Add square matrix scaling graph

[llewelld]

Adds a graph to the benchmarking README.md to show the time taken to multiply square matrices of increasing size.

Results CSV file, graph PNG and graph SVG file are included.

[mhauru]

LGTM

[mhauru]

The performance of C naive with 10 threads is really interesting. Why does it lose some of its edge against single threaded when matrices get bigger?

Nothing to do with whether to merge this though.

[llewelld]

The performance of C naive with 10 threads is really interesting. Why does it lose some of its edge against single threaded when matrices get bigger?

Thanks for the review. I'll merge in, but I'd also be interested to know the answer to your question.

In practice, it's losing its edge only in comparison with the transposed version, no? At the top end it's consistently faster than the non-transposed versions. Or am I misreading?

In this case, maybe it's a sign that the different memory cache levels are being exhausted?

Having reviewed these two, would you be up for taking a look at #40 as well? No obligation, but it would be great to get your feedback on it.

[llewelld]

Rebased.