Re: [PATCH 00/14] Modularize schedutil

[Quentin Perret]

Hi Rafael,

On Tuesday 12 May 2020 at 19:30:52 (+0200), Rafael J. Wysocki wrote:
> I don't see any technical reason for making schedutil modular in the
> context of GKI other than to make the GKI image smaller, but I don't
> expect that to be significant enough.

The fact that we can make the image smaller, and we give vendors one
less reason to not-want GKI _is_ desirable IMO.

  $ size vmlinux.*
     text	   data	    bss	    dec	    hex	filename
  19225963	9601976	 491084	29319023	1bf5f6f	vmlinux.after
  19230599	9603236	 491084	29324919	1bf7677	vmlinux.before

^ that's with the series applied. 'before' means sugov is =y, and
'after' is sugov =m. So modularizing saves just over 4K on text, and a
bit of data too. Is it significant? Maybe not. But it's quite likely
that those who don't use schedutil will find any unnecessary byte to be
one too many.

I just checked the size of modules in the default arm64 defconfig, and
the median is ~4K of text. The average is a little bigger, but mostly
because of a small number of really large modules (nouveau being the
prime the example). So all in all, the sugov module is not particularly
small by comparison with other things that have been modularized. A lot
of small things can lead to significant savings at the end.

Thanks,
Quentin