I am satisfied.

That is one architecture.  I suggest checking which others can use
the same treatment.


Frederic Cambus <[email protected]> wrote:

> On Mon, May 31, 2021 at 12:57:47PM +0200, Mark Kettenis wrote:
> 
> > > +option           FONT_SPLEEN8x16
> > > +option           FONT_SPLEEN12x24
> > > +option           FONT_SPLEEN16x32
> > > +option           FONT_SPLEEN32x64
> > > +
> > >  option           RAMDISK_HOOKS
> > >  option           MINIROOTSIZE=7360
> > >  
> > > Does this look reasonable?
> > 
> > I would skip some sizes.  8x16 is readable on any screen size where
> > 12x24 would be picked.  And maybe 16x32 is good enough for 4K screens
> > as well?
> 
> Right, the 16x32 variant is small but readable on my 27" 4k monitor, so
> it sounds reasonable to skip the 32x64 version for now. For the 12x24
> version I agree that it's not stricly necessary, just more comfortable.
> > 
> > > If it does and if we want to go this way, I can try to build a release
> > > and check if MINIROOTSIZE must be bumped on RAMDISK_CD. Then we could do
> > > the same for i386, armv7 and arm64.
> > 
> > I'm all for it, but last time this came up Theo didn't like it and
> > suggested adding code to scale up the fonts instead.  I really don't
> > think you want to upscale the 8x16 font to 32x64.  But if we add the
> > 16x32 font, upscaling that to 32x64 for the really big screens might
> > be an option and a reasonable compromise?
> 
> Indeed, if we want to go this way, I believe upscaling the 16x32 version
> to 32x64 would give good enough results on 4k screens.
> 
> > But figuring out how much things grow by adding the 16x32 font would
> > be a good start.
> 
> So I built three miniroot images using the same source tree, and
> results are below. There was no need to bump MINIROOTSIZE.
> 
> Result of running df -h on mounted vnode disks:
> 
> /dev/vnd0a     4.3M    4.1M    146K    97%    /mnt0
> /dev/vnd1a     4.3M    4.1M    146K    97%    /mnt1
> /dev/vnd2a     4.3M    4.1M    146K    97%    /mnt2
> 
> Sizes (in bytes) of compressed kernels extracted from miniroot images:
> 
> -r-xr-xr-x  1 root  wheel  4215406 Jun  1 14:47 bsd (baseline)
> -r-xr-xr-x  1 root  wheel  4217267 Jun  1 15:33 bsd (+Spleen 16x32)
> -r-xr-xr-x  1 root  wheel  4218986 Jun  1 20:26 bsd (+Spleen 12x24 and 16x32)
> 
> Size difference from baseline:
> 
> With Spleen 16x32:           +1861 bytes
> With Spleen 12x24 and 16x32: +3580 bytes
> 
> I would be fine with adding only the 16x32 version, but still provided
> numbers for the 12x24 in case we decide the size difference is small
> enough for its addition to be worthwhile.
> 

Reply via email to