[Bug rtl-optimization/81423] [6/7/8 Regression] Wrong code at -O2

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81423

--- Comment #6 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Segher Boessenkool from comment #5)
> There are a few issues here.
> 
> 1) I cannot reproduce it: a trunk x86_64-linux compiler stores to
> memory in that insn 20 (the testcase from comment 3).

Just re-checked with todays SVN. There are two (insn 20) RTXes; one in the
function itself, the second one (the problematic one) in the inlined copy in
the main function.