search

atsamd-rs / atsamd

Target atsamd microcontrollers using Rust
https://matrix.to/#/#atsamd-rs:matrix.org
Apache License 2.0
558 stars 199 forks source link

SAMD5x/E5x PACs missing register field TC.CTRLA.DMAOS #147

Closed proman21 closed 4 years ago

proman21 commented 4 years ago

Another SVD mismatch with the reference manual.

Pretty much what it says on the tin. The DMAOS field of the CTRLA register for TC peripherals is not present in the SVD files for all processors in the SAMD5x/E5x family, but is present in the Reference Manual. All possible TC mode configurations (COUNT8, COUNT16, COUNT32) in the associated PACs are missing this field.

jacobrosenthal commented 4 years ago

We could use this issue as an opportunity to discuss how wed like to implement patching. In em32 I stole all the stm32 patching infrastructure. Its all based on python. That commit is here so you can get an idea of what it took https://github.com/em32-rs/efm32hg-pac/commit/97bc1eb248cf00d90b154e074d3e693976bd344c

sajattack commented 4 years ago

Yeah, we've already slightly altered the SVDs from their originals. Would you be able to diff them with originals and turn those into patches?

jessebraham commented 4 years ago

The patching from stm32-rs is in the process of being split out to stm32-rs/svdtools.

Might want to wait until stm32-rs/stm32-rs#331 has been merged.

proman21 commented 4 years ago

Not to be a issue purist but it would be better to file a separate issue for the SVD patching. Just cause there are some other things to chat about this issue.

Which is that the reference manual is not consistent w.r.t the DMA Oneshot functionality of the TC and TCC peripherals. Chiefly, because the TC peripheral does not have a DMA One-shot command in the Control B register but the relevant configuration option exists in the Control A register. So either the DMA One-shot functionality is supposed to match the TCC peripheral's implementation, or it is supposed to have no DMA One-shot functionality.

I'd recommend holding off on any changes to this part of the SVD until I can verify the actual behaviour in silico.

sajattack commented 4 years ago

Is this fixed with the new svds?