Peter Peres writes:

        [...]

        I suppose that the only thing that might make a 70 usec delay
under
        Windows possible, would be a boot virus that adds itself to the
API
        somehow (in cooperation with a VxD that you write probably ).

        It is just barely possible that one could write an assembly VxD
that takes
        over complete control for 70 usec. I have never tried this.

This is perfectly possible, though I agree it's all a bit nasty under
Windows. There is an (undocumented ;-/ ) 'virtual Timer device' funtion
VTDdelay (or similar) which delays in units of ~840nSec. This is done by
monitoring the timer chip, which is reprogrammed under windows to a
faster rate than under DOS (The VTD driver then uses this to simulate
the timing behaviour that DOS etc. expects).

I did a bit of experimentation with the actual delay given by this, and
found, as you might expect, that with the VxD calling overhead you
couldn't just do a straight multiple of the 840nSec, at lower multiples
(< say 50uSec) at least. I used the Real Cycle counter on the Pentium to
calibrate the thing. I needed delays of ~25uSec and ~450uSec, and
achieved them.

        HTH
        jon Nicoll
--
Jon Nicoll   Symbionics Ltd.  Cowley Road  Cambridge CB4 4WS  UK
Tel. +44 1223 421025   Fax +44 1223 421031 mailto:jkn@symbionics.co.uk