(point being: if you do have a go at it, the more its easy to replicate your systems config etc, the better :) ) On Fri, Dec 21, 2018 at 9:23 PM Carter Schonwald wrote:

i suspect not :)

note well: this sort of stuff is super tricky to measure well / reproducibly.

I remember some discussions I had with folks a few years ago where I learned it was a bit tricky to replicate the configurations of various systems for the benchmarks!

On Fri, Dec 21, 2018 at 11:37 AM Andrew Martin wrote:

...

Are there any easy-to-build benchmarks for the GHC event manager? There's some stuff in https://github.com/tibbe/event, but that repo hasn't been touched in 9 years, and it looks like the benchmarks in it were not pulled into GHC itself. I've found some stuff for benchmarking the event manager against stdio (an alternative event manager than builds on top of libuv), but I wanted to check and see if there are any other benchmark suites out there.

-- -Andrew Thaddeus Martin _______________________________________________ Libraries mailing list Libraries@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries

Yeah, this can’t be emphasized enough: different setups and workloads will measure different things / possibly contradict if you don’t look at what they measure carefully. On Fri, Dec 21, 2018 at 11:11 PM Ben Gamari wrote:

On December 21, 2018 9:23:39 PM EST, Carter Schonwald < carter.schonwald@gmail.com> wrote:

...

i suspect not :)

note well: this sort of stuff is super tricky to measure well / reproducibly.

I also don't know of any benchmarks beyond the recent work on stdio. However, Carter is quite right in saying that this sort of thing is quite tricky to benchmark properly.

Cheers,

- Ben

-- Sent from my Android device with K-9 Mail. Please excuse my brevity.