smeuletz (alexsmeu.delete@this.yahoo.com) on January 7, 2020 3:07 am wrote:
> And again, why not at least deprecate stuff that became meaningless 10 years ago so that
> in 20 years, when we use quantum computers we don't stumble on sched_yield again (joke).

IMHO sched_yield() is still a good interface for an application to tell the kernel:
I, as an application, am treating blocks of data in a bursty manner, and I have just finished one of such a burst of data.
I have been using quite a large block of memory cache for that block of data, but I will not need such content again because next time I will have another different block of data.
So right now, if you have something else to do on the current processor, that is the right time to start doing it, but please next time you schedule me you give me another "complete time slice" (or at least as much time as you gave me last time) - so I will run efficiently and will refill a big block of memory cache with my new data.
I.e. it would be inefficient for me to start filling a big block of memory cache right now and then be interrupted randomly by a timeslice limit, by another task which would wipe out my content in the memory cache.
Even quantum computers in 10 years might use memory cache...