https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67712
Oleg Endo <olegendo at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2017-08-16
Ever confirmed|0 |1
--- Comment #4 from Oleg Endo <olegendo at gcc dot gnu.org> ---
(In reply to Joseph S. Myers from comment #3)
> The patch committed limits the unrolling, but not necessarily as much as
> desirable (given the bug is about an 8-byte case and the limit is 32 bytes).
The unrolled byte checking tail doesn't seem to make sense at all. It will run
for 4 bytes at most, because the 4-byte loop (at L5 with cmp/str in the above
example) will have already determined that there is a byte mismatch in one of
the 4 bytes.