SCIENCE
GPU Support Included in Latest libflame Release
The fifth milestone release (r4648) of libflame announced. It is a modern replacement for the most-used functionality of the LAPACK linear algebra library. libflame is provided as free software licensed under the GNU Lesser General Public License (LGPL).
The main improvement since version 4.0 is that libflame now supports parallel execution using multiple GPUs through the SuperMatrix runtime system. By linking libflame with CUBLAS for the execution of BLAS routines on a single GPU, the SuperMatrix runtime system schedules operations to each GPU and manages the explicit movement of data. This release includes support for single and double precision real and complex floating point operations. A comprehensive table of dense linear algebra operations supported within libflame along with the ones with GPU support may be found at http://z.cs.utexas.edu/wiki/flame.wiki/libflame/.
An in-depth description of the effort required to support parallel execution using multiple GPUs and a first introduction to libflame, as well as a comprehensive user's guide, may be found at http://www.cs.utexas.edu/users/flame/publications/. For general information, including links to source code distributions, please visit the FLAME project home page at http://www.cs.utexas.edu/users/flame/.