https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125921
Kael Franco <kaelfandrew at gmail dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
See Also| |https://github.com/llvm/llv
| |m-project/pull/206242#issue
| |comment-4817634815
Summary|Missed size optimization of |Missed target optimization
|MAX/MIN (x, -1) |of MAX/MIN (x, -1)
--- Comment #2 from Kael Franco <kaelfandrew at gmail dot com> ---
https://godbolt.org/z/eKhbd7rK4
With -O3 -march=novalake on GCC 16.1,
f2 uses one less instruction and smaller than f.
https://godbolt.org/z/feWx9G4o5
With -O3 -march=armv9.2-a on ARM64 GCC 16.1,
f2 and i2 uses one less instruction and is smaller than f/i.
