Re: Benchmarks: Graphics vs. batch, Linux vs. Windows

[donalson]

Thanks Marcus,

  I think I may have bitten off more than I can (or want to chew).
Comparing between Swarm and something else if turning out to be really
tricky if I want to be fair to all concerned.  This test is driven both by
my dissertation (which has a section on prepackaged simulators vs.
building your own sim from scratch) and my present modeling task which is
a model of environmental and predation effects on mussel beds.  In the
latter case the PI's want a C++/Windows based interface so I have learned
how to create Dialog Boxes, child windows and threads etc. etc. etc. in
Windows over the last month along with building a specialized list object
again etc. etc. etc.

The general argument is that the generelist (SWARM) by default will be
slower that the specialist (a model with specialized lists, schedule,
etc.)  With sim's that can run a month or more, if overhead can be cut by
25% (for example) this is actually non-trivial.  As I understand it, in
many of the hightech computer games critical sections are still written in
assembly.

The problem is that it is becomming hard to figure out how to be "fair".
Just an example, Swarm now updates the ModelSwarm probe with the new
values of dynamic variables.  This is overhead which I must be accounted
for when comparing.  Now I know about this, so I can just remove all the
probe variables and use default values, but there may be other things that
I have missed.  The further I go the more I wonder whether I am just
jumping into quicksand...

Well, I'll continue on it a bit longer and report whatever I find out with
the proviso that the results might not hold up on appeal.

Cheers,

   D3

"Marcus G. Daniels" wrote:

> >>>>> "D3" == donalson  <address@hidden> writes:
>
> D3> I have been trying to find a way in Windows to get the system time
> D3> (not swarm time) at the beginning and end of a run so that I can
> D3> subtract them to get "real" elapsed time.
>
> $ time program
>
> Should do it...where "program" is your simulation.
>
> `time' will report real time on both NT and Windows 98.  NT will also
> report user (scheduled time for the simulation itself) and system
> time spent in the system on behalf the the simulation (e.g. I/O).
>
>                   ==================================
>    Swarm-Support is for discussion of the technical details of the day
>    to day usage of Swarm.  For list administration needs (esp.
>    [un]subscribing), please send a message to <address@hidden>
>    with "help" in the body of the message.

--
*********************************************************************
* Doug Donalson                 Office: (805) 893-2962
* Ecology, Evolution,           Home:   (805) 961-4447
* and Marine Biology            email address@hidden
* UC Santa Barbara
* Santa Barbara Ca. 93106
*********************************************************************
*
*   The most exciting phrase to hear in science, the one that
*   heralds new discoveries, is not "EUREKA" (I have found it) but
*   "That's funny ...?"
*
*       Isaac Asimov
*
*********************************************************************



                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.