Libraries

l1_logreg uses BLAS and LAPACK:

• Basic Linear Algebra Subprograms (BLAS)are routines which perform basic linear algebra operations such as vector and matrix multiplication. CBLAS is a C implementation (interface) of BLAS.
• The Linear Algebra PACKage (LAPACK), is a software library for numerical computing written in Fortran 77.

BLAS and LAPACK define interfaces, but the implementations are not necessarily optimized to a particular machine. In this case, you may speed up l1_logreg a lot by using optimized libraries like vendor specific BLAS/LAPACK (MKL, ACML, etc.) libraries or ALTAS.

• Automatically Tuned Linear Algebra Software (ATLAS) is a software library for linear algebra. It provides an open source implementation of BLAS APIs for C and Fortran77.
• Intel Math Kernel Library (MKL) and AMD core Math Library (ACML) are vendor specific, highly optimized math libraries for their own platform.

For best performance, we recommend you to use vendor specific libries, or ATLAS. Recent versions of MATLAB are also using vendor specific BLAS/LAPACK libraries internally.

• Intel MKL http://www.intel.com/cd/software/products/asmo-na/eng/perflib/mkl/index.htm
• AMD ACML http://developer.amd.com/acml.jsp
• ATLAS http://math-atlas.sourceforge.net