Implement RP2350 initialization clocks, uart, and (partial) USB support

[cibomahto]

This patch enables the clocks, uart, and partial USB support for the 2350.

I left the default console as 'uart' because if it is set to USB, using 'fmt' causes a panic. USB transfers work fine if only a single byte is transferred at a time.

Another issue is that the clock speed for RP2350 is set to 125MHz, because there were a number of places where this was hardcoded when setting up peripherals. This should at least be increased to 150MHz, perhaps as a new constant in machine_rp2_2xxxx.

[deadprogram]

Once you make the fix I suggest @cibomahto you should rebase this branch against the rp2350-add branch, since it is now rebased against the latest dev. If you do that, this PR should pass all CI tests.

[deadprogram]

@cibomahto I had to rebuild the cached LLVM because of reasons unrelated to this PR.

All tests are now passing! 🥳

[deadprogram]

@cibomahto looks like you need to run go fmt on src/machine/machine_rp2350_rom.go

[cibomahto]

@deadprogram Everything passes again

[deadprogram]

I just confirmed that the bootloader changes work as expected @cibomahto great work!

@soypat IMO this PR can now be merged into #4459 and that PR can be set to ready for review

[soypat]

We can merge this but I can see a lot of things I'm going to want to unexport before next release

[deadprogram]

but I can see a lot of things I'm going to want to unexport before next release

@soypat do you mean functions that you think should not be exported from the package?

[soypat]

@deadprogram yes. A cursory glance showed a lot of USB and Clock exported functions which did not need to be exported or could better be hid from package scope behind methods on a type

[deadprogram]

A cursory glance showed a lot of USB and Clock exported functions which did not need to be exported or could better be hid from package scope behind methods on a type

That is probably the case with the current rp2040 implementation as well. In any case we can address those things separately.

[deadprogram]

Now rebasing this branch into the rp2350-add branch for further development. Thank you very much @cibomahto and @soypat

[cibomahto]

@deadprogram yes. A cursory glance showed a lot of USB and Clock exported functions which did not need to be exported or could better be hid from package scope behind methods on a type

Yes, apologies that I'm not fluent enough in Go to know the right way to do this yet :-)

[cibomahto]

Another thing for whatever list these go on, the USB driver both chips are almost exactly the same*, but the CMSIS register names are slightly different between the two. It would probably be best to merge them back together if possible.

• Except for an extra init in the rp2350, and the rp2350 not needing the errata from the rp2040