On Thursday, 3 August 2017 at 14:00:31 UTC, Matt wrote:
Meanwhile, the blog post Laeeth gave you shows Mir doing
better on matrix multiplication benchmarks than Eigen,
significantly better when dealing with complex numbers.
I mean by now we should all be jaded enough not to simply take
toy benchmarks as gospel for which is actually fastest in a
non-trivial application.
That's why I didn't make such a general claim and noted that the
linked benchmarks only dealt with matrix multiplication. :P
I don't doubt mir is really fast, though.
Yes, the benchmarks are indicative, but it's up to you come up
with a benchmark that characterizes your workload better.
