[TEST/REVIEW] cpu time accounting patch, step 2

[Julian Elischer]

Andrew Gallatin wrote:

>Poul-Henning Kamp writes:
> > In message <17388.44976.250463.383429 at grasshopper.cs.duke.edu>, Andrew Gallatin
> >  writes:
> > 
> > >Here are some benchmark results from my "AMD Athlon(tm) 64 X2 Dual
> > >Core Processor 3800+" running an SMP kernel for both loopback and
> > >10GbE TCP networks.  The executive summary is that there's a 41%
> > >improvement in loopback pingpong, and a 12.5% improvement in a
> > >pingpong test with a remote linux machine over 10GbE.  Since the linux
> > >machine isn't infinately fast, the improvement is probably a bit more
> > >than 12.5% ;)
> > 
> > That's a LOT more than I had hoped for...
>
>Unfortunately, even after your patch, we are still about 38% slower
>than linux x86_64 on the same box for loopback ping-pong, and 32%
>slower for ping-pong over 10GbE.  (bandwidth is lower for streaming
>tests, and CPU utilization is much, much much higher in FreeBSD as
>well).
>
>I think you nailed the biggest source of overhead, but there is
>apparently a lot more performance that we can get out of the hardware.
>I'd love to see you commit this.
>  
>
robert watson has a change that apparently helps with loopback times.

>Drew
>
>x loopback_rr.after
>+ loopback_rr.rhel4
>+--------------------------------------------------------------------------+
>|   x                                                                   +  |
>|x  x  xx                                                          +   ++ +|
>| |_MA__|                                                            |_AM_||
>+--------------------------------------------------------------------------+
>    N           Min           Max        Median           Avg        Stddev
>x   5      38590.71      40167.44      39301.22      39463.23      645.1375
>+   5      53349.28      54884.98      54388.49     54274.174     566.90395
>Difference at 95.0% confidence
>        14810.9 +/- 885.686
>        37.531% +/- 2.24433%
>        (Student's t, pooled s = 607.282)
>
>
>x 10GbE_rr.after
>+ 10GbE_rr.rhel4
>+--------------------------------------------------------------------------+
>|  x                                                                    +  |
>|xxx                                                            +   +   + +|
>| AM                                                              |___A_M_||
>+--------------------------------------------------------------------------+
>    N           Min           Max        Median           Avg        Stddev
>x   6      33323.05      33703.95      33614.14     33556.268     137.33838
>+   5       43448.3      44966.41      44610.41     44331.738     606.53307
>Difference at 95.0% confidence
>        10775.5 +/- 571.321
>        32.1116% +/- 1.70258%
>        (Student's t, pooled s = 417.112)
>_______________________________________________
>freebsd-current at freebsd.org mailing list
>http://lists.freebsd.org/mailman/listinfo/freebsd-current
>To unsubscribe, send any mail to "freebsd-current-unsubscribe at freebsd.org"
>  
>