Re: [PATCH v3 4/8] pwm: lpss: Include headers we are direct user of

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Tue, Sep 27, 2022 at 05:10:53PM +0200, Uwe Kleine-König wrote:
> On Tue, Sep 27, 2022 at 05:47:19PM +0300, Andy Shevchenko wrote:
> > For the sake of integrity, include headers we are direct user of.
> > 
> > While at it, add missed struct pwm_lpss_boardinfo one and replace
> > device.h with a forward declaration. The latter improves compile
> > time due to reducing overhead of device.h parsing with entire train
> > of dependencies.
> 
> Hm, I copied the cmdline for the compiler from a V=1 build and only run
> the compiler on drivers/pwm/pwm-lpss-pci.c.
> 
> With #include <device.h> I got:
> 
> 	real	0m0.421s
> 	user	0m0.354s
> 	sys	0m0.066s
> 
> With struct device; I got:
> 
> 	real	0m0.431s
> 	user	0m0.378s
> 	sys	0m0.052s
> 
> Are the numbers for you considerably different?

Why Ingo created thousands of patches to do something similar? Because for
a single user you won't see a big difference, but when amount of small pieces
are gathered together, you definitely will.

> > +struct device;

...

> > +struct pwm_lpss_boardinfo;
> 
> Hmm, I wonder why there is no compiler warning without that declaration.
> At least in my builds. Do you see a warning? IMHO it's better to fix
> that be swapping the order of struct pwm_lpss_chip and struct
> pwm_lpss_boardinfo.

Have I told about warning? It's a proper C programming style.
You don't have a warning because all pointers are considered to be the same,
but it is better style to explicitly point that out.

OTOH you may go and clean all the forward declarations of the data structures
in the kernel, but I believe the work wouldn't be appreciated much.

-- 
With Best Regards,
Andy Shevchenko