Update pygamer to HAL 0.17 #750
Conversation
ianrrees
force-pushed
the
port_pygamer_to_v2_b
branch
from
1550cb5 to
2f712ab
Compare
|
From some not-very-thorough testing today, I think fixing the timer-based neopixel issues will at least require some work in ws2812-timer-delay-rs because the timing it generates (via GPIO and waiting on a supplied timer) isn't quite right on SAMD51 with the new HAL. I've tried adjusting the library locally, and have the examples working well with opt-level="s", but then dialing it up to 1 or higher causes problems again. On paper, and based on the WS2812 datasheet I found, the current delays in that crate shouldn't work, so I think it's likely that fixing them for this case would just move the problem to some other user of that crate. Although the SPI neopixel example isn't working at present, that architecture seems a lot more reasonable to me; in particular the timer+GPIO approach would be a stretch to make work on a 48MHz SAMD21, but should be fine using a SERCOM on that part. I'm thinking that the appropriate thing might be to move the PyGamer neopixel examples away from the timer-based crate and to the SPI-based one. |
|
Debugging the SPI-driven WS2812 examples revealed that there are some subtle changes in behaviour of SPI writes between the old and new HAL. The old behaviour doesn't work reliably for me in release profile (the pixels show the rainbow, but flicker quite a bit), and the new behaviour doesn't work at all for WS2812 despite the SPI HAL implementation behaving correctly according to the So, at present, I think that the neopixel examples are broken for reasons outside of atsamd-rs. I've opened #752 to track that, since it's not specific to PyGamer although some of the issues are worse with a string of multiple neopixels. The question is whether we want to get those examples working before moving forward with this PR. My vote is to remove the neopixel examples from PyGamer until we can get them working, and proceed with this PR. |
Port the pygamer BSP to the `gpio::v2` and `sercom::v2::spi` APIs. While implementing this, I discovered that the pygamer uses an undocumented combination of pins for `Sercom1`. It uses PA17, PB22 & PB23. The datasheet indicates that PA17 is in `IoSet1`, while the other two pins are in `IoSet3`. It shouldn't be possible to use these together, but it appears there is an undocumented `IoSet`. Add this as `UndocIoSet1` for `Sercom1`. Also fix a warning from a trinket_m0 example.
ianrrees
force-pushed
the
port_pygamer_to_v2_b
branch
6 times, most recently
from
71362c4 to
7f711c6
Compare
This leaves NeoPixel examples in a broken state, it appears work will need to be done in WS2812 crates and/or the HAL to get timing working again.
These compile, but don't currently work reliably, due to changes in the HAL changing the timing of the asynchronous serial signals to the WS2812
ianrrees
force-pushed
the
port_pygamer_to_v2_b
branch
from
7f711c6 to
20bac54
Compare
ianrrees
changed the title
Summary
Initial work by @bradleyharden for v2 SERCOM, with more updates to account for improvements in our project and others.
Still some issues to work through, but at this stage all the examples compile:
CHANGELOG.mdfor the BSP