Re: [PATCH v2] pwm: bcm2835: Support apply function for atomic configuration

[Sean Young]

Hi Uwe,

On Fri, Dec 04, 2020 at 10:55:25PM +0100, Uwe Kleine-König wrote:
> On Fri, Dec 04, 2020 at 11:40:36AM +0000, Sean Young wrote:
> > On Fri, Dec 04, 2020 at 12:21:15PM +0100, Uwe Kleine-König wrote:
> > > On Fri, Dec 04, 2020 at 12:42:15AM +0100, Lino Sanfilippo wrote:
> > > > On 29.11.20 at 19:10, Uwe Kleine-König wrote:
> > > > > You're storing an unsigned long long (i.e. 64 bits) in an u32. If
> > > > > you are sure that this won't discard relevant bits, please explain
> > > > > this in a comment for the cursory reader.
> > > > 
> > > > What about an extra check then to make sure that the period has not been truncated,
> > > > e.g:
> > > > 
> > > > 	value = DIV_ROUND_CLOSEST_ULL(state->period, scaler);
> > > > 
> > > > 	/* dont accept a period that is too small or has been truncated */
> > > > 	if ((value < PERIOD_MIN) ||
> > > > 	    (value != DIV_ROUND_CLOSEST_ULL(state->period, scaler)))
> > > > 		return -EINVAL;
> > > 
> > > I'd make value an unsigned long long and check for > 0xffffffff instead
> > > of repeating the (expensive) division. (Hmm, maybe the compiler is smart
> > > enough to not actually repeat it, but still.)
> > 
> > I wonder where you got that idea from.
> 
> I don't know how to honestly answer your question.
> Which idea do you mean? That divisions are expensive? Or that compilers
> might be smart? And do you consider it a good idea? Or do you disagree?

I had already made this exact suggestion -- and you had replied to my
email making that suggestion -- before you emailed this. Granted, I said
u64 and U32_MAX rather than unsigned long long and 0xffffffff.

However, I should not have sent that snotty email. It's irrelevant.

My apologies.


Sean