How can the C++ Eigen library perform better than specialized vendor libraries?

Question

I was looking over the performance benchmarks: http://eigen.tuxfamily.org/index.php?title=Benchmark

I could not help but notice that eigen appears to consistently outper

Answer 1

About the comparison ATLAS vs. Eigen

Have a look at this thread on the Eigen mailing list starting here:

• http://listengine.tuxfamily.org/lists.tuxfamily.org/eigen/2012/07/msg00052.html

It shows for instance that ATLAS outperforms Eigen on the matrix-matrix product by 46%:

• http://listengine.tuxfamily.org/lists.tuxfamily.org/eigen/2012/07/msg00062.html

More benchmarks results and details on how the benchmarks were done can be found here:

• Intel(R) Core(TM) i5-3470 CPU @ 3.20GHz:

  http://www.mathematik.uni-ulm.de/~lehn/bench_FLENS/index.html

• http://sourceforge.net/tracker/index.php?func=detail&aid=3540928&group_id=23725&atid=379483

Edit:

For my lecture "Software Basics for High Performance Computing" I created a little framework called ulmBLAS. It contains the ATLAS benchmark suite and students could implement their own matrix-matrix product based on the BLIS papers. You can have a look at the final benchmarks which also measure Eigen:

• http://apfel.mathematik.uni-ulm.de/~lehn/sghpc/gemm/page14/index.html#toc5

You can use the ulmBLAS framework to make your own benchmarks.

Also have a look at

• Matrix-Matrix Product Experiments with uBLAS
• Matrix-Matrix Product Experiments with BLAZE