OK, I'm slowing doing some bits and pieces in my little Ruby OS to 816-ify it... and there's one little thing that I'm finding just weird. I'm probably missing something completely obvious, but right now can't see it ...

For the most part I'm running in native 816 mode with 8-bit registers.

I have a memory-fill routine. Thought I'd be clever and use the 816 block-move instruction. The code takes the value from the command line via an 'atoi' routine (understands decimal, $hex or %bin) and stores it in a 16-bit memory location in page 0 called atoi0 and atoi1, so then ...



The aix8 and aix16 instructions are macros:



This is ca65 and .a8/16 and .i8/16 are used to indicate to the assembler to use 16 bit or 8 for Acc or XY.

The assembler list output is as I'd expect:




Anyway, here is the odd bit. This works for all values I've tested except zero. If I ask it to fill with zeros, then it crashes in an obscure way and scribbles over memory with a repeating 4-byte pattern: $37, $00, $35, $37 ...

I've tried changing the lda atoi0 with lda #$0 and it fails the same way (but works for other values I've tried).

So have I missed something obvious, being stupid, or something else?

Any clues welcome!

Cheers,

-Gordon

_________________
--
Gordon Henderson.
See my Ruby 6502 and 65816 SBC projects here: https://projects.drogon.net/ruby/

Please forgive me if these are silly questions, but:
1) do you have any NMIs and
2) is your subroutine located in bank 0 RAM, causing you to sled all the way to your cold start after the move?

_________________
Got a kilobyte lying fallow in your 65xx's memory map? Sprinkle some VTL02C on it and see how it grows on you!

Mike B. (about me) (learning how to github)

Yes - there is only one bank of 64K in this system. The routine should clear from $0000 to $CFFF.

The code lives in higher RAM - the addresses in the listing output are relocatable ones generated by ca65.

-Gordon

_________________
--
Gordon Henderson.
See my Ruby 6502 and 65816 SBC projects here: https://projects.drogon.net/ruby/

OK. I have worked it out.

It's the count.

According to the book, you set count -1 in the 16-bit A register.

So I was setting the address of the start of the OS ($D000) minus 1. I've "manually" written the first byte, and I'm copying one less (to location 1), so I really ought to have -2 in there.

And when I do that it works.

It wasn't working before because I've not got the '816 BRK vector setup, so if it jumped to $D000 after the copy, it hits a BRK which goes to ($FFE6) in '816 mode which "blows up". With all the other numbers I tried, it seemed to just work - or skipped over them. Lucky I guess, but led me down the wrong path.

So, ok, as you were, please have a chuckle at me in my Gordon is a moron moment (old UK folks will know that one)

Cheers,

-Gordon

_________________
--
Gordon Henderson.
See my Ruby 6502 and 65816 SBC projects here: https://projects.drogon.net/ruby/

Nicely done. I knew you probably had the best chance of solving the issue, being reasonably competent and situated right there and all ...

https://www.youtube.com/watch?v=LTKORcr1jhY

_________________
Got a kilobyte lying fallow in your 65xx's memory map? Sprinkle some VTL02C on it and see how it grows on you!

Mike B. (about me) (learning how to github)