Pardiso 4.1.0

[Petros_Mamales]

Hi,
Are there any plans to include pardiso 4.1 in mkl in the near future?
Thank you, P-

[Alexander_K_Intel2]

Hi,

Which feature from PARDISO 4.1 areyou interested in?

With best regards,

Alexander Kalinkin

[Petros_Mamales]

I was trying to build the ipopt optimization package, and thenew pardiso was said to include functionality which would be necessary for it. You can refer to:

O. Schenk, A. Wachter, and M. Hagemann. Matching-based preprocessing algorithms to the solution

of saddle-point problems in large-scale nonconvex interior-point optimization.

In the official IPOPT page there is mention of the MKL pardiso being not suitable for the build. And, related,in the pardiso page, there are ipopt binaries for sale.
Thank you.

[Petros_Mamales]

Coming back to my initial question, is there any plans for pardiso 4.1?
Thank you, in advance.

[Alexander_K_Intel2]

Hi,

Now Im investigating your issue trying to find a workaround. Nevertheless, there is a feature request to develop features from UB PARDISO 4.1 so Ive added the information from this topic into this tracker.

With best regards,

Alexander Kalinkin

[Peter_Hollinger]

Quoting Alexander Kalinkin (Intel)

Hi,

Which feature from PARDISO 4.1 areyou interested in?

With best regards,

Alexander Kalinkin

One feature I miss in MKL's PARDISO is an option to solve the transposed problem (equivalently, solve the normal problem for a matrix in CSC format rather than CSR). In PARDISO 4.1.2, that option is invoked by setting IPARM(12)=1.

Another useful option is to compute the determinant; in 4.1.2: IPARM(33)=1.

[Alexander_K_Intel2]

Hi,

We have feature request to implement support of solving system with transpose matrix so if you provide data about your company or university I could include it in this feature request. If you want to get determinant of symmetric indefinite matrix please look example namedpardiso_sym_getdiag_c in MKL example folder - there is way to get it.

With best regards,

Alexander Kalinkin