Intel® Math Kernel Library

Odd dgesv results [Now with pardiso]


I'm trying to refactor a piece of code that solves a simple Ax=b linear problem using MKL. From the manual, I saw that dgesv is the function I should call.

A is an n-by-n Jacobian matrix (with just 5 non-zero diagonals), while b and x are arrays of size n.

To solve the problem I use this snippet of code

Using MPI parMETIS with cluster_sparse_solver


I am optimizing the `cluster_sparse_solver` runtime. In my case, the majority of the runtime is taken by phase `11`, symbolic factorization, with METIS. Additionally, only a single node is used in an otherwise `MPI`-enabled application.

I was wondering if there is a way to use `parMETIS` for fill-reducing ordering, in order to benefit from the cluster environment. One thing that would help tremendously is the source code for `cluster_sparse_solver`.

The version of MKL in question is mkl 11.2u3, which was bundled with composer_xe 2015 3.187.


DftiCommitDescriptor hangs my application


I've been struggling with this for a while, hope you can help.

I'm creating a shared library on Linux (Centos 6.7) using FFT's implemented with the MKL. My library works fine when used together with my own test program, but fails when used from a 3rd party application.

The problem has been pinned down to the FFT's. Print-output arrives on screen after calling DftiCreateDescriptor and DftiSetValue, and the status is 0. But when I subsequently call DftiCommitDescriptor the 3rd party application keeps running at 100% cpu and doesn't show any print-output anymore.

Intel® Math Kernel Library abonnieren