[Bug translation/80958] [8 regression] gcc.target/i386/pr70021.c FAILs

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80958

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Note the loop in question does very sparse accesses which have high load/store
cost (in the attached dump), it also contains conditionals so the more precise
accounting might have just pushed the cost over the boundary, not changing it
by much.