On 24/04/15 09:23, Martin Lucina wrote:
Thanks to Andrew and Antti for prompting me to do these tests. I'll add "write some reproducible benchmarking scripts" to my growing TODO list, it will be useful to track these results over time.
Actually, I wanted to reply to this point more specifically already earlier, but thinking about Maud and science I forgot everything else.
Yes, 100% I agree it would be useful to track the results over time. I'd be happy just being able to look at the progression of a number and use some old-fashioned human intellect to filter out disinteresting runs (e.g. ones where I can suspect that the host was overloaded). The numbers would both show that performance is going in the right direction, and also quickly catch bugs such as the recent one of all threads being always runnable.
Who knows, someone might even be able to generate some science out of those numbers.
