How to write LUT Exec assembly code?

Rayman

Maybe I forgot, but how does one write assembly code for LUT exec that PNut can load?

I've got one cog that is full and had to move code to HUB for hubexec.
That's all good and straightforward, but maybe LUT exec code would be faster...

cgracey

Just ORG it at $200 and move it into the lut at runtime using SETQ2+RDLONG.

Rayman

Thanks. That sounds simple enough...

Can I still use "fit" for the cog code?
I.e., can I put the org $200 after fit $1f5 ?

cgracey

Rayman wrote: »

Thanks. That sounds simple enough...

Can I still use "fit" for the cog code?
I.e., can I put the org $200 after fit $1f5 ?

Yes. And you can put another fit, like $300, if you want to limit your code to the first half of the lut. Just do a JMP $200 (or, to your label) to execute it.

Rayman

Ok, I just got around to trying this...

This loads LUT:

'copy in LUT exec code
              setq2     #512-1
              rdlong    0,##@MainLutExecCode

And the code goes here:

DAT 'Main LUT exec code
org     $200
MainLutExecCode


              fit       $3FF

Interestingly, this appears to work even if LUT code is above or below $400.

This is nice. I also like how I can copy this code to hubexec area like this:

DAT 'Main Hubexec code
              orgh
              'org

Not sure why, but using "org" here breaks it.

I've also found that self-modifying code doesn't work in either LUT or HUB areas. Obvious, I suppose.

potatohead

Yes, the self modify happens in COG registers.

When in LUT and HUB, it's necessary to think of the COG memory as a bunch of registers.

When in COG, the code is in the registers, which makes the self modify possible and practical.

I personally shift gears, from a memory to memory model (COG), to a more traditional load, store model (HUB, LUT)