Using Intel MKL can save development, debug and maintenance time in the long run because today's code will run optimally on future generations of Intel processors with minimal effort. Intel has engineered this ready-to-use, royalty-free library, to allow you to focus on and deliver features your customers have requested.
What’s New: Version 11.2 Features
The Cluster Parallel Direct Sparse Solver extends the capabilities of Intel MKL PARDISO, enabling users to solve large distributed sparse systems of equations on clusters. Benchmark results demonstrate up to 2x performance improvement over MUMPS*1
Small Matrix Multiply performance improvements deliver performance boosts of 1.3X on average for small problem sizes (less than 20x20).2
- Significant performance improvement for small matrices (for 4x4 to 20x20 matrices) over Intel® MKL 11.1.1 for S/C/ZDGEMM.
- Further performance improvement (up to 2x) over Intel® MKL 11.1.1 through reduced call/error checking overhead and new inline functions.