rec.autos.simulators

Something came flashing in my mind while discussing benchmarking issues at
a client. All benchmarks have to run on big volumes because of the CPU's
(x86) inability in terms of time precision to go higher than 1/16th of a
second. So here we are fighting for hundreds, sometimes thousands of a
second for a hotlap and all that for a 'fake' time??? Is there something I
am missing, some kind of trick used by the developers or is it all random
past the 1/16 limit?
Anybody could maybe enlight me on this?

**********************************************************
* Jean-Francois De Rudder    Tel   : +27 21 4196005      *
* Sybase SA                  Fax   : +27 21 4196009      *

**********************************************************

Hell, try to run with no details and almost no texturing at all in GP2
'to ensure that the frame rate does not fall too much with traffic'
and run a leisurely lap of 1.38 around Monza. Check your watch, which
you used to check the time. See it report 1.33. So the time runs faster
in GP2 if details are low enough. Put on full details and textures,
and the same 1.38 lap takes over three minutes to perform.

That is how much of a fake time it is. And how can you trust anyone
has driven their hot lap in 'real time' ???? You have _all_ the reaction
time you want (not that sure it helps much, but it must be better than
the game running faster than RL).

So, don't worry about thousandths when you can worry about ***y
MINUTES!

And of course, there is no possibility of adjusting the details so that
the time is closest to real time, because with cars on track, the details
and processor load will change . . . *aargh*

--
'I have something to say!   |            JJ Karhu
 It's better to burn out,   |     'The Imm***Immortal'  \o
   than to fade away!'      |    -=========================|xxxxxxo

Jason
 ---
Check out the NFS World Records Site
http://www.twogoons.com/nfs/

As you noted, most sims only calculate 10-30 frames per second. How can it
be more accurate than that. Well, if you know the position of the car on
the frame prior to crossing the start/finish line, and you the know the
position of the car on the frame after crossing the line, you can
interpolate between the two frames to get a pretty close approximation of
when the car actually crossed the line.

However, I have noticed that NASCAR and ICR tend to be pretty "chunky" with
lap times. I've run two identical lap speeds, down to the thousandths of a
MPH back to back, a feat I doubt I could ever reproduce in real life.

Dave (davids) Sparks
Late Night League
http://www.sequoia-dev.com/Hawaii/latenite.html

Dave Sparks wrote

Wow,

I doubt I could drive even ONE lap in real life without spinning,
crashing
burning and dying.  Or at least stalling the damn car on the way out of
the pits.  <g>

Jet

--
Chuck Stuart - Mesquite TX USA

Mark
_________________________________

"..every lap I think `Oh bollocks!
This is dreadful..`" Damon Hill
__________________________________

Dave (davids) Sparks
Late Night League
http://www.sequoia-dev.com/Hawaii/latenite.html

Yep this makes sense - To work out a precise lap time all you need it the
precise detail of the track and the line being driven, combined with a
knowledge of the speed the car in travelling at any moment in time.  I
don't think that any precise timeing ability is required.

 ___________________
/                   \
|   Jimmy Furlong   |        Aerodynamics are for people
|                   |         who cannot build engines  
|  Dublin, Ireland  |
\___________________/            - Enzo Ferrari

No he hasn't, he just hasn't strengthened it....

                     _________________________________
         __    _____|                                 |_____    __
________|  |__|    :|           John Wallace          |     |__|  |________

  \    :|  |::|    :|        Team WW Racing TSW       |     |::|  |     /
    >  :|  |::|    :|_________________________________|     |::|  |   <
  /    :|__|::|____/       * Sim Racing News *         \____|::|__|     \
/______:/  \::/ http://sneezy.dcn.ed.ac.uk/simnews/index.htm \::/  \._____\
               http://www.math.ohio-state.edu/~harmon/simnews

If the contrary, and view others hotlap, first you'll find even with
the correct graphic details on your own PC, that the hotlap will be
choppy and with a stopwatch the innaccuracy of session time might be as
high as 4-7 sec off real time.

As far as the performance editor currently available, the drivers power
can't be changed  + the best thing to validate a hotlap would be sending
the PERDATA and setup and try to match speed at least on a mainstraight
and long fast corner, for the rest, a setup from others is good only
for reference since driving style (with same steering device) are
different.

See ya later, Eric

: > Something came flashing in my mind while discussing benchmarking issues at
: > a client. All benchmarks have to run on big volumes because of the CPU's
: > (x86) inability in terms of time precision to go higher than 1/16th of a
: > second. So here we are fighting for hundreds, sometimes thousands of a
: > second for a hotlap and all that for a 'fake' time??? Is there something I
: > am missing, some kind of trick used by the developers or is it all random
: > past the 1/16 limit?
: > Anybody could maybe enlight me on this?

the technical explantion for this is that the timer (PIT) is
reprogrammed by the game to run at a higher speed.  normally it runs at
16.8 or so ticks per second (hence your 1/16th), but it can be programmed
to run way too fast.  

the actual fixed speed of the timer chip is 1,193,181 Hz.

the reason this doesnt affect your computer's
ability to keep time (ie dos clock time) is that the game only calls the
function to update the clock time every so many cycles to keep it at
~16.8.  if the game didnt do this, your clock would run in fast forward.  

early versions of games (ie heretic) didnt put the timer back to normal
when dumping to dos, and hence players had accelerated clocks are playing
the game.. pretty funny.

well this post didnt turn out so clear... so if u dont understand, ask.
-carlos

I'm not getting into a math argument because I would be completely
unarmed but let me give a different answer to the original question. The
TimeGetTime API function returns a long interger containing the number
of ms since Windows was started. Using this function, you can easily
measure events to a thousand of a second.

--
Chuck Stuart - Mesquite TX USA  
VBTrace 96 is a runtime Procedure Trace, XRef,
Debug, Profiler and much more. Shareware $49.
http://www.apexsc.com/vb/ftp/coop/cstuart

           Juergen.
\       /
\\     //
\\\ U ///
==ELITE==  Juergen Athen   http://iraul1.ira.uka.de/~s_athen/ja.html
 \\\-///   jAmeson on IRC, NightFall
   =^=
    ^      TSC Astoria Karlsruhe