🚀 go-pugleaf

In comp.sys.transputer <1992Sep2.003439.2817@usage.csd.unsw.OZ.AU>
Ross Mark <rossm@aix04.csd.unsw.OZ.AU> writes :

> I am a 4th year computer engineering student at the University
> of New South Wales, doing a thesis project on transputers.
> I wish to memory map some devices on to a transputer. What I would
> like to know has anyone done this before and if so was there any 
> problems in doing so. My other question are there any TRAM boards
> that give full access to all the pins of the transputer. From what
> I can see I will have to design my own board. Something I will 
> happily do but would prefer not to. Even if some one allready has
> designed their own board it would be a great help if I could get 
> a copy of the circuit.

Someone once told me that the reason the (standard) TRAMS do not give you
direct access to the Transputer data and address lines is *specifically* to
prevent you from trying to memory map a device to the Transputer. The idea
is that you should use *link adapters* for that function so that all your 
I/O is done through channels. It's not a bad idea as far as modularity goes 
and I might have even considered it if the link speed were anywhere near the
I/O rate that you would get from direct memory access. Since that is not the
case, memory-mapped hardware still lives safely in the world of Transputers. 

There is only one problem that I am aware of in interfacing real hardware to
Transputers: if your device has a FIFO-type interface you have to make sure
that any dma to that device is *never* interrupted. I think this was a
recognized bug/feature on the T8's but I also had the same problem on T2's.
I *really* hope that Inmos fixes this problem with the T9000. Otherwise a
Transputers interface to a memory-mapped device looks like that of any other
processor. You might want to ask Inmos for some applications notes if you are
looking for circuits to get you started.

Jacob Hirbawi
JcbHrb@CERF.net