This page has not been completed. Parts may be missing or reorganized before completed. Information is provided as-is and may have errors.
While not a multitasking operating system, the Prizm’s OS appears to have support for a fixed amount of “processes”, which only run when in foreground. A process usually comes into foreground on user request.
The amount of processes supported appears to be two. One is the main process, and the other is the child process. All built-in OS apps and add-ins run as the child process, except eActivity, which runs as the main process. When a strip inside a eActivity file is open, it runs as the child process, and alternating between the strip and the eActivity document equates to selecting what process is in foreground.
When another app or add-in is selected on the Main Menu, or when another strip is launched (or the running app inside the strip is changed), the old app in the child process terminates, and the new one takes place. GetKey appears to play an important role in process switching, as it is one of the few syscalls that handle proper Main Menu access and switching between a strip and the eActivity document.
It is not clear in which process things like the Main Menu or screenshot capture screen run - it is actually possible that there’s a third process set up for the Main Menu.
Each process has a different stack.
The main process’ stack has a size of 458751 bytes, in the address range 0x880F0000 to 0x8815FFFF on the CG10/20, or 0x8C0F0000 to 0x8C15FFFF on the CG50.
The child process’ stack, which is the one normally used, has a size of 524288 bytes, in the address range 0x88160000 to 0x881DFFFF (0x8C160000 to 0x8C1DFFFF on the CG50). This is also used by the static add-in RAM (static RAM comes from the start, the stack grows downwards, from the end). The start of the child process' stack is virtualized to 0x08100000, at least in the case of add-ins.
Stacks, along with the heap, are saved to flash on power off and restored when turning back on, meaning the power off actually works more like hibernation.
The Prizm OS has at least one heap, with a size of about 128 KB, located in the address range 0x881E0000 to 0x881FFFFF on the CG10/20, or 0x8C1E0000 to 0x8C1FFFFF on the CG50.
The implementation appears to have some bugs; namely, one can’t allocate one big buffer at the first try, and it appears to not manage free blocks properly, making fragmentation go too high at a certain point and basically making the heap useless. If the heap is used heavily, it may also get to a point (usually after freeing many buffers) where the OS reports there’s more heap available than the total heap size - and usually the values are quite out of range, like 10 MB. There is some more discussion on this forum thread.
The heap has the usual set of related syscalls.
On the CG50, the heap expands to 3MB (3072KB) when running the built-in Python add-in. This can be confirmed by opening [Test mode](https://prizm.cemetech.net/OS_Information/Secret_Key_Combinations/ in the application and opening “STACK/HEAP REMAIN”. The start of the heap remains the same, but the end address changes to 0x8C4DFFFF, utilizing some of the extra RAM available on this model. Add-ins cannot use this heap area via the usual malloc and free syscalls, but it is possible to write directly into its memory area, starting at 0x8C200000 to avoid the 128KB heap. Using this may be unsafe if Casio uses memory area for something else in the future so it is recommended to check the OS version and confirm that the area is zeroed out before using it.
The stack can provide a bigger continuous RAM area than the heap (up to about 500 KiB, depending on static RAM requirements, versus about 128 KiB). This, plus the limitations described above, means that to get the most out of the memory available to add-ins, one has to use non-standard (“incorrect”) practices such as preferably using the stack instead of the heap.