🚀 go-pugleaf

In article <1993Jan27.020312.4883@nuscc.nus.sg>, eletanjm@nuscc.nus.sg (Tan Jin Meng) writes...
>b259phl@utarlg.uta.edu (PHILIP LIPPEL) writes:
>[ other stuff deleted ]
> 
>: The second Labmaster card is being used primarily for digital i/o from a 2-dimen
>: sional position-sensitive detector. Expected data rate is up to 100k events
>: /second. Each even requires a sixteen bit data transfer. All parts of this
>: system are " under development". Eventually the data taker will be a 486/33,
>: but we have transferred data at 100kHz on the 486/33 using a preliminary, 
>: DOS-based, program running in a DOS window. 
> 
>100kHz!!!! Is this correct??? This works out to 10 microseconds per
>event - which would make it almost an order of magnitude faster than the
>probable OS/2 context switch time.
> 
>Has anyone done realtime graphics display on an OS/2 machine?
>(DOS,WINOS2 or PM). What's the performance esp. with respect to timing
>jitter?
> 
>jin meng
Yes thats the correct rate. Context switching isn't really relevant-- it only
adds some time at the beginning of an acquisition cycle. We use a FIFO to store
a bunch of 16 bit numbers (representing the positions at which a particle hit
our detector). When the FIFO is half-full it generates an interrupt; you have
half the time to fill, then, to respond to the interrupt and empty the FIFO.
100kHz avg transfer rate is no problem. Whether or not you have time to do
anythig else while acquiring data at this rate depends on your machine, and
on how much you have to do with the received data. In our case we are just 
building a histogram in memory which amounts to a VERY short assembly program--
increment memory location X. 
P. Lippel