Sparse Linear Algebra Functions in Intel® Math Kernel Library

  • Overview

Sparse matrix algorithms are encountered in a broad range of important scientific computing applications. Intel® Math Kernel Library (Intel® MKL) offers a powerful set of functions that can be used to build a complete solution to many sparse linear systems. This webinar gives an overview on Intel MKL’s sparse linear algebra component. Highlights include Sparse BLAS functions, Direct solvers for sparse linear systems, Iterative solvers and Eigensolvers for sparse matrices based on the FEAST algorithm.

View Slides

Benchmark results were obtained prior to the implementation of recent software patches and firmware updates intended to address exploits referred to as "Spectre" and "Meltdown". Implementation of these updates may make these results inapplicable to your device or system.

Software and workloads used in performance tests may have been optimized for performance only on Intel® microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information, see Performance Benchmark Test Disclosure.



Are the sparse linear solvers MPI or OpenMP based? namely, can it run on multiple nodes in a cluster like PETSc? Also, how does its performance compare with PETSc?

mad\dskapoor's picture

It has been a while since @Wadud M asked this question, but let me clarify for the benefit of others who might have the same question. The sparse linear solvers in Intel MKL library are multithreaded only. There is no direct MPI support. However, libraries like PETSc and Trilinos that do provide MPI support for sparse linear solvers, can be configured to use the Intel MKL library on a per node basis. This way the (multithreaded, single node) optimizations available in Intel MKL library can be used by MPI applications too.

Wadud, the Parallel Direct sparse solver for Clusters API has been introduced since Intel(R) MKL v.11.3. 

I would like to ask about RCI ISS in newest Intel MKL solver. Does it work well with EBE-PCG (element by element preconditioner conjugate gradient) method or can I build my code of EBE-PCG by using Intel MKL solver? 

Thank you very much!


Add a Comment

Have a technical question? Visit our forums. Have site or software product issues? Contact support.