Recall Arithmetic: The haves and have-nots

[[kby]]

(05-22-2019 12:57 PM)Didier Lachieze Wrote:  

(05-22-2019 09:33 AM)[kby] Wrote:  Can’t really buy that explanation for the -65. Storage arithmetic is not merged so it would not be odd to have recall arithmetic unmerged.

Well, with unmerged recall arithmetic it takes one more step compared to standard recall. For example: RCL + 2 is 3 steps, while RCL 2 + is 2 steps because RCL 2 is a partially merged instruction.
Without recall arithmetic you loose the T-register, but does it justify the full implementation of recall arithmetic? I would say that during the design of the HP-65 (the first programmable pocket calculator) the designers may not have seen any advantage to implementing recall arithmetic.

I agree that programmatically it is of limited use. But sometimes keeping something in T in particular is useful because it duplicates downward as the stack drops. Sometimes the extra step might still be worth it.

That said it should also not have cost “a lot” unless the ROMs were full. It should not have taken anything special to implement the programming had the sequence already “just worked” like STO+-*/ does—no extra op-codes, no extra instruction buffers. The actual code for the ROMs to implementthe sequence should have been fairly trivial to port from the -45, one would think.