSizeTransformFluorophoresKernel, localWorkSizeTransformFluorophoresKernel, null);
// computeCommandQueue.ExecuteTask(transformFluorophoresKernel, transformFluorophoresEvents);
float[] transformedFluorophores = new float[fluorophores.Length];
computeCommandQueue.ReadFromBuffer(fluorophoresCoords, ref transformedFluorophores, true, null);
computeCommandQueue.Finish();
//TODO remove, only for testing
// for (int i = 0; i < transformedFluorophores.Length; i++)
// {
// Console.WriteLine(transformedFluorophores[i]);
// }
// /TODO remove, only for testing
stopwatch.Stop();
Console.WriteLine("Transform fluophores duration:\n\t" + stopwatch.Elapsed);
stopwatch.Reset();
stopwatch.Start();
// fluorophoresCoords are now transformed (done in place)
////////////////////////////////////////////////////////////////
// Create Buffers Convolve Fluorophores
////////////////////////////////////////////////////////////////
const int convolve_kernel_lwgs = 16;
int totalBuffer = (int) Math.Ceiling(pixelCount / (float)convolve_kernel_lwgs) * convolve_kernel_lwgs;
ComputeBuffer<float> resultImage = new ComputeBuffer<float>(computeContext, ComputeMemoryFlags.WriteOnly, totalBuffer);
/////////////////////////////////////////////
// Create the transformFluorophoresKernel
/////////////////////////////////////////////
ComputeKernel convolveFluorophoresKernel = computeProgram.CreateKernel("convolve_fluorophores");
/////////////////////////////////////////////
// Set the convolveFluorophoresKernel arguments
/////////////////////////////////////////////
convolveFluorophoresKernel.SetMemoryArgument(0, resultImage);
convolveFluorophoresKernel.SetValueArgument(1, _imageDimensionX);
convolveFluorophoresKernel.SetValueArgument(2, _imageDimensionY);
convolveFluorophoresKernel.SetMemoryArgument(3, fluorophoresCoords);
convolveFluorophoresKernel.SetLocalArgument(4, convolve_kernel_lwgs);
convolveFluorophoresKernel.SetValueArgument(5, fluorophores.Length / 4);
/////////////////////////////////////////////
// Configure the work-item structure
/////////////////////////////////////////////
long[] globalWorkOffsetTransformConvolveFluorophoresKernel = null;
long[] globalWorkSizeTransformConvolveFluorophoresKernel = new long[] { pixelCount };
long[] localWorkSizeTransformConvolveFluorophoresKernel = new long[] {convolve_kernel_lwgs};
////////////////////////////////////////////////////////
// Enqueue the convolveFluorophoresKernel for execution
////////////////////////////////////////////////////////
computeCommandQueue.Execute(convolveFluorophoresKernel, globalWorkOffsetTransformConvolveFluorophoresKernel, globalWorkSizeTransformConvolveFluorophoresKernel, localWorkSizeTransformConvolveFluorophoresKernel, null);
float[] resultImageData = new float[totalBuffer];
computeCommandQueue.ReadFromBuffer(resultImage, ref resultImageData, true, null);
computeCommandQueue.Finish();
for (int i = 0; i < pixelCount; i++)
{
Console.WriteLine(resultImageData[i]);
}
Console.WriteLine("Writing data to file...");
// CsvData.WriteToDisk("..\\..\\..\\output.csv", resultImageData);
TiffData.WriteToDisk(resultImageData, _saveFilename, _imageDimensionX, _imageDimensionY);
Bitmap bitmap = new Bitmap(_imageDimensionX, _imageDimensionY);
float max = resultImageData.Max();
float scale = 255/(float)max;
// for (int r = 0; r < _imageDimensionY; r++)
// {
// for (int c = 0; c < _imageDimensionX; c++)
// {
// float value = resultImageData[c*(r + 1)];
// Color newColor = Color.FromArgb((int)(value * scale), (int)(value * scale), (int)(value * scale));
// bitmap.SetPixel(c,r, newColor);
// }
// }
ushort[] ushortdata = new ushort[resultImageData.Length];
for (int i = 0; i < resultImageData.Length; i++)
{
ushortdata[i] = (ushort)resultImageData[i];
}
uint[] convertGray16ToRgb = ConvertGray16ToRGB(ushortdata, 16);
byte[] bytes = new byte[convertGray16ToRgb.Length * 4];
//
// int[] resultImageData2 = new int[resultImageData.Length];
//
for (int index = 0; index < convertGray16ToRgb.Length; index++)
{
// resultImageData2[index] = (int)(scale*resultImageData[index]);
byte[] bytes1 = BitConverter.GetBytes(convertGray16ToRgb[index]);
bytes[index] = bytes1[0];
bytes[4 * index + 1] = bytes1[1];
bytes[4 * index + 2] = bytes1[2];
bytes[4 * index + 3] = bytes1[3];
}
//
// for (int r = 0; r < _imageDimensionY; r++)
// {
// for (int c = 0; c < _imageDimensionX; c++)
// {
// float value = resultImageData2[c*(r + 1)];
// Color newColor = Color.FromArgb((int)(value), (int)(value), (int)(value));
// bitmap.SetPixel(c,r, newColor);
// }
// }
// bitmap.Save("c:\\temp.bmp");
using (MemoryStream ms = new MemoryStream(bytes))
{
Image image = Bitmap.FromStream(ms);
image.Save("c:\\temp.bmp");
}
Console.WriteLine("Writing data to file... done");
stopwatch.Stop();
Console.WriteLine("Convolve fluophores duration:\n\t" + stopwatch.Elapsed);
Console.WriteLine("Computing... done");
}
private static bool prepTris()
{
int width = frame.GetLength(0);
int height = frame.GetLength(1);
List <tris> tris = new List <tris>();
List <vec> poss = new List <vec>();
List <vec> rots = new List <vec>();
int[] options = new int[4];
vec[] cam = new vec[3];
cam[0] = eye.toVec();
Vector look = new Vector(0, 0, 1, 0);
Vector up = new Vector(0, 1, 0, 0);
look *= Matrix4x4.rotationXMatrix(ConsoleSettings.cameraRot.x);
look *= Matrix4x4.rotationYMatrix(ConsoleSettings.cameraRot.y);
cam[1] = look.toVec();
cam[2] = up.toVec();
foreach (Mesh m in ConsoleRenderer.Meshes)
{
foreach (Triangle t in m.tris)
{
tris.Add(t.toTris());
poss.Add(m.position.toVec());
rots.Add(m.rotation.toVec());
}
}
options[0] = tris.Count;
options[1] = width;
options[2] = height;
if (tris.Count <= 0)
{
return(false);
}
if (lastTriCount != tris.Count)
{
gTris = new ComputeBuffer <tris>(context, ComputeMemoryFlags.ReadWrite | ComputeMemoryFlags.CopyHostPointer, tris.ToArray());
gOptions = new ComputeBuffer <int>(context, ComputeMemoryFlags.ReadWrite | ComputeMemoryFlags.CopyHostPointer, options);
gCam = new ComputeBuffer <vec>(context, ComputeMemoryFlags.ReadWrite | ComputeMemoryFlags.CopyHostPointer, cam);
gPos = new ComputeBuffer <vec>(context, ComputeMemoryFlags.ReadWrite | ComputeMemoryFlags.CopyHostPointer, poss.ToArray());
gRot = new ComputeBuffer <vec>(context, ComputeMemoryFlags.ReadWrite | ComputeMemoryFlags.CopyHostPointer, rots.ToArray());
gTexture = new ComputeBuffer <float>(context, ComputeMemoryFlags.ReadWrite | ComputeMemoryFlags.CopyHostPointer, texture);
}
else
{
commands.WriteToBuffer(tris.ToArray(), gTris, false, eventList);
commands.WriteToBuffer(options, gOptions, false, eventList);
commands.WriteToBuffer(cam, gCam, false, eventList);
commands.WriteToBuffer(rots.ToArray(), gRot, false, eventList);
commands.WriteToBuffer(poss.ToArray(), gPos, false, eventList);
commands.WriteToBuffer(texture, gTexture, false, eventList);
}
prepTrisKernel.SetMemoryArgument(0, gTris);
prepTrisKernel.SetMemoryArgument(1, gOptions);
prepTrisKernel.SetMemoryArgument(2, gRot);
prepTrisKernel.SetMemoryArgument(3, gPos);
prepTrisKernel.SetMemoryArgument(4, gCam);
commands.Execute(prepTrisKernel, null, new long[] { tris.Count }, null, eventList);
return(true);
}
/// <summary>
/// Copy the material data to the GPU.
/// </summary>
/// <param name="commandQueue"></param>
public void syncBuffer(ComputeCommandQueue commandQueue)
{
// TODO: try non-blocking sends.
commandQueue.WriteToBuffer<Material>(_materialArray, Buffer, true, null);
}
public static void VectorSumHandler(HttpListenerContext context)
{
HttpListenerRequest request = context.Request;
HttpListenerResponse response = context.Response;
string vecSum = @"
__kernel void vectorSum(__global float *v1, __global float *v2, __global float *v3) {
int i = get_global_id(0);
v3[i] = v1[i] + v2[i];
}
";
int size = 100000;
float[] v1_ = new float[size];
float[] v2_ = new float[size];
float[] v3_ = new float[size];
for (var i = 0; i < size; i++)
{
v1_[i] = (float)i;
v2_[i] = (float).5f;
}
var platform_ = ComputePlatform.Platforms[0];
ComputeContextPropertyList properties = new ComputeContextPropertyList(platform_);
ComputeContext ctx = new ComputeContext(ComputeDeviceTypes.Gpu, properties, null, IntPtr.Zero);
ComputeCommandQueue commands = new ComputeCommandQueue(ctx, ctx.Devices[0], ComputeCommandQueueFlags.None);
ComputeProgram program = new ComputeProgram(ctx, vecSum);
try
{
program.Build(null, null, null, IntPtr.Zero);
Console.WriteLine("program build completed");
}
catch
{
string log = program.GetBuildLog(ctx.Devices[0]);
}
ComputeBuffer <float> v1, v2, v3;
v1 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.ReadOnly | ComputeMemoryFlags.CopyHostPointer, v1_);
v2 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.ReadOnly | ComputeMemoryFlags.CopyHostPointer, v2_);
v3 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.WriteOnly | ComputeMemoryFlags.CopyHostPointer, v3_);
long[] worker = { size };
commands.WriteToBuffer(v1_, v1, false, null);
commands.WriteToBuffer(v2_, v2, false, null);
ComputeKernel sumKernal = program.CreateKernel("vectorSum");
Console.WriteLine("kernal created");
sumKernal.SetMemoryArgument(0, v1);
sumKernal.SetMemoryArgument(1, v2);
sumKernal.SetMemoryArgument(2, v3);
commands.Execute(sumKernal, null, worker, null, null);
Console.WriteLine("Executed");
commands.ReadFromBuffer <float>(v3, ref v3_, false, null);
StringBuilder sb = new StringBuilder();
for (int i = 0; i < size; i++)
{
sb.AppendFormat("{0} + {1} = {2}<br>", v1_[i].ToString(), v2_[i].ToString(), v3_[i].ToString());
}
var sum_expression_result = sb.ToString();
string responseString = string.Format("<html><body>{0}</body></html>", sum_expression_result);
byte[] buffer = Encoding.UTF8.GetBytes(responseString);
response.ContentLength64 = buffer.Length;
Stream output = response.OutputStream;
output.Write(buffer, 0, buffer.Length);
output.Close();
}
public static void CustomHandler(HttpListenerContext context)
{
HttpListenerRequest request = context.Request;
HttpListenerResponse response = context.Response;
string responseString;
Stream body = request.InputStream;
Encoding encoding = request.ContentEncoding;
StreamReader reader = new StreamReader(body, encoding);
var data = reader.ReadToEnd();
var props = JsonSerializer.Deserialize <OpenCLPOCO>(data);
string vecSum = props.code;
int size = props.v1.Length;
float[] v1_ = props.v1;
float[] v2_ = props.v2;
float[] v3_ = new float[size];
var platform_ = ComputePlatform.Platforms[0];
ComputeContextPropertyList properties = new ComputeContextPropertyList(platform_);
ComputeContext ctx = new ComputeContext(ComputeDeviceTypes.Gpu, properties, null, IntPtr.Zero);
ComputeCommandQueue commands = new ComputeCommandQueue(ctx, ctx.Devices[0], ComputeCommandQueueFlags.None);
ComputeProgram program = new ComputeProgram(ctx, vecSum);
try
{
program.Build(null, null, null, IntPtr.Zero);
Console.WriteLine("program build completed");
}
catch
{
string log = program.GetBuildLog(ctx.Devices[0]);
}
ComputeBuffer <float> v1, v2, v3;
v1 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.ReadOnly | ComputeMemoryFlags.CopyHostPointer, v1_);
v2 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.ReadOnly | ComputeMemoryFlags.CopyHostPointer, v2_);
v3 = new ComputeBuffer <float>(ctx, ComputeMemoryFlags.WriteOnly | ComputeMemoryFlags.CopyHostPointer, v3_);
long[] worker = { size };
commands.WriteToBuffer(v1_, v1, false, null);
commands.WriteToBuffer(v2_, v2, false, null);
try
{
ComputeKernel sumKernal = program.CreateKernel("vectorSum");
Console.WriteLine("kernal created");
sumKernal.SetMemoryArgument(0, v1);
sumKernal.SetMemoryArgument(1, v2);
sumKernal.SetMemoryArgument(2, v3);
commands.Execute(sumKernal, null, worker, null, null);
Console.WriteLine("Executed");
commands.ReadFromBuffer <float>(v3, ref v3_, false, null);
StringBuilder sb = new StringBuilder();
for (int i = 0; i < size; i++)
{
sb.AppendFormat("Ω({0}, {1}) = {2}<br>", v1_[i].ToString(), v2_[i].ToString(), v3_[i].ToString());
}
var sum_expression_result = sb.ToString();
responseString = string.Format("{0}", sum_expression_result);
} catch (Exception e)
{
responseString = e.Message;
}
byte[] buffer = Encoding.UTF8.GetBytes(responseString);
response.ContentLength64 = buffer.Length;
Stream output = response.OutputStream;
output.Write(buffer, 0, buffer.Length);
output.Close();
}