Refactor VGA driver code to iteratively assign internal registers from main RAM about jcap HOT 8 CLOSED

It's a shame really. Those things have been figured out a long time ago ...

Check out this link, it's a generic loop transferring a long per hub cycle in either direction. There is also a counter based version pre-dating this one which - while OK for me - wasn't generally accepted (and sometimes you really don't have a spare counter). Anyway, use (one of) them if you want speed ...

from jcap.

I apologize for the mess Parallax made during forum transition(s) which is the reason I stopped participating.

from jcap.

@konimaru Thanks, looks like just the thing I need. Hadn't even gotten around to Googling for a solution yet so thanks for saving me a step! Seems like a solution to this takes up just as much real estate as explicitly setting variables, no matter how you do it.

from jcap.

Your call. I - obviously - use this in most of my drivers (where applicable). And the setup is only required once. In the end it's really down to timing: it fits or it doesn't :)

from jcap.

@konimaru True story! I'm definitely riding the assembly-timing learning curve into outer space. I'm actually going to have to do a refactor of my VGA driver to populate a line buffer during the visible porch/sync regions, instead of racing to grab a new tile every time I need it. As-is the driver works fine up to 15 horizontal 16x16 tiles, or 16 horizontal 8x8 tiles. Past that, de-sync.

from jcap.

Satisfied by commit a2ff0f2

from jcap.

For the record, I just realised you only wanted a one-off load not something repeating (where speed might be important).

from jcap.

@konimaru no problem! There's an entirely realistic chance the driver will change in the future to make that necessary, and being aware of the execution-time conscious approach is useful.

from jcap.