public static void MyFirstBlasEmulatorTest()
{
Console.WriteLine("MyTest()");
// Get GPU device
CudafyModes.Target = eGPUType.Emulator;
GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target);
// Create GPGPUBLAS (CUBLAS Wrapper)
using (GPGPUBLAS blas = GPGPUBLAS.Create(gpu))
{
const int N = 100;
float[] a = new float[N];
float[] b = new float[N];
float[] c = new float[N];
float alpha = -1;
float beta = 0;
float[] device_a = gpu.CopyToDevice(a);
float[] device_b = gpu.CopyToDevice(b);
float[] device_c = gpu.CopyToDevice(c);
int m = 10;
int n = 10;
int k = 10;
cublasOperation Op = cublasOperation.N;
blas.GEMM(m, k, n, alpha, device_a, device_b, beta, device_c, Op);
gpu.CopyFromDevice <float>(device_c, c);
}
}
public static float[] CallGPU()
{
CudafyModes.Target = eGPUType.OpenCL;
CudafyModes.DeviceId = 0;
CudafyTranslator.Language = eLanguage.OpenCL;
CudafyModule km = CudafyTranslator.Cudafy(ePlatform.Auto, eArchitecture.OpenCL, typeof(GPU));
GPGPU gpu = CudafyHost.GetDevice(eGPUType.OpenCL, 0);
gpu.LoadModule(km);
km.Serialize();
float[] input = Utils.GenerateRandomVector();
float[,,] NN = Utils.GenerateRandomMatrix().AsSingleDimension();
float[] output = new float[Utils.N];
Stopwatch gpuSW = new Stopwatch();
gpuSW.Start();
float[] dev_output = gpu.Allocate <float>(output);
float[] dev_input = gpu.CopyToDevice(input);
float[,,] dev_NN = gpu.CopyToDevice(NN);
gpu.Launch(Utils.GRID_SIZE, Utils.BLOCK_SIZE).CalculateNeuralNetwork(dev_input, dev_NN, dev_output);
gpu.CopyFromDevice(dev_output, output);
gpu.FreeAll();
gpuSW.Stop();
Console.WriteLine("GPU: " + gpuSW.ElapsedMilliseconds);
return(output);
}
//
// http://stackoverflow.com/questions/18628447/cudafy-throws-an-exception-while-testing
//
private static void BlasSample(int deviceId)
{
CudafyModes.Target = eGPUType.Emulator;
GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target, deviceId);
CudafyModes.DeviceId = deviceId;
eArchitecture arch = gpu.GetArchitecture();
CudafyModule km = CudafyTranslator.Cudafy(arch);
gpu.LoadModule(km);
GPGPUBLAS blas = GPGPUBLAS.Create(gpu);
const int N = 100;
float[] a = new float[N];
float[] b = new float[N];
float[] c = new float[N];
float alpha = -1;
float beta = 0;
float[] device_a = gpu.CopyToDevice(a);
float[] device_b = gpu.CopyToDevice(b);
float[] device_c = gpu.CopyToDevice(c);
int m = 10;
int n = 10;
int k = 10;
cublasOperation Op = cublasOperation.N;
blas.GEMM(m, k, n, alpha, device_a, device_b, beta, device_c, Op);
throw new NotImplementedException();
}
public static void Main()
{
Console.WriteLine("CUDAfy Example\nCollecting necessary resources...");
CudafyModes.Target = eGPUType.Cuda; // To use OpenCL, change this enum
CudafyModes.DeviceId = 0;
CudafyTranslator.Language = CudafyModes.Target == eGPUType.OpenCL ? eLanguage.OpenCL : eLanguage.Cuda;
//Check for available devices
if (CudafyHost.GetDeviceCount(CudafyModes.Target) == 0)
{
throw new System.ArgumentException("No suitable devices found.", "original");
}
//Init device
var gpu = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
Console.WriteLine("Running example using {0}", gpu.GetDeviceProperties(false).Name);
//Load module for GPU
var km = CudafyTranslator.Cudafy();
gpu.LoadModule(km);
//Define local arrays
var a = new int[N];
var b = new int[N];
var c = new int[N];
// allocate the memory on the GPU
var dev_c = gpu.Allocate <int>(c);
// fill the arrays 'a' and 'b' on the CPU
for (var i = 0; i < N; i++)
{
a[i] = i;
b[i] = i * i;
}
// copy the arrays 'a' and 'b' to the GPU
var dev_a = gpu.CopyToDevice(a);
var dev_b = gpu.CopyToDevice(b);
gpu.Launch(1, N).add(dev_a, dev_b, dev_c);
// copy the array 'c' back from the GPU to the CPU
gpu.CopyFromDevice(dev_c, c);
// display the results
for (var i = 0; i < N; i++)
{
Console.WriteLine("{0} + {1} = {2}", a[i], b[i], c[i]);
}
// free the memory allocated on the GPU
gpu.FreeAll();
Console.WriteLine("Done!");
Console.ReadKey();
}
public void SetUp()
{
//CudafyModes.Architecture = eArchitecture.sm_30;
_gpu = CudafyHost.GetDevice(eArchitecture.sm_30, CudafyModes.DeviceId);
Assert.IsFalse(_gpu is OpenCLDevice, "OpenCL devices are not supported.");
_cm = CudafyModule.TryDeserialize();
if (_cm == null || !_cm.TryVerifyChecksums())
{
_cm = CudafyTranslator.Cudafy(eArchitecture.sm_30);
Console.WriteLine(_cm.CompilerOutput);
_cm.TrySerialize();
}
_gpu.LoadModule(_cm);
inputIntArray = new int[] { 0x17, 0x01, 0x7f, 0xd1, 0xfe, 0x23, 0x2c, 0xa0, 0x00, 0xcf, 0xaa, 0x7a, 0x35, 0xf4, 0x04, 0xbc,
0xe9, 0x6d, 0xb2, 0x55, 0xb0, 0xc8, 0x10, 0x49, 0x76, 0x17, 0x92, 0xab, 0xf3, 0xf2, 0xab, 0xcb }; // arbitrary values
d_inputIntArray = _gpu.CopyToDevice(inputIntArray);
d_outputIntArray = _gpu.Allocate <int>(WARP_SIZE);
gpuIntResult = new int[WARP_SIZE];
cpuIntResult = new int[WARP_SIZE];
inputFloatArray = new float[] { 1.7f, -37.03f, 2147.6436f, -0.1f, 7.7f, 99.99f, -809.142f, -0.1115f,
1.0f, 2.0f, 3.0f, 5.0f, 7.5f, 0.1001f, 11.119f, -9.0f,
7749.9847f, -860249.118843f, 0.0f, -2727745.586215f, 12.0f, -11.0f, 77.77f, 22.0f,
377.1112f, -377.1112f, 0.12345f, -0.12345f, 0.11111f, -0.11111f, 700000f, -14f }; // arbitrary values
d_inputFloatArray = _gpu.CopyToDevice(inputFloatArray);
d_outputFloatArray = _gpu.Allocate <float>(WARP_SIZE);
gpuFloatResult = new float[WARP_SIZE];
cpuFloatResult = new float[WARP_SIZE];
}
public void SetUp()
{
_gpu = CudafyHost.GetDevice(CudafyModes.Architecture, CudafyModes.DeviceId);
_byteBufferIn = new byte[N];
_byteBufferOut = new byte[N];
_sbyteBufferIn = new sbyte[N];
_sbyteBufferOut = new sbyte[N];
_ushortBufferIn = new ushort[N];
_ushortBufferOut = new ushort[N];
_uintBufferIn = new uint[N];
_uintBufferOut = new uint[N];
_ulongBufferIn = new ulong[N];
_ulongBufferOut = new ulong[N];
_cplxDBufferIn = new ComplexD[N];
_cplxDBufferOut = new ComplexD[N];
_cplxFBufferIn = new ComplexF[N];
_cplxFBufferOut = new ComplexF[N];
SetInputs();
ClearOutputsAndGPU();
}
public static void Execute()
{
CudafyModule km = CudafyModule.TryDeserialize();
if (km == null || !km.TryVerifyChecksums())
{
km = CudafyTranslator.Cudafy(typeof(Generic <ushort, ushort>), typeof(SimpleGeneric));
km.Serialize();
}
GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target);
gpu.LoadModule(km);
var input = new Generic <ushort, ushort>();
input.A = 187;
int[] devoutput = gpu.Allocate <int>(1);
gpu.Launch(1, 1, "Kernel", input, devoutput);
int output;
gpu.CopyFromDevice(devoutput, out output);
Console.WriteLine("Simple Generic: " + ((output == 1) ? "PASSED" : "FAILED"));
}
public void InitGPU()
{
// Work around for bug in Cudafy trying to find the path..
var os64Bit = Environment.Is64BitOperatingSystem;
if (os64Bit)
{
var dir = Environment.GetEnvironmentVariable("ProgramFiles");
Environment.SetEnvironmentVariable("ProgramFiles", "C:\\Program Files\\");
dir = Environment.GetEnvironmentVariable("ProgramFiles");
}
if (Gpu == null)
{
Gpu = CudafyHost.GetDevice(_gpuType, 0);
//Blas = GPGPUBLAS.Create(Gpu);
if (_gpuType == eGPUType.Cuda)
{
Blas = new SharpBLAS(Gpu);
Rand = GPGPURAND.Create(Gpu, curandRngType.CURAND_RNG_PSEUDO_DEFAULT);
Rand.SetPseudoRandomGeneratorSeed((ulong)RandomHelpers.Next(9999));
}
CudafyTranslator.GenerateDebug = true;
Debug.WriteLine("CUDA workdir = " + CudafyTranslator.WorkingDirectory);
Console.WriteLine("Recompile module");
CudafyTranslator.Language = eLanguage.Cuda;
var km = CudafyTranslator.Cudafy(eArchitecture.sm_30);
km = CudafyTranslator.Cudafy();
km.Serialize();
Gpu.LoadModule(km);
}
}
/// <summary>
/// Вызов и исполнение одной элементарной функции по имени функции
/// </summary>
/// <param name="function"></param>
public static void Execute(string function)
{
Debug.Assert(_indexes1.Last() == _sequencies1.Length);
Debug.Assert(_indexes2.Last() == _sequencies2.Length);
CudafyModule km = CudafyTranslator.Cudafy();
GPGPU gpu = CudafyHost.GetDevice();
gpu.LoadModule(km);
// copy the arrays 'a' and 'b' to the GPU
int[] devIndexes1 = gpu.CopyToDevice(_indexes1);
int[] devIndexes2 = gpu.CopyToDevice(_indexes2);
int[] devSequencies1 = gpu.CopyToDevice(_sequencies1);
int[] devSequencies2 = gpu.CopyToDevice(_sequencies2);
int[,] devMatrix = gpu.Allocate(_matrix);
int rows = _matrix.GetLength(0);
int columns = _matrix.GetLength(1);
dim3 gridSize = Math.Min(15, (int)Math.Pow((double)rows * columns, 0.33333333333));
dim3 blockSize = Math.Min(15, (int)Math.Pow((double)rows * columns, 0.33333333333));
gpu.Launch(gridSize, blockSize, function,
devSequencies1, devIndexes1,
devSequencies2, devIndexes2,
devMatrix);
// copy the array 'c' back from the GPU to the CPU
gpu.CopyFromDevice(devMatrix, _matrix);
// free the memory allocated on the GPU
gpu.FreeAll();
}
/// <summary>
/// Вызов и исполнение функции проверки что массив отсортирован
/// </summary>
public static void ExecuteSorted(int direction = 1)
{
CudafyModule km = CudafyTranslator.Cudafy();
GPGPU gpu = CudafyHost.GetDevice();
gpu.LoadModule(km);
int[] devA = gpu.Allocate(_a);
int[] devB = gpu.Allocate(_b);
int[] devC = gpu.Allocate(_c);
int[] devD = gpu.Allocate(D);
gpu.CopyToDevice(_a, devA);
gpu.Launch(1, 1).Split(devA, devB, devC, _middle);
gpu.Launch(_gridSize, _blockSize).Sorted(devA, devB, devC, devD, 0, direction);
gpu.Launch(1, 1).Sorted(devA, devB, devC, devD, 1, direction);
gpu.CopyFromDevice(devD, D);
// free the memory allocated on the GPU
gpu.FreeAll();
}
public void Test_TwoThreadTwoGPU()
{
_gpu0 = CudafyHost.CreateDevice(CudafyModes.Target, 0);
_gpu1 = CudafyHost.CreateDevice(CudafyModes.Target, 1);
_gpu0.EnableMultithreading();
_gpu1.EnableMultithreading();
bool j1 = false;
bool j2 = false;
for (int i = 0; i < 10; i++)
{
Console.WriteLine(i);
Thread t1 = new Thread(Test_TwoThreadTwoGPU_Thread0);
Thread t2 = new Thread(Test_TwoThreadTwoGPU_Thread1);
t1.Start();
t2.Start();
j1 = t1.Join(10000);
j2 = t2.Join(10000);
if (!j1 || !j2)
{
break;
}
}
_gpu0.DisableMultithreading();
_gpu0.FreeAll();
_gpu1.DisableMultithreading();
_gpu1.FreeAll();
Assert.IsTrue(j1);
Assert.IsTrue(j2);
}
public void SetUp()
{
_gpu = CudafyHost.CreateDevice(CudafyModes.Target);
_hostInput = new double[N * BATCH];
_hostInputCplx = new ComplexD[N * BATCH];
_hostOutput = new double[N * BATCH];
_hostOutputCplx = new ComplexD[N * BATCH];
_devInput = _gpu.Allocate(_hostInput);
_devInputCplx = _gpu.Allocate(_hostInputCplx);
_devInter = _gpu.Allocate <double>(N * 2 * BATCH);
_devInterCplx = _gpu.Allocate <ComplexD>(N * BATCH);
_devOutput = _gpu.Allocate(_hostOutput);
_devOutputCplx = _gpu.Allocate(_hostOutputCplx);
_fft = GPGPUFFT.Create(_gpu);
for (int b = 0; b < BATCH; b++)
{
for (int i = 0; i < N; i++)
{
ComplexD cf = new ComplexD();
cf.x = (double)((10.0F * Math.Sin(100 * 2 * Math.PI * i / N * Math.PI / 180)));
cf.y = (double)((10.0F * Math.Sin(200 * 2 * Math.PI * i / N * Math.PI / 180)));
_hostInput[i + b * N] = cf.x;
_hostInputCplx[i + b * N] = cf;
}
}
}
internal TensorOpGpu()
{
//CudafyTranslator.GenerateDebug = true;
Module = CudafyTranslator.Cudafy();
Gpu = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
Gpu.LoadModule(Module);
}
internal static AnswerStruct GetAnswer()
{
using (var gpu = CudafyHost.GetDevice())
{
gpu.LoadModule(CudafyTranslator.Cudafy());
var answer = new AnswerStruct[BlocksPerGrid];;
var gpuAnswer = gpu.Allocate(answer);
gpu.Launch(BlocksPerGrid, ThreadsPerBlock,
GpuFindPathDistance, gpuAnswer);
gpu.Synchronize();
gpu.CopyFromDevice(gpuAnswer, answer);
gpu.FreeAll();
var bestDistance = float.MaxValue;
var bestPermutation = 0L;
for (var i = 0; i < BlocksPerGrid; i++)
{
if (answer[i].distance < bestDistance)
{
bestDistance = answer[i].distance;
bestPermutation = answer[i].pathNo;
}
}
return(new AnswerStruct
{
distance = bestDistance,
pathNo = bestPermutation
});
}
}
public static void cudaTranspose(ref MathNet.Numerics.LinearAlgebra.Double.DenseMatrix dm)
{
GPGPU gpu = CudafyHost.GetDevice(eGPUType.Cuda);
GPGPUBLAS blas = GPGPUBLAS.Create(gpu);
int cols = dm.ColumnCount, rows = dm.RowCount;
int restRows = rows - cols;
//double[] a = dm.Storage.ToColumnMajorArray();
double[] a = dm.SubMatrix(0, cols, 0, cols).Storage.ToColumnMajorArray();
double[] b = dm.SubMatrix(cols, restRows, 0, cols).Storage.ToColumnMajorArray();
dm = null;
double[] a_d = gpu.CopyToDevice <double>(a);
a = null;
double[] c_d = gpu.Allocate <double>(cols * cols);
double[] x_d = gpu.CopyToDevice <double>(new double[] { 1 });
blas.GEMV(cols, cols, 1, c_d, x_d, 0, x_d, Cudafy.Maths.BLAS.Types.cublasOperation.T);
a = new double[cols * rows];
gpu.CopyFromDevice <double>(c_d, 0, a, 0, cols * cols);
gpu.FreeAll();
a_d = gpu.CopyToDevice <double>(b);
b = null;
c_d = gpu.Allocate <double>(restRows * cols);
x_d = gpu.CopyToDevice <double>(new double[] { 1 });
blas.GEMV(restRows, cols, 1, c_d, x_d, 0, x_d, Cudafy.Maths.BLAS.Types.cublasOperation.T);
gpu.CopyFromDevice <double>(c_d, 0, a, cols * cols, restRows * cols);
gpu.FreeAll();
dm = new MathNet.Numerics.LinearAlgebra.Double.DenseMatrix(cols, rows, a);
}
public static void Execute()
{
int i = 0;
foreach (GPGPUProperties prop in CudafyHost.GetDeviceProperties(CudafyModes.Target, false))
{
Console.WriteLine(" --- General Information for device {0} ---", i);
Console.WriteLine("Name: {0}", prop.Name);
Console.WriteLine("Platform Name: {0}", prop.PlatformName);
Console.WriteLine("Device Id: {0}", prop.DeviceId);
Console.WriteLine("Compute capability: {0}.{1}", prop.Capability.Major, prop.Capability.Minor);
Console.WriteLine("Clock rate: {0}", prop.ClockRate);
Console.WriteLine("Simulated: {0}", prop.IsSimulated);
Console.WriteLine();
Console.WriteLine(" --- Memory Information for device {0} ---", i);
Console.WriteLine("Total global mem: {0}", prop.TotalMemory);
Console.WriteLine("Total constant Mem: {0}", prop.TotalConstantMemory);
Console.WriteLine("Max mem pitch: {0}", prop.MemoryPitch);
Console.WriteLine("Texture Alignment: {0}", prop.TextureAlignment);
Console.WriteLine();
Console.WriteLine(" --- MP Information for device {0} ---", i);
Console.WriteLine("Shared mem per mp: {0}", prop.SharedMemoryPerBlock);
Console.WriteLine("Registers per mp: {0}", prop.RegistersPerBlock);
Console.WriteLine("Threads in warp: {0}", prop.WarpSize);
Console.WriteLine("Max threads per block: {0}", prop.MaxThreadsPerBlock);
Console.WriteLine("Max thread dimensions: ({0}, {1}, {2})", prop.MaxThreadsSize.x, prop.MaxThreadsSize.y, prop.MaxThreadsSize.z);
Console.WriteLine("Max grid dimensions: ({0}, {1}, {2})", prop.MaxGridSize.x, prop.MaxGridSize.y, prop.MaxGridSize.z);
Console.WriteLine();
i++;
}
}
static void Main(string[] args)
{
CudafyModule km = CudafyTranslator.Cudafy();
GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
gpu.LoadModule(km);
int numFrames = numberOfSeconds * framesPerSecond;
InitializeParticles();
File.WriteAllText("length.txt", numFrames.ToString());
for (int i = 0; i < numFrames; i++)
{
DateTime frameStart = DateTime.Now;
Simulate(gpu);
Bitmap frame = Render(gpu, i);
TimeSpan frameTime = DateTime.Now - frameStart;
Console.WriteLine("Frame " + i + " complete. Time: " + frameTime.TotalMilliseconds + "ms");
}
}
public static bool TestGpuDoublePrecision(int DeviceId)
{
if (DeviceId > CudafyHost.GetDeviceCount(eGPUType.OpenCL))
{
return(false);
}
try
{
CudafyModes.Target = eGPUType.OpenCL;
CudafyTranslator.Language = eLanguage.OpenCL;
CudafyModule km = CudafyTranslator.Cudafy();
GPGPU gpu = CudafyHost.GetDevice(eGPUType.OpenCL, DeviceId);
gpu.LoadModule(km);
double c;
double[] dev_c = gpu.Allocate <double>();
gpu.Launch().add_double(2.5d, 7.5d, dev_c);
gpu.CopyFromDevice(dev_c, out c);
gpu.Free(dev_c);
return(c == 10.0d);
}
catch
{ return(false); }
}
public void SetUp()
{
_gpu = CudafyHost.CreateDevice(CudafyModes.Target);
Console.WriteLine(_gpu.GetDriverVersion());
_fft = GPGPUFFT.Create(_gpu);
_hostInput = new float[N * BATCH];
_hostInputCplx = new ComplexF[N * BATCH];
_hostOutput = new float[N * BATCH];
_hostOutputCplx = new ComplexF[N * BATCH];
_devInput = _gpu.Allocate(_hostInput);
_devInputCplx = _gpu.Allocate(_hostInputCplx);
_devInter = _gpu.Allocate <float>(N * 2 * BATCH);
_devInterCplx = _gpu.Allocate <ComplexF>(N * BATCH);
_devOutput = _gpu.Allocate(_hostOutput);
_devOutputCplx = _gpu.Allocate(_hostOutputCplx);
Console.WriteLine(_fft.GetVersion());
for (int b = 0; b < BATCH; b++)
{
for (int i = 0; i < N; i++)
{
ComplexF cf = new ComplexF();
cf.x = (float)((10.0F * Math.Sin(100 * 2 * Math.PI * i / N * Math.PI / 180)));
cf.y = (float)((10.0F * Math.Sin(200 * 2 * Math.PI * i / N * Math.PI / 180)));
_hostInput[i + b * N] = cf.x;
_hostInputCplx[i + b * N] = cf;
}
}
}
/// <summary>
/// Выполнение сортировки слияниями
/// Пример использования:
/// CudafySequencies.SetSequencies(arrayOfArray,arrayOfArray);
/// CudafySequencies.Execute("Compare");
/// var compare = CudafySequencies.GetMartix();
/// CudafyArray.SetArray(Enumerable.Range(0,n).ToArray());
/// CudafyArray.SetCompare(compare);
/// CudafyArray.MergeSort();
/// var indexesOfSorted = CudafyArray.GetArray();
/// </summary>
public static void MergeSort(int direction = 1)
{
CudafyModule km = CudafyTranslator.Cudafy();
GPGPU gpu = CudafyHost.GetDevice();
gpu.LoadModule(km);
int[] devA = gpu.Allocate(_a);
int[] devB = gpu.Allocate(_b);
gpu.CopyToDevice(_a, devA);
for (int i = 0; i < _ceiling; i++)
{
int gridSize = Math.Min(15, (int)Math.Pow((_length >> i) + i, 0.333333333333));
int blockSize = Math.Min(15, (int)Math.Pow((_length >> i) + i, 0.333333333333));
gpu.Launch(gridSize, blockSize)
.MergeLinear(((i & 1) == 0) ? devA : devB, ((i & 1) == 0) ? devB : devA, i, 0,
_length,
direction);
}
gpu.CopyFromDevice(((_ceiling & 1) == 0) ? devA : devB, _a);
// free the memory allocated on the GPU
gpu.FreeAll();
}
public static void primaGPU()
{
CudafyModule modul_kernel = CudafyTranslator.Cudafy();
GPGPU vga = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
vga.LoadModule(modul_kernel);
Stopwatch waktu = new Stopwatch();
waktu.Start();
int[] list_cpu = new int[KONSTANTA_THREAD];
int[] list_cpy = new int[KONSTANTA_THREAD];
int[] list = vga.Allocate <int>(KONSTANTA_THREAD);
vga.Launch(KONSTANTA_THREAD, 1).ModulAtomic(list);
vga.CopyFromDevice(list, list_cpy);
vga.FreeAll();
int index = 0;
for (int z = 0; z < list_cpy.Length; z++)
{
if (list_cpy[z] != -1)
{
list_cpu[index] = list_cpy[z];
//Console.WriteLine(list_cpu[index]);
index++;
}
}
waktu.Stop();
TimeSpan ts = waktu.Elapsed;
String total = ts.Seconds.ToString();
Console.WriteLine("Total GPU ------ {0} detik> ", total);
}
public static void eksekusi()
{
CudafyModule kernel_modul = CudafyTranslator.Cudafy();
GPGPU vga = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
vga.LoadModule(kernel_modul);
Stopwatch waktu = new Stopwatch();
waktu.Start();
int[] array_vga = vga.Allocate <int>(KONSTANTA_THREAD);
int[] array_hasil = new int[KONSTANTA_THREAD];
//long[] matriks1 = vga.Allocate<long>(KONSTANTA_THREAD);
//long[] matriks2 = vga.Allocate<long>(KONSTANTA_THREAD);//new int[KONSTANTA_THREAD];
//long[] matriks3 = vga.Allocate<long>(KONSTANTA_THREAD); //[KONSTANTA_THREAD];
vga.Launch(KONSTANTA_THREAD, 1).fungsiAtomic(array_vga);
vga.CopyFromDevice(array_vga, array_hasil);
vga.FreeAll();
//for(int z = 0; z < array_hasil.Length; z++)
//{
// Console.WriteLine("Hasil Ekstrak----" + array_hasil[z]);
//}
vga.FreeAll();
waktu.Stop();
TimeSpan ts = waktu.Elapsed;
String total = ts.Milliseconds.ToString();
Console.WriteLine("Total VGA ------ > " + total);
}
public void Test_TwoThreadCopy()
{
_gpu = CudafyHost.GetDevice(eGPUType.Cuda);
_gpuuintBufferIn3 = _gpu.Allocate(_uintBufferIn1);
_gpuuintBufferIn4 = _gpu.Allocate(_uintBufferIn1);
_gpu.EnableMultithreading();
bool j1 = false;
bool j2 = false;
for (int i = 0; i < 10; i++)
{
Console.WriteLine(i);
SetInputs();
ClearOutputs();
Thread t1 = new Thread(Test_TwoThreadCopy_Thread1);
Thread t2 = new Thread(Test_TwoThreadCopy_Thread2);
t1.Start();
t2.Start();
j1 = t1.Join(10000);
j2 = t2.Join(10000);
if (!j1 || !j2)
{
break;
}
}
_gpu.DisableMultithreading();
_gpu.FreeAll();
Assert.IsTrue(j1);
Assert.IsTrue(j2);
}
public static void Execute()
{
CudafyModule km = CudafyTranslator.Cudafy(typeof(ParamsStruct), typeof(ImpliedVolatile));
_gpu = CudafyHost.GetDevice(CudafyModes.Target);
_gpu.LoadModule(km);
ParamsStruct[] host_par = new ParamsStruct[1];
ParamsStruct[] result = new ParamsStruct[1];
host_par[0].OP = 96.95;
host_par[0].Price = 1332.24;
host_par[0].Strike = 1235;
host_par[0].TD = 31;
host_par[0].R = 0.0001355;
host_par[0].Q = 0.0166;
host_par[0].N = 100;// 1000;
host_par[0].kind = 1;
ParamsStruct[] dev_par = _gpu.CopyToDevice(host_par);
float[] PA = _gpu.Allocate <float>(1001);
_gpu.Launch(1, 1, "impliedVolatile", dev_par, PA);
_gpu.CopyFromDevice(dev_par, 0, result, 0, 1);
Console.WriteLine("I={0}, B={1}", result[0].i, result[0].B);
//Console.ReadKey();
}
public void ExeTestKernel()
{
GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target, 0);
eArchitecture arch = gpu.GetArchitecture();
CudafyModule km = CudafyTranslator.Cudafy(arch);
gpu.LoadModule(km);
int[] host_results = new int[N];
// Either assign a new block of memory to hold results on device
var dev_results = gpu.Allocate <int>(N);
// Or fill your array with values first and then
for (int i = 0; i < N; i++)
{
host_results[i] = i * 3;
}
// Copy array with ints to device
var dev_filled_results = gpu.CopyToDevice(host_results);
// 64*16 = 1024 threads per block (which is max for sm_30)
dim3 threadsPerBlock = new dim3(64, 16);
// 8*8 = 64 blocks per grid , just for show so you get varying numbers
dim3 blocksPerGrid = new dim3(8, 8);
//var threadsPerBlock = 1024; // this will only give you blockDim.x = 1024, .y = 0, .z = 0
//var blocksPerGrid = 1; // just for show
gpu.Launch(blocksPerGrid, threadsPerBlock, "GenerateRipples", dev_results, dev_filled_results);
gpu.CopyFromDevice(dev_results, host_results);
}
public static uint[] Evaluate(ulong[] hands, int numCards)
{
// Translates this class to CUDA C and then compliles
CudafyModule km = CudafyTranslator.Cudafy();//eArchitecture.sm_20);
// Get the first GPU and load the module
GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
gpu.LoadModule(km);
int blockSize = 256;
int blockx = hands.Length / blockSize;
if (hands.Length % blockSize != 0)
{
blockx++;
}
ulong[] dev_hands = gpu.Allocate <ulong>(hands.Length);
uint[] dev_ranks = gpu.Allocate <uint>(hands.Length);
gpu.CopyToDevice(hands, dev_hands);
gpu.StartTimer();
gpu.Launch(blockx, blockSize).evaluate(dev_hands, numCards, hands.Length, dev_ranks);
var ts = gpu.StopTimer();
uint[] toReturn = new uint[hands.Length];
gpu.CopyFromDevice(dev_ranks, toReturn);
return(toReturn);
}
public void SetUp()
{
_gpu = CudafyHost.GetDevice();
_sparse = GPGPUSPARSE.Create(_gpu);
_blas = GPGPUBLAS.Create(_gpu);
_solver = new Solver(_gpu, _blas, _sparse);
}
public void SetUp()
{
_gpu = CudafyHost.CreateDevice(CudafyModes.Target, CudafyModes.DeviceId);
_uintBufferIn1 = new uint[N];
_uintBufferOut1 = new uint[N];
_uintBufferIn2 = new uint[N];
_uintBufferOut2 = new uint[N];
}
private void PrintOutAviableDevices()
{
Console.WriteLine("Printing out avaiable devices...");
Console.WriteLine("For now this code will work only with Cuda devs...");
var numberOfCudaDevices = CudafyHost.GetDeviceCount(eGPUType.Cuda);
var numberOfOpenClDevices = CudafyHost.GetDeviceCount(eGPUType.OpenCL);
var numberOfEmulatorDevices = CudafyHost.GetDeviceCount(eGPUType.Emulator);
Console.WriteLine("{0} devices of type Cuda found", numberOfCudaDevices);
Console.WriteLine("{0} devices of type OpenCl found", numberOfOpenClDevices);
Console.WriteLine("{0} devices of type Emulator found", numberOfEmulatorDevices);
Console.WriteLine("Attempting to print out detailed info about Cuda devices..");
var cudaDevicesProperties = CudafyHost.GetDeviceProperties(eGPUType.Cuda);
if (cudaDevicesProperties.Count() != numberOfCudaDevices)
{
Console.WriteLine("Something is terribly off! Number of cuda devices differ from received properites");
}
foreach (var cudaDeviceProperties in cudaDevicesProperties)
{
Console.WriteLine(@"---");
PrintOutObjectPublicProperties(cudaDeviceProperties);
Console.WriteLine(@"---");
}
Console.WriteLine("Attempting to print out detailed info about openCl devices..");
var openClDevicesProperties = CudafyHost.GetDeviceProperties(eGPUType.OpenCL);
if (openClDevicesProperties.Count() != numberOfOpenClDevices)
{
Console.WriteLine("Something is terribly off! Number of openCl devices differ from received properites");
}
foreach (var openClDeviceProperties in openClDevicesProperties)
{
Console.WriteLine(@"---");
PrintOutObjectPublicProperties(openClDeviceProperties);
Console.WriteLine(@"---");
}
Console.WriteLine("Attempting to print out detailed info about emulator devices..");
var emulatorDevicesProperties = CudafyHost.GetDeviceProperties(eGPUType.Emulator);
if (emulatorDevicesProperties.Count() != numberOfEmulatorDevices)
{
Console.WriteLine("Something is terribly off! Number of emulator devices differ from received properites");
}
foreach (var emulatorDeviceProperties in emulatorDevicesProperties)
{
Console.WriteLine(@"---");
PrintOutObjectPublicProperties(emulatorDeviceProperties);
Console.WriteLine(@"---");
}
}
public void Initialize(int DeviceId)
{
CudafyModes.Target = eGPUType.OpenCL;
CudafyTranslator.Language = eLanguage.OpenCL;
CudafyModule km = CudafyTranslator.Cudafy();
Gpu = CudafyHost.GetDevice(eGPUType.OpenCL, DeviceId);
Gpu.LoadModule(km);
}