https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63672
Peter Bergner <bergner at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |bergner at gcc dot gnu.org
--- Comment #1 from Peter Bergner <bergner at gcc dot gnu.org> ---
Andi, did this bug ever get fixed? I ask, since we're hitting a similar issue
on POWER with our __builtin_tbegin() builtin. I know we need a minimum of a
memory barrier on our HTM builtins, but I also wonder whether a full
optimization barrier is needed to stop even non loads and stores from being
moved outside of the transaction body. The following shows a test case that
seems to need a full optimization barrier:
https://www.sourceware.org/ml/libc-alpha/2015-08/msg00880.html
