atomics: do not emit consume barrier for atomic_rcu_read

Currently we emit a consume-load in atomic_rcu_read. Because of
limitations in current compilers, this is overkill for non-Alpha hosts
and it is only useful to make Thread Sanitizer work.

This patch leaves the consume-load in atomic_rcu_read when
compiling with Thread Sanitizer enabled, and resorts to a
relaxed load + smp_read_barrier_depends otherwise.

On an RMO host architecture, such as aarch64, the performance
improvement of this change is easily measurable. For instance,
qht-bench performs an atomic_rcu_read on every lookup. Performance
before and after applying this patch:

$ tests/qht-bench -d 5 -n 1
Before: 9.78 MT/s
After: 10.96 MT/s

Backports commit 15487aa132109891482f79d78a30d6cfd465a391 from qemu
This commit is contained in:
Emilio G. Cota 2018-02-24 03:28:09 -05:00 committed by Lioncash
parent 87ef2a2c5f
commit ab569f5cde
No known key found for this signature in database
GPG key ID: 4E3C3CC1031BA9C7

View file

@ -70,12 +70,22 @@ void _ReadWriteBarrier(void);
__atomic_store(ptr, &_val, __ATOMIC_RELAXED); \
} while(0)
/* Atomic RCU operations imply weak memory barriers */
/* See above: most compilers currently treat consume and acquire the
* same, but this slows down atomic_rcu_read unnecessarily.
*/
#ifdef __SANITIZE_THREAD__
#define atomic_rcu_read__nocheck(ptr, valptr) \
__atomic_load(ptr, valptr, __ATOMIC_CONSUME);
#else
#define atomic_rcu_read__nocheck(ptr, valptr) \
__atomic_load(ptr, valptr, __ATOMIC_RELAXED); \
smp_read_barrier_depends();
#endif
#define atomic_rcu_read(ptr) \
({ \
typeof(*ptr) _val; \
__atomic_load(ptr, &_val, __ATOMIC_CONSUME); \
atomic_rcu_read__nocheck(ptr, &_val); \
_val; \
})