public void Test_BLAS2_GBMV()
{
ClearBuffer(hiMatrixA);
ClearBuffer(hiMatrixACBC);
ClearBuffer(hiVectorXM);
ClearBuffer(hiVectorXN);
ClearBuffer(hiVectorYM);
ClearBuffer(hiVectorYN);
CreateBandedMatrix(hiMatrixA, M, N, KL, KU);
CompressBandedMatrixToCBC(hiMatrixA, hiMatrixACBC, M, N, KL, KU);
FillBuffer(hiVectorXM);
FillBuffer(hiVectorXN);
FillBuffer(hiVectorYM);
FillBuffer(hiVectorYN);
diMatrixA = _gpu.CopyToDevice(hiMatrixACBC);
diVectorXM = _gpu.CopyToDevice(hiVectorXM);
diVectorXN = _gpu.CopyToDevice(hiVectorXN);
// Test without transpose
diVectorYM = _gpu.CopyToDevice(hiVectorYM);
_blas.GBMV(M, N, KL, KU, Alpha, diMatrixA, diVectorXN, Beta, diVectorYM);
_gpu.CopyFromDevice(diVectorYM, gpuResultM);
for (int i = 0; i < M; i++)
{
double cpuResult = 0.0;
for (int j = 0; j < N; j++)
{
cpuResult += Alpha * hiMatrixA[GetIndexColumnMajor(i, j, M)] * hiVectorXN[j];
}
cpuResult += Beta * hiVectorYM[i];
Assert.AreEqual(cpuResult, gpuResultM[i]);
}
// Test with transpose
diVectorYN = _gpu.CopyToDevice(hiVectorYN);
_blas.GBMV(M, N, KL, KU, Alpha, diMatrixA, diVectorXM, Beta, diVectorYN, cublasOperation.T);
_gpu.CopyFromDevice(diVectorYN, gpuResultN);
for (int j = 0; j < N; j++)
{
double cpuResult = 0.0;
for (int i = 0; i < M; i++)
{
cpuResult += Alpha * hiMatrixA[GetIndexColumnMajor(i, j, M)] * hiVectorXM[i];
}
cpuResult += Beta * hiVectorYN[j];
Assert.AreEqual(cpuResult, gpuResultN[j]);
}
_gpu.FreeAll();
}
