Support SPI DMA

[guysv]

potentially closes #98, see comment there.

must admit this looks even slower than straight spi register to callback path... we probably want to squeeze the CPU better here.

[urish]

Thanks! Looks like there are some missing calls:

1. in doTX, after pulling an item from the txFIFO, I think we should call updateDMATx
2. in completeTransmit, after pushing an into to the rxFIFO, I think we should call updateDMARx

Also, please format the code with prettier (it should happen automatically when you commit, not sure why it was not working for you)

[guysv]

i just moved the calls to updatefifos seems related to there.

[guysv]

I think something broke, DMA SPI works fine but direct SPI writes x4 the data, ie every spi.write is translated into 4 register writes. very weird.

[urish]

Sounds weird indeed, how did you observe that behavior?

[guysv]

I added a micropython test that checks the spi call. should be ok. don't know what happened to what i found in the earlier message 🤣

[urish]

Thank you!

[urish]

Looks like the SPI test is failing with older releases of MicroPython (e.g. 1.19.1: https://github.com/wokwi/rp2040js/actions/runs/6247824728/job/17028285253?pr=131)

Mind looking into that?

[guysv]

looks like I forgot to pass the expected SPI messages as parameters to the CI npm test invocation. please try again.