Pardiso Schur Complement Memory Allocation

Raffael_C_ — Wed, 23 Mar 2016 18:52:45 GMT

Hello,

I was just looking at the Schur Complement functionality of Pardiso and was studying the documentation for quite some time but there are still two things that I don't understand:

- How much memory should I allocate for the vector x when iparm(36)=1 or 2? (Obviously it can be bigger than n (the number of equations).

- What is the format in which the matrix is returned. It seems to be row-major dense format?

It would be very helpful if the documentation was more clear about this.

Thank you for your help,

Raffael

HI Raffael,

Alexander_K_Intel2 — Thu, 24 Mar 2016 04:25:37 GMT

HI Raffael,

- How much memory should I allocate for the vector x when iparm(36)=1 or 2? (Obviously it can be bigger than n (the number of equations).

[akalinki] Schur complement is the dense matrix K by K where K is number of unit element in perm array, so you need to provide array with size K*K*sizeof(data)

- What is the format in which the matrix is returned. It seems to be row-major dense format?

[akalinki] correct

Thanks,

Alex

topic Pardiso Schur Complement Memory Allocation in Intel® oneAPI Math Kernel Library

Pardiso Schur Complement Memory Allocation

HI Raffael,