Re: Current ULE state and plans (was Re: HMP scheduling on FreeBSD)

On Jan 22, 2026, at 11:09 AM, Rodney W. Grimes <freebsd-rwg@gndrsh.dnsmgr.net> wrote:
> 
>> On 1/19/26 15:51, Olivier Certner wrote:
>>>> Why remove functioning code that people actually use?
>>> 
>>> How can you be sure it is functioning correctly?  Have you reviewed it?  Has it been tested on a variety of workloads?  What does guarantee it will not break, more or less badly, when some code surrounding it is changed, even when these changes are done carefully?
>>> 
>>> And who is using it, exactly? You count for 1. Given the past threads on the topic, I expect a few more, but can we even reach 10 or 20, I wouldn't bet on it.
>>> [...]
>> ...
>> Anecdotally, I don't personally observe the purported interactivity
>> improvement ascribed to ULE, either, but I can't give you numbers on it.
> 
> I actually find ULE to be BAD on interactive systems, often blocking my
> login shell for long enough that I think the system might be crashed
> when a heavy compute (make world for example) is running.

I have been meaning to ask since the start of this thread.....

Is there a set of tests that people can run to measure
important characteristics of schedulers under various
conditions (# of cores, multi-cpu, amoutn of memory,
interactivity, network traffic, NFS etc.)? The same can also
help with testing scheduler changes to see their effect on
these characteristics. And can help with trying out new
scheduling disciplines. And may be such testing can inspire
peopel who want the 4BSD to improve it as well! No synthetic
tests will sufficiently test real world conditions but at
least they can provide a baseline. Better still may be they
can be run against Linux as well!

Alternative question: what tests are currently used to test
ULE vs 4BSD schedulers and what can be done to improve them,
make them more comprehensive?

I hope such a testsuite would allow people to objectively see
which scheduler is better (under what conditions).

Thanks!
Bakul