[optimux]

Hi guys,

I just run into a dense asymmetric rank-deficient system (see attached fig). This system comes from a Fourier-Chebyshev approximation of thermal convection in an annulus. Its condition number is as high as 1.0e14 and would increase if we use higher spatial resolution. The author uses dgetrf to do LU factorization and then dgetrs to solve a similar system, but they're not the same.

Any suggestions? Thanks.

PS: for example, a 514*514 matrix, its rank is 507, condition number is 9.0e14.

Best,
Mingming

[optimux]

Originally Posted by optimux

Hi guys,

I just run into a dense asymmetric rank-deficient system (see attached fig). This system comes from a Fourier-Chebyshev approximation of thermal convection in an annulus. Its condition number is as high as 1.0e14 and would increase if we use higher spatial resolution. The author uses dgetrf to do LU factorization and then dgetrs to solve a similar system, but they're not the same.

Any suggestions? Thanks.

PS: for example, a 514*514 matrix, its rank is 507, condition number is 9.0e14.

Best,
Mingming

Hello,

This dense matrix can be solved by LAPACK routines DGETRF and DGETRS, also MATLAB \ operator will work. So this post will end here.

Thanks
Mingming