Bug ID: 30471
Summary: Power 128bit atomics support
Component: new bugs
Right now a
atomicrmw xchg i128* %2, i128 13 acq_rel
get's lowered to
This is problematic to handle for JIT compilers since the family of `__sync`
libcalls are implemented as builtins in clang and gcc. (I have not found a
library that implements them). I am seeking guidance on how a JIT frontend is
supposed to handle the `__sync` libcalls.
Alternatively llvm needs to lower 128bit atomics to the corresponding
definitions in libatomic/compiler-rt (which happens in most other cases.)
You are receiving this mail because:
You are on the CC list for the bug.
llvm-bugs mailing list