AOCL-LAPACK delivers a high-performance implementation of the Linear Algebra PACKage (LAPACK). LAPACK offers routines for solving linear systems, least-squares problems, eigenvalue and singular-value problems, and the related matrix factorizations. The library is extensible, easy to use, and distributed under a modified BSD open source license. Applications that depend on the standard Netlib LAPACK interfaces can adopt AOCL-LAPACK with little or no change to their source. AOCL-LAPACK provides C and Fortran interfaces and a C++ template-based interface for a subset of the APIs.
AOCL-LAPACK is built on libFLAME, a project originally developed by members of the Science of High-Performance Computing (SHPC) https://shpc.oden.utexas.edu/ group at the Institute for Computational Engineering and Sciences https://www.oden.utexas.edu/ at The University of Texas at Austin https://www.utexas.edu/. The upstream libflame repository is available on GitHub (flame/libflame). AMD maintains an “amd” fork of libflame on its GitHub, where key routines are optimized for AMD “Zen”-based architectures as part of AOCL-LAPACK.
Starting with AOCL 5.1, AOCL-LAPACK conforms to the LAPACK 3.12.0 specification. When used together with AOCL-BLAS—which contains optimizations for AMD “Zen”-based processors, AOCL-LAPACK delivers high-performance LAPACK functionality on AMD platforms.
AOCL-LAPACK API Guide documentation is available at https://docs.amd.com/go/en-US/63860-AOCL-LAPACK