public void TestAddCUDA()
{
//Alloc device memory
int count_rows_a = 2;
int count_shared = 2;
int count_cols_b = 2;
double alfa = 1.0;
double beta = 1.0;
CudaDeviceVariable <double> A = new double[] { 0, 0, 0, 0 };
CudaDeviceVariable <double> B = new double[] { 0, 0, 0, 0 };
CudaDeviceVariable <double> C = new double[] { 0, 0, 0, 0 };
//Clean up
CudaBlas blas = new CudaBlas();
blas.Gemm(Operation.NonTranspose, Operation.NonTranspose,
count_rows_a, count_shared, count_cols_b,
alfa,
A, count_rows_a,
B, count_shared,
beta,
C, count_rows_a);
//Copy data back to host
double[] result = C;
ToolsArray.print(result);
Assert.AreEqual(0, result[0], 0.001);
// Um right this should be moved
blas.Dispose();
}
protected virtual void Dispose(bool disposing)
{
if (disposing && !_disposed)
{
_blas.Dispose();
_cuda.Dispose();
//if(_solver.IsValueCreated)
// _solver.Value.Dispose();
_numerics.Dispose();
_disposed = true;
}
}
public override void Dispose()
{
cuBlas.Dispose();
cudaContext.Dispose();
}
static void Main(string[] args)
{
int N = 275;
float[] h_A;
float[] h_B;
float[] h_C;
float[] h_C_ref;
CudaDeviceVariable <float> d_A;
CudaDeviceVariable <float> d_B;
CudaDeviceVariable <float> d_C;
float alpha = 1.0f;
float beta = 0.0f;
int n2 = N * N;
int i;
float error_norm;
float ref_norm;
float diff;
CudaBlas handle;
/* Initialize CUBLAS */
Console.WriteLine("simpleCUBLAS test running.");
handle = new CudaBlas();
/* Allocate host memory for the matrices */
h_A = new float[n2];
h_B = new float[n2];
//h_C = new float[n2];
h_C_ref = new float[n2];
Random rand = new Random(0);
/* Fill the matrices with test data */
for (i = 0; i < n2; i++)
{
h_A[i] = (float)rand.NextDouble();
h_B[i] = (float)rand.NextDouble();
//h_C[i] = (float)rand.NextDouble();
}
/* Allocate device memory for the matrices */
d_A = new CudaDeviceVariable <float>(n2);
d_B = new CudaDeviceVariable <float>(n2);
d_C = new CudaDeviceVariable <float>(n2);
/* Initialize the device matrices with the host matrices */
d_A.CopyToDevice(h_A);
d_B.CopyToDevice(h_B);
//d_C.CopyToDevice(h_C);
/* Performs operation using plain C code */
simple_sgemm(N, alpha, h_A, h_B, beta, h_C_ref);
/* Performs operation using cublas */
handle.Gemm(Operation.NonTranspose, Operation.NonTranspose, N, N, N, alpha, d_A, N, d_B, N, beta, d_C, N);
/* Allocate host memory for reading back the result from device memory */
h_C = d_C;
/* Check result against reference */
error_norm = 0;
ref_norm = 0;
for (i = 0; i < n2; ++i)
{
diff = h_C_ref[i] - h_C[i];
error_norm += diff * diff;
ref_norm += h_C_ref[i] * h_C_ref[i];
}
error_norm = (float)Math.Sqrt((double)error_norm);
ref_norm = (float)Math.Sqrt((double)ref_norm);
if (Math.Abs(ref_norm) < 1e-7)
{
Console.WriteLine("!!!! reference norm is 0");
return;
}
/* Memory clean up */
d_A.Dispose();
d_B.Dispose();
d_C.Dispose();
/* Shutdown */
handle.Dispose();
if (error_norm / ref_norm < 1e-6f)
{
Console.WriteLine("simpleCUBLAS test passed.");
return;
}
else
{
Console.WriteLine("simpleCUBLAS test failed.");
return;
}
}
static void Main(string[] args)
{
int N = 275;
float[] h_A;
float[] h_B;
float[] h_C;
float[] h_C_ref;
CudaDeviceVariable<float> d_A;
CudaDeviceVariable<float> d_B;
CudaDeviceVariable<float> d_C;
float alpha = 1.0f;
float beta = 0.0f;
int n2 = N * N;
int i;
float error_norm;
float ref_norm;
float diff;
CudaBlas handle;
/* Initialize CUBLAS */
Console.WriteLine("simpleCUBLAS test running.");
handle = new CudaBlas();
/* Allocate host memory for the matrices */
h_A = new float[n2];
h_B = new float[n2];
//h_C = new float[n2];
h_C_ref = new float[n2];
Random rand = new Random(0);
/* Fill the matrices with test data */
for (i = 0; i < n2; i++)
{
h_A[i] = (float)rand.NextDouble();
h_B[i] = (float)rand.NextDouble();
//h_C[i] = (float)rand.NextDouble();
}
/* Allocate device memory for the matrices */
d_A = new CudaDeviceVariable<float>(n2);
d_B = new CudaDeviceVariable<float>(n2);
d_C = new CudaDeviceVariable<float>(n2);
/* Initialize the device matrices with the host matrices */
d_A.CopyToDevice(h_A);
d_B.CopyToDevice(h_B);
//d_C.CopyToDevice(h_C);
/* Performs operation using plain C code */
simple_sgemm(N, alpha, h_A, h_B, beta, h_C_ref);
/* Performs operation using cublas */
handle.Gemm(Operation.NonTranspose, Operation.NonTranspose, N, N, N, alpha, d_A, N, d_B, N, beta, d_C, N);
/* Allocate host memory for reading back the result from device memory */
h_C = d_C;
/* Check result against reference */
error_norm = 0;
ref_norm = 0;
for (i = 0; i < n2; ++i)
{
diff = h_C_ref[i] - h_C[i];
error_norm += diff * diff;
ref_norm += h_C_ref[i] * h_C_ref[i];
}
error_norm = (float)Math.Sqrt((double)error_norm);
ref_norm = (float)Math.Sqrt((double)ref_norm);
if (Math.Abs(ref_norm) < 1e-7)
{
Console.WriteLine("!!!! reference norm is 0");
return;
}
/* Memory clean up */
d_A.Dispose();
d_B.Dispose();
d_C.Dispose();
/* Shutdown */
handle.Dispose();
if (error_norm / ref_norm < 1e-6f)
{
Console.WriteLine("simpleCUBLAS test passed.");
return;
}
else
{
Console.WriteLine("simpleCUBLAS test failed.");
return;
}
}