jeantil commented on PR #2814:
URL: https://github.com/apache/james-project/pull/2814#issuecomment-3305367985

   go ahaead as part of 
   > Other improvements from your side are welcome too (maybe in another PR).
   
   I might just try to disable all this forkcount tuning and see how it affects 
build times. 
   
   > So forkCount does indeed help more or less.
   Results on your machine have no guarantees to translate to results on the CI 
machines, parallelization is highly hardware dependant. Modern developper 
machines have insane amounts of resources compared to workers which most of the 
time work on virtualized CPUs and RAM which is shared with an unknowable number 
of other processes. While doing performance benchmarking before, I have seen 
context switching on the physical CPU use up to 50% of the CPU compute 
capacity. 
   
   Performance can only be measured and tuned in and for a specific context 
which is usually highly hardware dependant. 
   This is also why I mentioned that I would autorebase and dual build for at 
least a couple week : to obtain enough datapoints on a range of CI workers that 
a statistical analysis has some hope of giving results since we cannot control 
the worker composition and the shared usage on the hypoervisors they run on. 
   Alternatively we would need to be able to provide our own controlled compute 
resources in which case a couple runs would be enough but I don't have that 
kind of money :D


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to