[Bug tree-optimization/100363] gcc generating wider load/store than warranted at -O3

[torvalds@linux-foundation.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100363

--- Comment #11 from Linus Torvalds <torvalds@linux-foundation.org> ---
(In reply to Linus Torvalds from comment #10)
> 
>       This particular code comes
> from some old version of zlib, and I can't test because I don't have the ARC
> background to make any sense of the generated code.

Heh. We upgraded to a "recent version" of zlib back in 2006: 

   "Upgrade the zlib_inflate implementation in the kernel from a patched
    version 1.1.3/4 to a patched 1.2.3"

but it turns out that the "do things a 16-bit word at a time" was a
kernel-local optimization for some very slow old PowerPC microcontroller.

The code in upstream zlib actually looks rather better (which is not saying
much, admittedly), doesn't have any 16-bit accesses, and we probably should
just try to synchronize with that instead.