Date   : Mon, 27 Jul 1992 10:25:28 -0400
From   : Jay Sage <sage@ll.mit.edu>
Subject: Re: Undocumented Z80 Opcodes

   Willy recently added, "As things turn out, the Hitachi '180 traps all
unused opcodes."  This will not help, of course, if the undocumented Z80
opcodes are no longer "undefined" in the Z180 or Z280.  They will be
processed according to their new implementation.

   I got a kick out of Joe Wright's comment, "The Jay Sage I know has both
Z80 and Z180 machines at hand.  My Jay would be answering these questions
instead of posing them."  OK, I admit it.  I was being lazy; I could have
looked this stuff up myself.  But it appeared that there were others with a
much more burning interest in the subject.

   I completely agree with Joe's implication that unsupported opcodes and
unsupported programming language features simply should NOT be used.  As far
as my home controller is concerned, here's my excuse.  That machine was
built about 12 or 13 years ago, when 2K EPROMs cost $50 each (and 50 dollars
were a lot more money).  The whole unit only had 1K of EPROM (and 256 bytes
of RAM!).  The poor yield (by today's standards) on 2716 2K EPROMs provided
lots of chips with only 1K working, and Intel sold those as 2758s for much
less money (see footnote below).  So it seemed like a good idea at the time
to save a few bytes by using some of the 8085's undocumented opcodes. 
Actually, in retrospect, it think it was still a bad idea -- it's just not
good practice.  It rarely imposes a significant cost to do things right, but
the cost of doing things wrong can be horrendous.

   Joe added, "It is academically interesting I guess."  I think so, too. 
In particular, I was interested to know how the manufacturers felt about
their undocumented codes and, specifically, the extent to which they felt
that they should carry them over to the new chips, even though they never
officially supported them.

-- Jay Sage

P.S. Several years later I was building some computer hardware at work.  In
an attempt to save the company a little money, I specified 2758 EPROMS
instead of 2716s.  Unfortunately, by then the yield on 2716s was so high
that there were no longer any half-defective 1K units, and 2758s were being
specially made only for back-compatibility.  Being low-volume items, they
commanded a premium price, and the company paid $50 each for them (instead
of the $10 I had written on the purchase requisition) when they could have
gotten 2716s for only about $20.  You gotta watch out for progress!



End of INFO-CPM Digest V92 Issue #74
************************************