Re: [Flent-users] irtt sleep test

From: Pete Heist <peteheist@gmail.com>
To: Dave Taht <dave.taht@gmail.com>
Cc: flent-users <flent-users@flent.org>
Subject: Re: [Flent-users] irtt sleep test
Date: Thu, 7 Dec 2017 09:36:40 +0100	[thread overview]
Message-ID: <83DA13E1-75DE-45E8-83C1-053AC4D3D701@gmail.com> (raw)
In-Reply-To: <CAA93jw6KQP-FFgdL8g5U1EYaD2GFpuPvVd9G9cj+m3mU5xh+BA@mail.gmail.com>

Thanks for the numbers. I wondered if irtt would run on that Archer.

1ms seems to be a kind of timing purgatory, where above that things look good, below that are the demon-things and right at 1ms it’s a mixture. There’s usually some discontinuity in percent error around 1us or 100ns like in your numbers. Below some point I guess you’re just testing cpu / system call overhead. I was going to remove those lower numbers then figured it’s fun to look at anyway.

Macs don’t seem to do as well as Linux at 1ms and above, which is what spawned the timer compensation stuff. The ‘hybrid’ timer can be useful still down to 100us or so, and the ‘busy’ timer is currently needed for anything below that. Looks like timerfds could be good down to 5us or so, without a modified kernel:

https://stackoverflow.com/questions/19927225/scheduling-events-at-microsecond-granularity-in-posix

OS/X 10.11.6, Core i7-2860QM (quad core late 2011):

tron:~/src/github.com/peteheist/irtt:% ./irtt sleep
Testing sleep accuracy...

Sleep Duration        Mean Error       % Error
           1ns             381ns       38194.1
          10ns             370ns        3703.5
         100ns             286ns         286.6
           1µs           9.041µs         904.2
          10µs          30.538µs         305.4
         100µs          60.746µs          60.7
           1ms         263.256µs          26.3
          10ms        2.000929ms          20.0
         100ms        3.309786ms           3.3
         200ms        4.183972ms           2.1
         500ms         2.53946ms           0.5

> On Dec 7, 2017, at 2:24 AM, Dave Taht <dave.taht@gmail.com> wrote:
> 
> good test. I wish I'd thought of it. Life below a ms is a PITA.
> 
> For comparison, an aarch64 and an x86_64:
> 
> root@nanopineo2:~# irtt sleep
> 
> Testing sleep accuracy...
> 
> Sleep Duration        Mean Error       % Error
> 
>           1ns          13.353µs     1335336.9
> 
>          10ns           14.34µs      143409.5
> 
>         100ns          13.343µs       13343.9
> 
>           1µs          12.791µs        1279.2
> 
>          10µs         148.661µs        1486.6
> 
>         100µs         150.907µs         150.9
> 
>           1ms         168.001µs          16.8
> 
>          10ms         131.235µs           1.3
> 
>         100ms         145.611µs           0.1
> 
>         200ms         162.917µs           0.1
> 
>         500ms         169.885µs           0.0
> 
> 
> d@nemesis:~$ irtt sleep
> 
> Testing sleep accuracy...
> 
> 
> Sleep Duration        Mean Error       % Error
> 
>           1ns             668ns       66831.9
> 
>          10ns             672ns        6723.7
> 
>         100ns             557ns         557.6
> 
>           1µs          57.749µs        5774.9
> 
>          10µs          63.063µs         630.6
> 
>         100µs          67.737µs          67.7
> 
>           1ms         153.978µs          15.4
> 
>          10ms         169.709µs           1.7
> 
>         100ms         186.685µs           0.2
> 
>         200ms         176.859µs           0.1
> 
>         500ms         177.271µs           0.0
> 
> 
> -- 
> 
> Dave Täht
> CEO, TekLibre, LLC
> http://www.teklibre.com
> Tel: 1-669-226-2619

_______________________________________________
Flent-users mailing list
Flent-users@flent.org
http://flent.org/mailman/listinfo/flent-users_flent.org