Re: [ublas] Matrix multiplication performance

Subject: Re: [ublas] Matrix multiplication performance
From: Michael Lehn (michael.lehn_at_[hidden])
Date: 2016-02-16 20:19:38

On 16 Feb 2016, at 17:09, Michael Lehn <michael.lehn_at_[hidden]> wrote:

>>> 
>>> About Blaze:  Do they have their own implementation of a matrix-matrix product?  It seems to require a
>>> tuned BLAS implementation (“Otherwise you get only poor performance”) for the matrix-matrix product.
>> I will check the benchmarks I run. I think I was using MKL with Blaze, but Blaze is taking it a step further (I am not sure how) and they are getting better performance than the underlying GEMM. Their benchmarks indicate that they are faster than MKL (https://bitbucket.org/blaze-lib/blaze/wiki/Benchmarks#!row-major-matrixmatrix-multiplication)
> 
> I started today with similar experiments on BLAZE and had closer look at their internal implementation.  By default
> they are calling an external BLAS backend.  On my machine I used the Intel MKL.   But you are right, they also have
> an internal implementation that can be used if no external BLAS is available.  I will publish the results on this page:
> 
> 	http://www.mathematik.uni-ulm.de/~lehn/test_blaze/index.html
> 
> At the moment the benchmarks for the internal BLAZE implementation for the matrix-matrix product seem to look
> poor.  I asked Klaus Iglberger (the author of BLAZE) to check the compiler flags that I have used.  So don’t take the
> current results as-is.
> 

I modified the benchmark code in favour for BLAZE as Klaus Iglberger’s suggested.  The results are now slightly
better.  However, compared to Intel MKL they are way off.  Similar results I got on (my) Haswell.  The problem seems
as stated by Klaus Iglberger:  "The optimization flags are fine. The performance difference is due to the lack of a
specifically optimized kernel.”  So on other platforms it might perform much better.

-----------------------------------------------------------------------------------
Dr. Michael Lehn
University of Ulm, Institute for Numerical Mathematics
Helmholtzstr. 20
D-89069 Ulm, Germany
Phone: [phone], Fax: [phone]
-----------------------------------------------------------------------------------