Multicore support for ESP32 and ESP32S3

[bjoernQ]

We should have ways to start and run code on the second core on dual-core chips

[jessebraham]

Duplicate of #28

[bjoernQ]

This is not a duplicate of the mentioned issue - starting to work in this

[bjoernQ]

@jessebraham @MabezDev I started to work on this and was wondering if I should spawn the code on app-cpu in the same way ESP32-HAL did, or if I should try to use a closure .... seems to work but unsure if I like it so here is how the new example looks like now

#![no_std]
#![no_main]

use core::sync::atomic::{AtomicI32, Ordering};

use esp32_hal::{clock::ClockControl, pac::{Peripherals, TIMG1}, prelude::*, CpuControl, RtcCntl, Timer};
use esp_println::println;
use nb::block;
use panic_halt as _;
use xtensa_lx::mutex::Mutex;
use xtensa_lx_rt::entry;

#[entry]
fn main() -> ! {
    let peripherals = Peripherals::take().unwrap();
    let system = peripherals.DPORT.split();
    let clocks = ClockControl::boot_defaults(system.clock_control).freeze();

    let mut timer0 = Timer::new(peripherals.TIMG0, clocks.apb_clock);
    let mut timer1 = Timer::new(peripherals.TIMG1, clocks.apb_clock);
    let mut rtc_cntl = RtcCntl::new(peripherals.RTC_CNTL);

    // Disable MWDT and RWDT (Watchdog) flash boot protection
    timer0.disable();
    timer1.disable();
    rtc_cntl.set_wdt_global_enable(false);

    timer0.start(1u64.secs());
    timer1.start(500u64.millis());

    let counter = xtensa_lx::mutex::SpinLockMutex::new(AtomicI32::new(0));

    let mut cpu_control = CpuControl {};
    let mut cpu1_fnctn = || { cpu1_start(&mut timer1, &counter); };
    let _guard = cpu_control.start_app_core(&mut cpu1_fnctn).unwrap();    

    loop {
        block!(timer0.wait()).unwrap();

        let count = (&counter).lock(|counter| counter.load(Ordering::Relaxed));
        println!("Hello World - Core 0! Counter is {}", count);
    }
}

fn cpu1_start(timer: &mut Timer<TIMG1>, counter: &xtensa_lx::mutex::SpinLockMutex<AtomicI32>) -> ! {
    println!("Hello World - Core 1!");
    loop {
        block!(timer.wait()).unwrap();

        (&*counter).lock(|counter| {
            counter.store(
                counter.load(Ordering::Relaxed).wrapping_add(1),
                Ordering::Relaxed,
            );    
        });
    }
}

I like that there are no static muts but

    let mut cpu1_fnctn = || { cpu1_start(&mut timer1, &counter); };
    let _guard = cpu_control.start_app_core(&mut cpu1_fnctn).unwrap();

Doesn't look too nice. The need to declare that cpu1_fnctn (to keep it alive when passing the reference to start_app_core) is probably understandable and not that weird.

But not so nice is that I can't move the timer1 into the closure because I struggle to call the FnOnce via the reference.

It's probably not that bad but before going further into that direction I just wanted to hear your opinions on that - if you don't like that I can revert that. On the other hand we can also keep that and see if it works good enough and change back if not ... it's not that much work