https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86005
--- Comment #3 from Alex Bradbury <asb at lowrisc dot org> --- (In reply to Andrew Waterman from comment #2) > I realize the documentation doesn't concur with me, but as long as gcc > and libgcc agree on the lock-freeness of the routines, I don't see the > harm. (wrt. rv32ia, at least.) Yes, for RV32IA it might be allowable - assuming you're able for the compiler to make the assumption that it knows that property of the __atomic_* implementation and that assumption won't/can't be invalidated by linking in another implementation which isn't lock-free.
