A 65C816 version of the "NMOS 6502 Opcodes" is inevitably going to overlap the 65C816 Programming Manual, so rather than just repeating the material, I took the opportunity to fill in the gaps and address some of the things that come up on the forum from time to time. Things like differences between 8-bit native mode and emulation mode (TXS is one that's not obvious at first blush).

There is a lot more 65(C)02 code than 65C816 code out there, so I wanted to address things that might come up when adapting code for native mode. Adapting to 8-bit native mode before updating routines for 16 bits is one way to do this incrementally.

I also wanted to point out the things that are actually different between the 65(C)02 and emulation mode. For example, JMP ($12FF) get the high byte from $1300 (like the 65C02) but takes 5 cycles (like the NMOS 6502), so it's not exactly like either.

Also, many simulator writers seem to be very interested in replicating absolutely everything -- even things that would never break any code that isn't a unit test. For example, V flag in decimal mode, the fact that the high byte of the LDA ($FF),Y pointer is $0000 not $0100, etc.

Somewhat surprisingly, very few people (even those who are otherwise very knowledgable about the 65C816 and the 6502 family generally) are really all that familiar with when page/bank wrapping does or does not occur. So that seemed like another good gap to fill.

Still, the thrust of the article is native mode. I used many 16-bit examples because 2 bytes allowed me to illustrate when wrapping did and did not occur with a single example.

this is the latest corrected version, converted to HTML.

Bruce Clark's long and detailed document about the 65C816 opcode behaviour is now available online - see the Tutorials and Primers section of the site. It's mostly about the native modes, and goes into some detail on the exact behaviour. He called for proofreaders a while ago, and this is the latest corrected version, converted to HTML.

although I had suggested to Bruce to use a different way of describing the registers (e.g., SP for the stack pointer, not S). Most machine language monitors don't use single letters to refer to the "control" registers, such as PC and SP.

Did machine-language monitors move away from the standard in order to reserve the single letters like S for commands?

5.9 (DIRECT)

Length: 2 bytes, $OP $LL

(direct) addressing uses a 16-bit pointer. When the e flag is 1 and the DL register is $00 (both conditions must be met), the address of the pointer is:

I think the doc is right - this is one of the obscure cases.