public void Test_BLAS2_TRMV()
{
ClearBuffer(hiVectorXN);
FillBuffer(hiVectorXN);
diVectorXN = _gpu.Allocate(hiVectorXN);
diMatrixA = _gpu.Allocate(hiMatrixANN);
// Lower triangle matrix
ClearBuffer(hiMatrixANN);
CreateBandedMatrix(hiMatrixANN, N, N, N - 1, 0);
_gpu.CopyToDevice(hiMatrixANN, diMatrixA);
// Test without transpose
_gpu.CopyToDevice(hiVectorXN, diVectorXN);
_blas.TRMV(N, diMatrixA, diVectorXN);
_gpu.CopyFromDevice(diVectorXN, gpuResultN);
for (int i = 0; i < N; i++)
{
double cpuResult = 0.0;
for (int j = 0; j < N; j++)
{
cpuResult += hiMatrixANN[GetIndexColumnMajor(i, j, N)] * hiVectorXN[j];
}
Assert.AreEqual(cpuResult, gpuResultN[i]);
}
// Test with transpose
_gpu.CopyToDevice(hiVectorXN, diVectorXN);
_blas.TRMV(N, diMatrixA, diVectorXN, cublasOperation.T);
_gpu.CopyFromDevice(diVectorXN, gpuResultN);
for (int j = 0; j < N; j++)
{
double cpuResult = 0.0;
for (int i = 0; i < N; i++)
{
cpuResult += hiMatrixANN[GetIndexColumnMajor(i, j, N)] * hiVectorXN[i];
}
Assert.AreEqual(cpuResult, gpuResultN[j]);
}
// Upper triangle matrix
ClearBuffer(hiMatrixANN);
CreateBandedMatrix(hiMatrixANN, N, N, 0, N - 1);
_gpu.CopyToDevice(hiMatrixANN, diMatrixA);
// Test without transpose
_gpu.CopyToDevice(hiVectorXN, diVectorXN);
_blas.TRMV(N, diMatrixA, diVectorXN, cublasOperation.N, cublasFillMode.Upper);
_gpu.CopyFromDevice(diVectorXN, gpuResultN);
for (int i = 0; i < N; i++)
{
double cpuResult = 0.0;
for (int j = 0; j < N; j++)
{
cpuResult += hiMatrixANN[GetIndexColumnMajor(i, j, N)] * hiVectorXN[j];
}
Assert.AreEqual(cpuResult, gpuResultN[i]);
}
// Test with transpose
_gpu.CopyToDevice(hiVectorXN, diVectorXN);
_blas.TRMV(N, diMatrixA, diVectorXN, cublasOperation.T, cublasFillMode.Upper);
_gpu.CopyFromDevice(diVectorXN, gpuResultN);
for (int j = 0; j < N; j++)
{
double cpuResult = 0.0;
for (int i = 0; i < N; i++)
{
cpuResult += hiMatrixANN[GetIndexColumnMajor(i, j, N)] * hiVectorXN[i];
}
Assert.AreEqual(cpuResult, gpuResultN[j]);
}
_gpu.FreeAll();
}
