Upgrade direct mode support

[gorsat]

Add SETDP and review/update and/or remove logic that attempts to automatically use direct mode.

[gorsat]

Currently SETDP expects a 16-bit value and uses the MSB as the page number. I did it this way so that address labels can be used directly: SETDP MYLABEL

This may cause some confusion since in reality DP is an 8-bit register, so assuming that SETDP takes a byte value is reasonable.

I think I'll modify the behavior such that if SETDP is given an 8-bit value, then that value will be interpreted as a page number whereas 16-bit values will behave just as they do now.

[gorsat]

SETDP now works with both 8-bit and 16-bit args.

• 8-bit: SETDP xx means DP is set to xx
• 16-bit: SETDP mmnn means DP is set to mm (it's a syntax error if nn != 00)
• Note: When invoking SETDP with a label as address, e.g., SETDP TABLE, the label must be page-aligned or an error will be raised.
• SETDP with no arg turns off automatic use of direct mode addressing until/unless SETDP is invoked again with an arg
• Note: SETDP <expression> is not yet supported

[ceharris]

FWIW, I think it's fairly common to see SETDP used with a label in an integer divide or right shift expression; e.g.

setdp VARIABLES / 256

I like your approach of interpreting the effective size of the operand 👍, but just wanted to mention that support for an operand expression would be useful at some point.

I also wonder whether a warning might be more appropriate than an error when the label isn't page aligned. Might I not write code that accounts for the position of my data in the target page, even when the label I use to "pin" the page number isn't page aligned? Just a thought...

[gorsat]

Simple expressions are now supported so SETDP VARIABLES/256 should work. (Note: the expression can't contain spaces.) This was an easy fix.

Regarding args that aren't page aligned: I feel that the support for expressions provides everything needed here. If you have a label like this:

MY_LABEL  ORG  $1234

and you want MY_LABEL to be on the direct page then you can just

SETDP MY_LABEL/256

But explicitly telling the assembler to set direct page to a non-page-aligned value feels like an error to me. I would do that accidentally and I'd want it to fail. Granted, that's a subjective take, and probably influenced by rust rapping my knuckles all the time :-)

At any rate, splitting the validation code into success/warn/fail (instead of just success/fail) is more investment than I'd like to make in this since I think expressions gets us all the way there. I'll make sure to document this example (issue #8).

[ceharris]

I agree -- if you support SETDP expressions then there's no reason not to throw an error if the expression evaluates to more than 255 and the least significant byte isn't zero.

[gorsat]

Looks like we're good for now on this front.