https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126114
rsaxvc at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rsaxvc at gmail dot com
--- Comment #4 from rsaxvc at gmail dot com ---
(In reply to William Gibbons from comment #2)
> Thanks. I think this will affect a range of architectures; not just ARM64.
> Peter mentions it also affects x86 (example in the link) and umulh doesn’t
> exist on ARM32.
On ARM32 a software umulh built out of 4x 32x32->64 bit multiply-accumulates is
faster than udiv on the new cores, and far faster than iterative division on
the old cores. See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106484
--- Comment #5 from rsaxvc at gmail dot com ---
(In reply to William Gibbons from comment #2)
> Thanks. I think this will affect a range of architectures; not just ARM64.
> Peter mentions it also affects x86 (example in the link) and umulh doesn’t
> exist on ARM32.
On ARM32 a software umulh built out of 4x 32x32->64 bit multiply-accumulates is
faster than udiv on the new cores, and far faster than iterative division on
the old cores. See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106484
