According to the definition used the arithmetic intensity is measured by operations per byte. This might not be adequat for haswell processors (and later). Due to the fused multiply-add\footnote{although called multiply-add there are 36 different slightly instructions} extension two floating point operations can be performed with a single instruction. 

- worse results for 4 threads @ NUMA-STREAM not necessarily expected
- better results for triad possibly due to combined storage in FMA
- striding for arrays

%%% Local Variables:
%%% mode: latex
%%% TeX-master: "../report"
%%% End: