> Just 2 changes from v1.06:
>
> o Changed the test of whether a process is fork()'ing "often".
> It's still not perfect I suppose, but I think it's an
> improvement in most situations.
>
> o Inherit p->run_q_sw in fork().
>
> ***
>
> Future plans for this: maybe add a syscall and a user-space util to allow
> fine-tuning of scheduling.
Hmmmm, surely a job for sysctl rather than YASC (yet another system call).
In reply to someone else's comments about sysctl a while back, the scheme
is indeed a mess as currently stands. The code for /proc based
modification and sysctl() call based modification should _really_
eventually hit a common code path.
Chris