nice configure

[Joshua Root]

Joshua Root wrote:
> Jordan K. Hubbard wrote:
>> On Jan 10, 2009, at 11:22 AM, Joshua Root wrote:
>>
>>> I just tested on 10.5.6 by running 'nice nbench', and both Activity
>>> Monitor and top show it using ~99% CPU...
>> A processes' nice factor can generally only be seen to produce results
>> when there is enough contention for CPU resources that it *needs* to be
>> "nice".  If there's little to no contention for CPU, however, then the
>> nice'd process is still going to get pretty much whatever it asks for.
>>
>> A much more realistic test would be to run n CPU-hungry processes, where
>> n is > the number of cores you have, and then nice one of them.
> 
> Are you saying that the problem only shows up when there is scheduling
> contention? I realise that it's normal for this example to use all the
> CPU despite being niced. It seemed like Adam and William were saying
> that any nice process will show 0% CPU usage.

Running 5 nbench instances, one of which is niced, on a 4-core machine:

 PID COMMAND      %CPU   TIME   #TH #PRTS #MREGS RPRVT  RSHRD  RSIZE  VSIZE
70605 top          8.3%  0:01.74   1    18     29  560K   200K  1156K
 18M
70604 nbench      58.9%  0:11.67   1    13     16  140K   232K   376K
 20M
70602 nbench      78.8%  0:20.36   1    13     16  140K   232K   376K
 20M
70600 nbench      81.9%  0:23.64   1    13     16  140K   232K   376K
 20M
70599 nbench      79.6%  0:26.08   1    13     16  140K   232K   376K
 20M
70598 nbench      79.8%  0:28.34   1    13     15   84K   232K   328K
 20M

Seems to be exactly as expected, and Activity Monitor shows essentially
the same numbers (they fluctuate a little of course).

- Josh