/eigen-magma

A fork of Eigen 3.2 to use MAGMA (GPU & CPU) as backend in the same way it does with Intel MKL.

Primary LanguageC++OtherNOASSERTION

Eigen Magma

MAGMA is a solid implementation of BLAS and linear algebra routines for multicore and GPU systems: http://icl.cs.utk.edu/magma/index.html

Optional support of MAGMA in Eigen for large dense matrix computations would allow applications written using Eigen to profit from GPU and multi-GPU support of MAGMA.

This project is a fork of Eigen 3.2.0 to use MAGMA (tested with version 1.4.0) as backend. MAGMA was compiled using the Intel compiler and using MKL as part of Parallel Studio 2013. It also uses CUDA and therefore CUBLAS 5.5

Sample project and benchmarks

The following sister project offers a sample CMake client of Eigen MAGMA and benchmark results in Gflop/s for all functions implemented:
https://github.com/bravegag/eigen-magma-benchmark