this is already done during platform init. it wasn't even sound in the
original implementation because futures would meddle with the nvic in
critical sections, while another (interrupt) executor could meddle with
the nvic without critical sections here. it is only accidentally sound
now and only if irq1 of both pios isn't used by user code. luckily the
worst we can expect to happen is interrupt priorities being set wrong,
but wrong is wrong is wrong.
since we never actually *disable* these interrupts for any length of
time we can simply enable them globally. we also initialize all pio
interrupt flags to not cause system interrupts since state machine
irqa are not necessarily meant to cause a system interrupt when set. the
fifo interrupts are sticky and can likewise only be cleared inside the
handler by disabling them.
dma does this too, also with 12 bits to check. this decreases code size
significantly (increasing speed when the cache is cold), frees up an
interrupt handler, and avoids read-modify-write cycles (which makes each
processed flag cheaper). due to more iterations per handler invocation
the actual runtime of the handler body remains roughly the
same (slightly faster at O2, slightly slower at Oz).
notably wakers are now kept in one large array indexed by the irq
register bit number instead of three different arrays, this allows for
machine code-level optimizations of waker lookups.
1403: Bump versions preparing for -macros and -executor release r=lulf a=lulf
I'd like to propose a new release of embassy-macros and embassy-executor, as there is a challenge with some of the features changing since 0.1.1 when using libraries that depend on 0.1.1 with applications that patch to use git versions.
Co-authored-by: Ulf Lilleengen <lulf@redhat.com>
1406: rp: DMA behaviour during flash operations r=Dirbaio a=kalkyl
This PR changes the old behaviour during flash operations where all DMA transfers were paused during the flash operation.
The new approach is to wait for any DMA operating in flash region to finish and let RAM transfers continue.
Co-authored-by: kalkyl <henrik.alser@me.com>
storing a full function pointer initialized to a resolver trampoline
lets us avoid the runtime cost of checking whether we need to do the
initialization.
rp-hal has done this very well already, so we'll just copy their entire
impl again. only div.rs needed some massaging because our sio access
works a little differently, everything else worked as is.
1378: Add ability to invert UART pins, take 2 r=Dirbaio a=jakewins
Same PR as before, except this now works :)
There was a minor hiccup in the UartRx code where the rx pin got passed as the tx argument, so the invert settings didn't get applied. With this fix, my local setup at least is happily reading inverted uart data.
Co-authored-by: Jacob Davis-Hansson <jake@davis-hansson.com>
1372: rp: add division intrinsics r=Dirbaio a=pennae
rp2040-hal adds division intrinsics using the hardware divider unit in the SIO, as does the pico-sdk itself. using the hardware is faster than the compiler_rt implementations, and more compact too.
since embassy does not expose the hardware divider in any way (yet?) we could go even further an remove the state-saving code rp2040-hal needs, but that doesn't seem to be worth it.
Co-authored-by: pennae <github@quasiparticle.net>
rp2040-hal adds division intrinsics using the hardware divider unit in
the SIO, as does the pico-sdk itself. using the hardware is faster than
the compiler_rt implementations, and more compact too.
This is useful in some cases where the surrounding circuit
for some reason inverts the UART signal, for instance if you're talking
to a device via an optocoupler.
A while ago `OutputOpenDrain` was made to implement `InputPin`,
something that allowed drivers for various one-wire protocols to be
written, but it's been lacking a `Wait` implementation — something
that's needed to write async versions of these drivers.
This commit also adds `get_level()` to `OutputOpenDrain`, since
`is_high()` and `is_low()` were already implemented, but `get_level()`
itself was missing.
1318: rp: Allow zero len reads for buffered uart r=Dirbaio a=timokroeger
Prevents the read methods from getting stuck forever.
cc `@MathiasKoch` can you test if this fixes the problem you described in the chat?
Co-authored-by: Timo Kröger <timokroeger93@gmail.com>
When using gpio pin changes for things like peripheral interrupts these
debug! calls flood defmt, making it difficult to find what you're
actually looking for.
1142: More rp2040 BufferedUart fixes r=Dirbaio a=timokroeger
* Refactor init code
* Make it possible to drop RX without breaking TX (or vice versa)
* Correctly handle RX buffer full scenario
Co-authored-by: Timo Kröger <timokroeger93@gmail.com>
* Only clear interrupt flags that have fired (so that we do not lose any error flags)
* Enable RX interrupt when a read is requested, disable it when the RX buffer is full
* Rework TX interrupt handling: its "edge" triggered by a FIFO threshold
When data is in the RX fifo the RX timeout interrupt goes high again even after clearing it.
The result is a deadlock because execution is stuck in the interrupt handler. No other code
can run to clear the receive buffer.
Enable and disable RX interrupts based on the buffer fill level.
Use the same approach for the TX code path.
1128: Add missing SPI pins r=Dirbaio a=pferreir
The SPI definitions lack the pins which are not accessible on the pico (but are so e.g. on the stamp).
Co-authored-by: Pedro Ferreira <pedro@dete.st>
959: Generic, executor-agnostic queue implementation r=ivmarkov a=ivmarkov
Hopefully relatively well documented.
Implementation relies on a fixed-size `SortedLinkedList` from `heapless`. (By default, for up to 128 timer schedules, but we can lower this number to - say - 64.)
As discussed earlier, on queue overflow, the `WakerRegistration` approach is utilized, whereas the waker that is ordered first in the queue is awoken to make room for the incoming one (which might be the waker that would be awoken after all!). Wakers are compared with `Waker::will_wake`, so the queue should actually not fill up that easily, if at all.
I've left provisions for the user to manually instantiate the queue using a dedicated macro - `generic_queue!` so that users willing to adjust the queue size, or users (like me) who have to use the queue in a complex "on-top-of-RTOS-but-the-timer-driver-calling-back-from-ISR" scenario can customize the mutex that protects the queue.
The one thing I'm not completely happy with is the need to call `{ embassy_time::queue::initialize() }` early on before any futures using embassy-time are polled, which is currently on the shoulders of the user. I'm open to any ideas where we can get rid of this and do it on the first call to `_embassy_time_schedule_wake`, without introducing very complex combinations of critical sections, atomics and whatnot.
Co-authored-by: ivmarkov <ivan.markov@gmail.com>
Co-authored-by: Dario Nieuwenhuis <dirbaio@dirbaio.net>
