Crashes

[thebetter1]

I've been wondering exactly what causes the calculator to crash. There is no "crash" opcode. I have also never found any specific illegal operations; for example, overflowing a register just sets the value to zero or whatever. Even if the return command is left out of the end, shouldn't the calculator just keep running the rest of the code in ram until it finds another return?

[magicdanw]

Well, generally a crash doesn't occur due to improper opcodes, or anything. The CPU usually doesn't initiate a crash. What usually happens is, a program on the calculator does something the OS doesn't like, such as try to execute code past a certain point in ram, or it overflows the stack, or it starts running rampant, hitting so many returns that eventually the stack becomes negative, or something that eventually causes either the CPU or the OS to go "hey, something's not right." The OS gets control via some error handler, and since the OS doesn't really have any error-recovery abilities (since a program can theoretically screw up all the ram) it just initiates a ram clear to start over fresh.

It's something like that. I might not have gotten all the correct things that cause a crash, but generally it's just that something causes the OS to initiate a ram clear, to prevent the calculator from going catatonic.

[brandonw]

Yeah, it'll keep going until it gets caught in an infinite loop or runs garbage code that makes no sense.

For those "I've hung and when I press ON I'm going to get 'RAM Cleared'", it's usually because it's gotten stuck in some sort of loop and RAM is completely corrupted/unreliable so there's no way to recover, and pressing ON will trigger an interrupt. Deep in the interrupt code, it will do a checksum on the RAM and then compare it with its stored checksum, and if they don't match, RAM reset.

Is there a specific type of crash or reset you're thinking about? There are lots.

[FloppusMaximus]

I'd hazard a guess that in that type of crash, what usually happens is the calculator eventually jumps to a location in RAM page zero, either by "returning" to a garbage address, or simply by running off the end of page one. That causes the calculator to reset and hence turn itself off, and when it turns back on it finds that the RAM checksum is invalid, so it clears the RAM.

As obnoxious as the execution restrictions are, this does at least make many crashes easier to recover from.

[darkstone knight]

doesnt the CPU crash when IP (or w/e) ever exceeds $C0000?

[benryves]

RAM page 0 is mapped to $C000~$FFFF by default; so yes, if PC>=$C000 the protection hardware triggers a reset.

[darkstone knight]

you can see RST 00h as a crash...

as well JP $0, CALL $0, and JP >$C000