Ken Shirriff has, as so often, a nice table to start with - especially nice to detect the 'undocumented' ones. All opcodes in lower case are 'undocumented'. With sorting his table according to the 'octal' (2-3-3) decoding logic the 8085 uses (*1), the first are nicely grouped where the 8080 only decoded
NOP - Not undocumented, but on the 8080 it filled the whole 0-x-0 group)
DSUB - 16 bit subtract
HL - BC
ARHL - Arithmetic Right shift HL
RDEL - Rotate (shift) DE Left into carry, fill in zero
RIM - Not undocumented, but added with the 8085)
LDHI - Load DE with HL plus Immediat8 (HL+i8->DE)
SIM - Not undocumented, but added with the 8085)
LDSI - Load DE with SP plus Immediat8 (SP+i8->DE)
The others are scattered in remaining holes for 16 bit operations and jumps (where they belong anyway)
RSTV - ReSTart if V is set
SHLX - Store HL indeXed by DE
JNK - Jump on Not K
LHLX - Load HL indeXed by DE
JK - Jump on K
JNK are about the (equally undocumented) K-flag (*2), while
RSTV is about the (similarly undocumented) V-Flag (*3).
(Jules did already a great write-up about those and how they may these may be translated into x86 in his answer to a related question)
These instructions would have made
DE a quite versatile register for 16 bit address calculation in indirect operations (*4). These instructions were present in all 8085, including second source, I ever tested (*5).
I seriously have no idea why they where kept 'secret', as they would have improved HLL handling a lot. Morse offers an explanation in his paper Intel Microprocessors: 8008-8086 (MS-Word Document) published in IEEE Computer, Vol 13, No. 10, page 46, October 1980
Several other instructions that had been contemplated were not made available because of the software ramifications and the compatibility constraints they would place on the forthcoming 8086.
This is not only incorrect, as these instructions were available, but, to me it sounds quite like hindsight, as the 8085 was delivered a full two years before the 8086, about the time the 8086 project started. Further all of the 16 bit instructions are (in some form) present in the 8086 (as well as the Overflow Flag) and translation would have been as straight forward as with the rest. Even more so, the only two instructions officially added,
SIM, are among the ones explicit not supported by the 8086. (*6)
*1 - Less obvious in Pastraiser's 8085 opcode table in classic hex notation - but maybe more like we're used to.
*2 - Which might have been the one intended for signed comparison, but that's a different story.
*3 - The V-Flag signals overflow, much like the V flag on the 6502. With
RSTV this might have been intended for easy insertation of overflow exception handling.
*4 - I really used them a lot, as
DE now became a quite useful register instead of being a mere second
HL usable via
XHNG. Address calculation for structures became quite doable without bloating the instruction set.
*5 - Mostly SIEMENS SAB8085, but also Toshiba, NEC, AMD and maybe others.
*6 - My personal speculation is that it wasn't, as so often, not a technical or design decision but made on management level to keep it '100%' 8080 compatible (whatever that meant for management blurb at that time) and thus suppressing the nice add ons engineering had made (*7).
SIM eventually as exception, as they where needed to support the new interrupt feature.
*7 - Which leads to another thought: Having the ability for fast offset calculation to a memory (
HL or stack pointer, 16 bit left shift for word address calculation and 16 bit indexed loads) is quite pleasing for HLL programming. So while Mr. Morse does rightfully deserve credit for making the 8086 with HLL in mind, these instructions indicate that others at Intel already tried to push into that direction before him and he knew about - and joined the trek.