tainter@ihlpg.ATT.COM (Tainter) (02/13/88)
In article <1168@trotter.usma.edu>, bill@trotter.usma.edu (Bill Gunshannon) writes: > In article <11691@brl-adm.ARPA>, milne@ICS.UCI.EDU (Alastair Milne) writes: > > microprogram, it is word-addressed, whereas the Terak is byte-addressed > > (this was to gain 128K of memory). > I don't know what TERAK you have but mine has 32K of word addressable > memory of which only 28K is really memory. The rest are vectors. No. Your Terak has 64K BYTEs of memory which they like to call 32K WORDs of memory. The memory is byte addressable but word addressed operands must be word aligned for the 16 bit data bus. On the otherhand the microengine only gives you word addressable memory ( i.e. the smallest memory reference available is 16 bits). By replacing your processor board with the LSI 11 QBUS version of the PDP 11/70 you get split I and D for 128K bytes of memory (64K words) minus the memory mapped IO and interrupt vector space. > > Also, the III.x p-code which is > > hardwired into the MicroEngine is very different from I.5 and II.x -- > > I assume you're running one of the latter on the Terak. > > I am running IV.0 and re-compiling all my programs wouldn't be a problem. Recompiling your own programs will be no problem. Recompiling all the precompiled products you use will be. I do believe their was a utility which translated I.5 and II p-codes to III. > It would be more than offset by getting rid of a hog like IV.0 is on the > TERAK. After you load the O/S you end out with somewhere around 12K words > but because of it being word addressed it really looks like much less to > an application. I wasn't aware that IV was available for the TERAK. I do have a version of IV for my Pinnacle 68000 box though. Does Terak IV.0 still have real ^NULL interrupts or are they now polled on IO like they are on my Pinnacle? All the enhancements of IV are not worth losing real ^NULL interrupts to get. IV.x is not only a hog but is substantially slower than I.5. Of course, it also does a lot more for you. I measured I.5 and came up with a range of 5 to 15 times as slow as native code (it gets really bad in function call overhead). Under IV.2 my measurements gave me a slow down of 12 to 25 times as slow as native code. > >>> Much text on the wonders of the TERAK deleted to save space <<< > > Good luck. Hang onto those Teraks -- they have virtues that have rarely > > been seen since. > You want to buy some??? CHEAP??? What do you class as cheap? <400? For systems that were $5000 ten years ago that seems about right. Are Teraks orphaned? The last I knew Calcomp had bought Terak and was looking to ditch it. That was a while ago though. > bill gunshannon --j.a.tainter