impl waveform with TIM OC Channel DMA #2410
Conversation
| if !original_enable_state { | ||
| self.disable(cc_channel); | ||
| } | ||
|
|
||
| self.set_duty(cc_channel, original_duty_state); | ||
|
|
||
| // Since DMA is closed before timer Capture Compare Event trigger DMA is turn off, | ||
| // this can almost always trigger a DMA FIFO error. | ||
| // | ||
| // optional TODO: | ||
| // clean FEIF after disable UDE | ||
| if !original_cc_dma_enabled { | ||
| self.inner.set_cc_dma_enable_state(cc_channel, false); | ||
| } | ||
|
|
||
| if !original_cc_dma_on_update { | ||
| self.inner.set_cc_dma_selection(Ccds::ONCOMPARE) | ||
| } |
There was a problem hiding this comment.
This also would apply to your original PR, but I think this needs special logic to also apply on Drop: if one uses e.g. embassy_time::with_timeout, the Transfer will be canceled as part of Transfer::drop, but this will not. Rust's std often creates a simple struct that executes a bunch of code in its drop handler to solve this problem, but you could probably make one that takes a closure.
There was a problem hiding this comment.
yeah, but we still need to wait dma::Transfer somehow expose a function to clean the flag.
|
And what's the point of the macro again? Are there some DMA channels that only work with some timer channels? |
At least for F4 MCU, the answer is yes. ST seems "randomly pick" the DMA channel mapping for TIMx_CHy. Each TIM has one TIMx_UP DMA Channel, but (at most?) 4 TIMx_CHy Channel. |
Notice:
.gen_waveform()to.waveform_up(), to match up with new.waveform_chx()P.S.
I have to copy&paste™ entire "waveform for UP" implement, and do a few modification. It's too tricky to exclude some code with a
$()?inside a MacroTranscriber:.waveform_up()requires user input atimer::Channel, but.waveform_chx()don't. I don't know how to achieve this inside one macro.