[RC5] Re: Needing testers (Mac OS X/G5)

Elektron elektron_rc5 at yahoo.ca
Mon Aug 9 13:58:13 EDT 2004


> Yep, it was busy with another d.net client, thats why benchmarks was 
> so low...
>
> Macs make me #2 at everyday top100. So I hope new version gives me 
> more stubs....
>
> When I stopped main client, I recieve next numbers from benchmark - 
> 23M, 30M, and 22M.

The only accurate benchmarks are where you boot to single-user (where 
the web interface asked me if the stats were accurate).

Sadly, "time" is not very accurate, either. With a lot of processes 
running, the CPU time used by an app is reported lower than it should 
be, so keys per usertime is too large.

> It has nothing to do with hyperthreading or something similar on the 
> G5. The "problem" is simply that Mac OS X' lowest priority level is 
> not the same as e.g. Linux' idle priority. There is no priority level 
> on Mac OS X where a process will only get the idle cycles. So even if 
> you run dnetc at priority 19 under Mac OS X, it'll slow down all other 
> applications somewhat (though other applications will get a larger 
> percentage of the available cycles).

FreeBSD did scheduling properly. Mac OS X's scheduler is icky, and 
though it supports nice 20, still manages to screw up scheduling (and 
top doesn't show the priority). When more processes are competing, it 
seems to do it better.

   UID   PID  PPID CPU PRI NI      VSZ    RSS WCHAN  STAT  TT       TIME 
COMMAND
   501  5279  7176   0   0 20    27320    372 -      RN    p2    0:50.49 
lag
   501  5280  7176   0   0  0    27320    372 -      R+    p2    1:03.75 
lag

   PID COMMAND      %CPU   TIME   #TH #PRTS #MREGS RPRVT  RSHRD  RSIZE  
VSIZE
  5280 lag         52.7%  1:50.92   1    11    21   104K   392K   372K  
26.7M
  5279 lag         39.2%  1:24.43   1    11    21   104K   392K   372K  
26.7M

   PID COMMAND      %CPU   TIME   #TH #PRTS #MREGS RPRVT  RSHRD  RSIZE  
VSIZE
  5293 lag         25.4%  0:04.45   1    11    20   104K   392K   372K  
26.7M
  5291 lag         25.3%  0:04.45   1    11    20   104K   392K   372K  
26.7M
  5292 lag         25.2%  0:04.39   1    11    20   104K   392K   372K  
26.7M
  5279 lag         16.4%  1:54.28   1    11    21   104K   392K   372K  
26.7M

I'm also not sure why a a one-liner allocates 26 megs of VM (OSX's 
memory management is horrible). If I remove redundant line feeds, 
lag.c:

int main() { do ; while (1); }

- Purr



More information about the rc5 mailing list