Advice on Casio Prizm development
A few years ago, I was very active in the Casio Prizm development community, having developed three notable add-ins, contributed to the Prizm wiki, libfxcg (my fork), and even done a bit of reverse-engineering (the calculator OS is closed-source and there is no official SDK), that resulted in the discovery of a couple of syscalls and more detailed documentation on some other ones. Because of this, once in a while I still get messages about my add-ins, which I’m happy to support when possible. Most annoyingly, I also get messages about Prizm development, usually about how to start making add-ins.
Why are these messages annoying? Because I don’t really know how to answer. When I started developing add-ins for the Prizm, I had little to no knowledge of the C programming language, and yet, despite the fact that add-ins can’t make use of all the stuff “normal” C programs can (the libc provided by libfxcg is incomplete; the filesystem uses a different API, there’s no threading, the stack is giant compared to the heap, etc.), I managed to learn it. It certainly helped that I had some previous experience with programming in other languages, even if it was just sloppy code, but I don’t have much of an idea of what to say to someone who intends to learn programming using the Prizm.
I usually end up saying that learning programming using the Prizm it’s a bad idea, probably coming across as extremely discouraging. However, I do hope it’s for the best, and that these people will still learn programming – just not by developing add-ins! Had my first contact with programming been through Prizm add-in development, most certainly I would have chosen other career path than computer engineering. I mean this seriously. I’m glad my first contact with programming was through sloppy Visual Basic code. Anyway, I already wrote a post on my programming experience – it needs updating, but it should do.
Learning how to program, even in a “easy” language and common platform, can be overwhelming; for a programmer that is used to higher-level programming, learning the Prizm, a poorly documented platform with a small developer community, can be overwhelming; combine learning how to program with learning a poorly documented embedded system, and it will most likely be very overwhelming. Of course, nothing will stop someone extremely motivated – hopefully, not even my less encouraging replies, or this blog post.
What follows is the partial reproduction of an email I recently sent, in reply to yet another of these inquiries on how to start developing for the Prizm. I have edited it to make it less specific to the situation of the person I was replying to. I’ll also use the terms “Prizm” and “fx-CG 50” interchangeably, as add-ins built for the former run, with a few exceptions due to sloppy coding (one of mine’s one of these exceptions…), on all Casio Prizm models: fx-CG 10, fx-CG 20 and fx-CG 50.
Do you have any previous programming experience? If not, I honestly do not recommend starting with the Prizm or any other Casio graphical calculator. If yes, then be aware that this is not an “easy” platform to develop for. Either way, here are a few reasons why:
- Prizm add-ins are written in C or very limited C++ (that might as well be considered C). By today’s standards, these are very low-level languages that require manual memory management and a very good awareness of the machine. They also provide very little protection from programmer mistakes. While some people had C as their first programming language, it is by no means a beginner-friendly language.
- Even if you already know C well, or if you learn it from any common book, tutorial or course, you’ll be disappointed to find out that much of the standard library is not present, or is insufficiently implemented, in the Prizm calculators.
- Add-in development for the Prizm was made possible through reverse-engineering and educated guesses based on what was known about previous models like the fx-9860G. While we now understand the essential things about the OS on these calculators, many things are yet to be known.
- Documentation is lacking and the community is not very large. This essentially means that you won’t be able to just google your way through many problems.
- Reverse-engineering/documentation and development efforts for the Prizm have basically stalled. You’ll also find many materials that mention the fx-CG 10/20, but since the fx-CG 50 is basically just a faster version of these with a mostly compatible OS (although some things like memory addresses have changed), almost everything you’ll ever need will still apply.
Now, I’m sorry if I came across as dismissive or as discouraging, I’m just trying to make sure you know what’s in front of you.
For Casio Prizm development specifically, this is where I can point you:
Prizm forums at Cemetech
Use these to ask any questions you might have and try to find solutions to any problems you encounter. There are also some guides there, mainly on how to set up the development environment (compiler and such), but I’m afraid they might be a bit out of date. However, as I said, development efforts have mostly stalled, so consider anything from 2014 or early 2015 as up-to-date. Specifically, do not follow the “[HOWTO] Prizm C Development” in there, as it is out of date.
This wiki contains much information on the calculator, the reverse-engineered OS functions (“syscalls”) that can be used from add-ins, etc. It also contains more up-to-date instructions on how to set up a development environment.
Personally, I have mostly moved on from Prizm development about three years ago, as I began pursuing a degree in Information Systems and Computer Engineering. Every year or so, I make a short comeback to fix urgent issues with my add-ins and eventually make them compatible with new OS versions and calculator models, as is the case of the fx-CG 50, as long as that does not require too much time/effort. As time passes and I work with other technologies, the more I realize more how “hard” of a platform the Prizm is, and the less motivated I am to build stuff for it again; the fact that I no longer use my fx-CG 20 nearly as much since high school, also doesn’t help.
I’m afraid I can’t help you much more, as I’ve forgotten much of what I knew about the Prizm, both the “theoretical” and “practical” knowledge, and I no longer have practical access to a development environment for it. I tried to put as much of my knowledge as I could into the Prizm wiki before I left, and I believe that the people that now frequent the Cemetech forums will be able to help you much better than I can.
I think that one day I might find some interesting in working on the Prizm again, but perhaps more from a reverse-engineering angle. As for the fun in developing for a constrained, embedded system, there are much more appealing constrained systems out there, like the ESP8266.
April 30, 2020 - 08:41
Years ago when I got a Prizm (2014/2015) I went down the path of trying to write C apps for it. I knew basic C, but getting that toolchain working was quite a step up from ‘apt install gcc’. After spending most of a Christmas break fighting through errors compiling the toolchain, I finally had a working Prizm toolchain (iirc I had to start the process on a linux machine and switch to windows to do the conversion to the special app format). I probably relied on your code and posts back in the day, so thank you!
Somehow after that I still went into computer engineering but it was years before I dared touch anything embedded that was more advanced than an arduino uno or raspberry pi. I now have experience with esp8266/32 and stm32, and I completely agree that picking a real MCU with a decent size community is the right way to go. Prizm development, even for someone with some experience, is far more frustrating and less rewarding than any of the common hobbyist platforms (esp32, avr, stm32 + other arm families).
It’s illogical how little support or docs Casio puts out, and a shame because the SH4 is a very unique CPU and with a proper environment the Prizm would make a great system for hands-on programming with its many nice buttons and large color screen. The new CG50 (essentially a Prizm with doubled CPU clock) does have MicroPython at least.
May 18, 2020 - 13:01
Thanks for sharing your story, and I’m sorry it took me so long to approve your comment – it got lost in the sea of spam.
I still think the Prizm would have been a bit more developer-friendly if it had become popular enough for there to be enough of a community to at least develop and maintain more user-friendly development tools. Unfortunately, there aren’t that many people with the necessary skills and motivation to mess with compiler toolchains and the like. At some point it also felt like the devices, at least the older models, were a bit prone to bricking, so that also scared people from “going deeper”. So the Prizm was always and will probably always be stuck in this limbo where there is lingering interest but never enough to kick off serious development of proper libraries, tools and documentation. Once the very few people behind those tools lost interest, myself included, the whole Prizm scene more or less came to a halt, other than the new occasional new add-in built using the same old outdated tools.