ADD: ARL support to vp20compiler

[dKoitka]

Can now use address relative indexing with passed in data. Useful for gpu vertex skinning

[JayFoxRox]

Please squash your changes; also, we usually try to avoid multi-line commit messages (= we only use a commit title, but no commit description).

Did you try addressing relative with offsets? Stuff like C[A0 + 5]? I'd appreciate some trivial samples in the PR description or references to projects which use these features (for maintainers so they can try the feature + users to learn about them).

[JayFoxRox]

Shouldn't this only work for PROGRAM_ENV_PARAM and be forbidden otherwise?

Also, to what source-reg does it apply? - shouldn't j be checked?
(Same goes for FLD_CONST or FLD_V which already existed - do we already prevent code from using it in multiple src registers?)

[JayFoxRox]

Would be cool if someone could finish this (or do minor cleanup and merge as-is, then create issues from review feedback).

[dKoitka]

Hey @JayFoxRox I'll jump on when I have some time, got distracted and forgot about updating this. I'll reply to some comments

[JayFoxRox]

That also works for me 👍 I was mostly trying to address the XboxDev maintainers.

This feature would be good to have, even if the implementation has minor issues (..which might not even exist, as I was simply asking about whether these cases are handled).

I consider this an issue with the XboxDev review process: It shouldn't delay 99.9%-done work for months.
We could squash-merge this as-is probably (and create issues from the remaining feedback).
Minor changes to the commit structure (which can't be fixed after merge) can be done by maintainers, imho.

If you address that anyway: Perfect!

[dKoitka]

I'll be happy to address the formatting issue, shouldn't take long. I don't remember seeing any specific case handled differently in the resource I was using to implement the ARL fix. However, if you'd like the ARL flag to be set in the only case I'm able to test, I'm confortable changing that. I believe it's only constants that I know for a fact work, I don't think R or V registers ever pop up for vp20compiler to process, but I think it's a possibility in some forum posts I remember looking at

[dKoitka]

I don't have any existing examples unfortunately, but could write one (which would piggy back off existing ones)

[dKoitka]

Not sure how to squash an existing merge request

[JayFoxRox]

I've fixed it for you.

---

In particular I did these steps in a terminal:

Set editor because I'll have to edit files as part of the git workflow

export EDITOR=nano

I downloaded your changes and made a working copy

git remote add dKoitka https://github.com/dKoitka/nxdk.git
git fetch dKoitka
git checkout dKoitka/fix/arl_support
git checkout -b fix/arl_support

I downloaded the latest nxdk version:

git fetch origin

I applied your changes to the latest nxdk version, -i[nteractively] which allows me to rename and edit things

git rebase -i origin/master

The interactive mode opens an editor now (nano for me) and I changed the second commit from pick to f[ixup] which means it gets combined into the other commit
I saved the filed and quit nano, so git started doing the pick and fixup step.

I then noticed another minor issue in the last commit and did this to edit the last commit (I also could have used r[eword] in an interactive rebase):

git commit --amend

In the editor, I renamed the previous commit and removed the long description.

Then I pushed my changes into your PR (which allows maintainers to edit it); force-pushing changes because I edited history / already public changes:

git push -u dKoitka fix/arl_support --force

[dKoitka]

thanks @JayFoxRox ! Sorry for ghosting like that, some life stuff happened, and I forgat again! Is the PROGRAM_ENV_PARAM comment you made the only thing remaining for this merge request?

[JayFoxRox]

Sorry for ghosting like that, some life stuff happened, and I forgat again!

No worries.

Is the PROGRAM_ENV_PARAM comment you made the only thing remaining for this merge request?

I think we can ignore that for now and turn it into a GitHub issue so we can check it later.

I assume someone will merge this, or I'll do a final review-pass and merge it in the coming days (although my life is also quite busy right now, so it might take a while).