rec.autos.simulators

I was wondering if anyone knows the story of what's going on with the
problem with running the DMA version of N2 with the Reactor (or other
Rendition board) with a Cyrix.  I have a P166+ on a Tyan TC I, and I get
the "jerky" controls behavior with N2 (and just pixel dust with ICR2)
that others have mentioned.

I heard (from this NG) that Intergraph was going to see Tyan about
something related to this(?), but the problem is with the Cyrix, not
just the Tyan MB, right?

Anyway, if anyone knows what's going on in with this issue, or if they
have gotten the DMA to work with Cyrix/Tyan, please pass it on.  Is it
Papy's "responsibility", or Intergraph's (and other video), or Cyrix's,
or am I never to get beyond running "n2rend" or "indysafe"?  Any patches
possible?

One last thing - I remember someone writing that with Rendition board,
one should not use "auto" on graphics options.  Is this true, and if so
why?  Is the "switching" on/off of an option going to slow things, or is
everything going to run slower even when no switching is occurring?

Thanks!

Here's the problem as best as I understand it:
A joystick is a "polled" device (you reset it, and spin in a loop
doing nothing but counting until the game card reports the capacitor
is chargerd; the longer your CPU sat there spinning, the more
resistance the joystick had and the higher the "count" (and from that,
you get the position))

When you put DMA into the mix, the CPU doesn't get as predictable
access to the game card (because DMA is taking control of the bus some
of the time), and the counts get a little off. The solution is to use
the Pentium instruction counter if available. This doesn't get skewed
by DMA like the loop count does. So, we query the CPU at the start of
the program, and if it says it's a "Pentium", we can use the accurate
joystick routines, and if it doesn't say "Pentium", we have to fall
back to the less accurate routines.

Your Cyrix chip doesn't report "Pentium", so we use the less accurate
routines. If you aren't using DMA, the less accurate routines are
"accurate enough" and everything is fine (except for your framerate).
If you are using DMA, framerate is great, but the joystick returns
unacceptable values...

---Jim Sokoloff, Papyrus

33/3 is 11MHz, which will cause the ISA bus transactions to take less
time. (And if the hardware can support it without glitches, it will
leave more time for the CPU to have control of the bus and do its
work.)

On your machine in N2, we should be using the RDTSC Pentium
instruction, so I wouldn't think it should affect the calibration
range reported...

Bottom line, if the 11MHz ISA bus works fine with your hardware and
you like how your games "feel" more, use that... :-)

---Jim Sokoloff, Papyrus

Thanks,