NEW: HP 15C Collectors Edition

[brouhaha]

(05-07-2023 11:54 PM)Valentin Albillo Wrote:  Such as Chess programming, for instance. See:
CPW - Chess Programming Wiki

Fully agree as far as Chess programming in assembly language goes. The 6502 would likely be just as inefficient for a Chess program written in C as it is for almost anything written in C. One of the fundamental problem is the lack of a suitable addressing mode for local variables in stack frames. Another is the lack of efficient 16-bit add and subtract. A Chess program written in 6502 assembly needs little of either, but for any non-trivial C program, the Z80 has the 6502 beat. (Being somewhat of a 6502 aficionado, it pains me to say that.) The 6502 makes up for that a little bit in having over 3x the memory bandwidth as the original Z80 when run at the same clock frequency, but newer Z80-compatible microcontrollers have significantly closed that gap.

There are techniques that a better C compiler could use to lessen the stack frame problem, involving static call graph analysis. Microchip's XC8 compiler for their 8-bit PIC microcontrollers does that. AFAIK, there is no existing C compiler for the 6502 that does it.

There's also no reason that any calculator designed today, other than a very cost-sensitive one, would need to use a 6502-derived microcontroller. At the time the Sunplus 6502-derived microcontroller was chosen fir use in various HP calculators, there was still a significant cost advantage to that 8-bit microcontroller compared to 32-bit microcontrollers (e.g. ARM). In the past decade that cost advantage has shrunk substantially.

One of the problems with making an ARM-based 17bii+ or 42s is that there aren't ARM microcontrollers with sufficient LCD drive capabilities for a passive dot matrix display of the required number of rows and columns. If an ARM microcontroller was used as the main processor, a second chip to drive the display would be required. It could either be a dedicated LCD driver chip, or a non-ARM microcontroller with suitable LCD drive capability.

For a midrange or high-end calculator, the added chip might not pose too much of a cost concern. It's doubtful that an ARM-based 17bii+ would offer enough improvement to justify the cost increase, but a higher-end calculator like a 28S or 42S might.

Note that the Swissmicros DM42 uses a VERY expensive LCD display with integrated electronics on the glass. (It's the same Sharp "memory LCD" that the late Richard Ottosen and I used in our 41 and 42 calculator prototypes.) That's part of why the DM42 is much more expensive than the HP 17bii+.