https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95751
Bug ID: 95751
Summary: [aarch64] Consider using ldapr for
__atomic_load_n(acquire) on ARMv8.3-RCPC
Product: gcc
Version: 10.1.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: andysem at mail dot ru
Target Milestone: ---
ARMv8.3-RCPC extension adds new ldapr[b/h] instructions for Load-AcquirePC
semantics that better matches C++ memory order semantics.
In particular, the core ldar instructions will wait until the previous stlr,
even on another address, completes on the same thread. The newer ldapr does not
wait for the previous stlr to complete, unless it is performed on the same
address.
Consider using these new instructions in __atomic_load_n(consume/acquire) when
-march=armv8.3 or higher is specified. Note that __atomic_load_n(relaxed) and
__atomic_load_n(seq_cst) should stay unchanged.
