[Bug] Incorrect pwm frequency

[mstumbra]

tim1(dp.TIM1, 144, 1.mhz().into(), &clocks)
leads to actual timer frequency of 0.5 MHz, as code doesn't set/takes into account tim1sw bit in RCC CFGR3 register
Same thing with other timers and possibly peripherals most likely. (Occurs only when MCU is running with max frequency)

[David-OConnor]

Note that center align modes will also cause incorrect freqs.

[Sh3Rm4n]

I have not had the time to dig into the PWM issues / PRs yet, but can this possible resolved or improved by #196?

[Piroro-hs]

Looks like current PWM and Timer implementation is wrong at source clock calculation.

stm32f3xx-hal/src/pwm.rs

Lines 283 to 284 in f8eb292

	// TODO: ppre1 is used in timer.rs (never ppre2), should this be dynamic?
	let clock_freq = clocks.$pclkz().0 * if clocks.ppre1() == 1 { 1 } else { 2 };

[David-OConnor]

As in, some timers use apb1 while others use apb2? Here's how I solved it, where paste is a macro crate that lets you build identifiers from parts, $apb is apb1 or apb2, and the apb1_timer() and apb2_timer() are traits the clocks struct implements. (Ie that calculation line is moved to the clock functionality; the timer module determines which to use, apb1 or apb2.)

paste! {
    let mut timer = Timer { clock_speed: clocks.[<$apb _timer>](), tim };
    // ...
}

Example trait implementation:

    fn apb2_timer(&self) -> u32 {
        if let ApbPrescaler::Div1 = self.apb2_prescaler {
            self.apb2()
        } else {
            self.apb2() * 2
        }
    }