In message <19990228230536.A433@marvin.ece.utexas.edu>,
"Brian C. Grayson" writes

>  I can send in a PR, but I am not qualified to actually do the
>changes -- I have no idea precisely where the inc and dec
>should be placed -- before or after setting the new int mask,
>etc.

Oh.  I thought the x86 memory-to-memory ops were atomic.  (Isn't that
how the memory-basd spl()s work?)  So provided you get them on the
right path, and declare the C variable as volatile, it doesn't matter
too much ;)

[multi-cpu support]

I dont know.  My point was more that this is not the right abstraction
to architect in at an MI level. But it's fine for uniprocessors with
atomic memory-to-memory inc/dec ops.

>> And somene should fix systat ps, so that instead of the braindead
>> load-average "histogram" it shows interrupt and system time. :)
>
>  Do you mean a la top (53% user, 20% system, 3% interrupt, 24%
>idle)?  _That's_ something I could handle doing myself!


Yes please :-/.  "systat ps" is pretty useless as a top replacement
without them.