Re: [PATCH 06/10] btrfs: uninline btrfs_bg_flags_to_raid_index

[David Sterba <dsterba@suse.cz>]

On Mon, Jul 26, 2021 at 08:34:20PM +0800, Qu Wenruo wrote:
> 
> 
> On 2021/7/26 下午8:15, David Sterba wrote:
> > The helper does a simple translation from block group flags to index to
> > the btrfs_raid_array table. There's no apparent reason to inline the
> > function, the translation happens usually once per function and is not
> > called in a loop.
> >
> > Making it a proper function saves quite some binary code (x86_64,
> > release config):
> >
> >     text    data     bss     dec     hex filename
> > 1164011   19253   14912 1198176  124860 pre/btrfs.ko
> > 1161559   19253   14912 1195724  123ecc post/btrfs.ko
> >
> > DELTA: -2451
> 
> My memory says there used to be some option to allow the compiler to
> uninline some functions, but I can't find it in the latest kernel.

Inlining is the compiler magic, the inline annotations are taken as
hints (that's what compiler people say) but they mostly work as written
in the source.

> It looks to me that this should really be something dependent on kernel
> config/compiler optimization.

Leaving the optimizations to the compiler is a good idea in general,
recompiling the same sources on a newer CPU target can lead to better
code compared to some hand-crafted code that may prevent the
optimizations. The one I uninlined here was an obvious "too big for
inline" and it was not a hot path. I'd rather remove such obvious cases
as we see them, the size and scope of the change is measurable.

> E.g. to allow -Os optimization to uninline such functions.

IIRC -Os is not recommended for kernel as it could generate worse code.
Also I don't think it's not a common default option so we'd have to live
with the worse code for most builds.