Comments on: Updated DragonFly multiprocessing benchmarks

By: Justin Sherrill

Justin Sherrill — Wed, 24 Oct 2018 15:15:14 +0000

I agree; I’d like to see same test, same hardware, different OS – but I’m not volunteering. I don’t have the hardware or time for it.

This does answer a more basic question of: is DragonFly handling multiprocessing on new hardware well? I am happy to see that.

By: Eric Olson

Eric Olson — Wed, 24 Oct 2018 06:43:54 +0000

The scaling looks good; however, scaling doesn’t tell whether things are fast or slow.

For that a direct comparison with Linux running the same test on the same hardware is important, because Linux development has been performance centered and carefully checked to avoid performance regressions over the years.

By: Justin Sherrill

Justin Sherrill — Tue, 23 Oct 2018 03:44:54 +0000

A benchmark can be an exploration of performance. In this case, it’s “how fully is DragonFly taking advantage of multiple CPUs?” I’d like to see comparative numbers too.

By: Anonymous

Anonymous — Tue, 23 Oct 2018 03:25:29 +0000

I’m super excited to see an updated benchmark after like 5 years.

What’s weird though is that the new benchmark doesn’t compare perf to either previous versions of Dfly nor other OSes.

It’s hard to know if this is a good or bad perf when no comparison is made.

By: Anonymous

Anonymous — Mon, 22 Oct 2018 18:48:39 +0000

Very, very impressive!