[Qemu-devel] [PATCH v2 04/12] qsp: use atomic64 accessors

["Emilio G. Cota" <cota@braap.org>]

With the seqlock, we either have to use atomics to remain
within defined behaviour (and note that 64-bit atomics aren't
always guaranteed to compile, irrespective of __nocheck), or
drop the atomics and be in undefined behaviour territory.

Fix it by dropping the seqlock and using atomic64 accessors.
This will limit scalability when !CONFIG_ATOMIC64, but those
machines (1) don't have many users and (2) are unlikely to
have many cores.

- With CONFIG_ATOMIC64:
$ tests/atomic_add-bench -n 1 -m -p
 Throughput:         13.00 Mops/s

- Forcing !CONFIG_ATOMIC64:
$ tests/atomic_add-bench -n 1 -m -p
 Throughput:         10.89 Mops/s

Signed-off-by: Emilio G. Cota <cota@braap.org>
---
 util/qsp.c | 49 ++++++++-----------------------------------------
 1 file changed, 8 insertions(+), 41 deletions(-)

diff --git a/util/qsp.c b/util/qsp.c
index b0c2575d10..c16af03f6a 100644
--- a/util/qsp.c
+++ b/util/qsp.c
@@ -84,13 +84,6 @@ struct QSPEntry {
     uint64_t n_acqs;
     uint64_t ns;
     unsigned int n_objs; /* count of coalesced objs; only used for reporting */
-#ifndef CONFIG_ATOMIC64
-    /*
-     * If we cannot update the counts atomically, then use a seqlock.
-     * We don't need an associated lock because the updates are thread-local.
-     */
-    QemuSeqLock sequence;
-#endif
 };
 typedef struct QSPEntry QSPEntry;
 
@@ -344,47 +337,16 @@ static QSPEntry *qsp_entry_get(const void *obj, const char *file, int line,
     return qsp_entry_find(&qsp_ht, &orig, hash);
 }
 
-/*
- * @from is in the global hash table; read it atomically if the host
- * supports it, otherwise use the seqlock.
- */
-static void qsp_entry_aggregate(QSPEntry *to, const QSPEntry *from)
-{
-#ifdef CONFIG_ATOMIC64
-    to->ns += atomic_read__nocheck(&from->ns);
-    to->n_acqs += atomic_read__nocheck(&from->n_acqs);
-#else
-    unsigned int version;
-    uint64_t ns, n_acqs;
-
-    do {
-        version = seqlock_read_begin(&from->sequence);
-        ns = atomic_read__nocheck(&from->ns);
-        n_acqs = atomic_read__nocheck(&from->n_acqs);
-    } while (seqlock_read_retry(&from->sequence, version));
-
-    to->ns += ns;
-    to->n_acqs += n_acqs;
-#endif
-}
-
 /*
  * @e is in the global hash table; it is only written to by the current thread,
  * so we write to it atomically (as in "write once") to prevent torn reads.
- * If the host doesn't support u64 atomics, use the seqlock.
  */
 static inline void do_qsp_entry_record(QSPEntry *e, int64_t delta, bool acq)
 {
-#ifndef CONFIG_ATOMIC64
-    seqlock_write_begin(&e->sequence);
-#endif
-    atomic_set__nocheck(&e->ns, e->ns + delta);
+    atomic_set_u64(&e->ns, e->ns + delta);
     if (acq) {
-        atomic_set__nocheck(&e->n_acqs, e->n_acqs + 1);
+        atomic_set_u64(&e->n_acqs, e->n_acqs + 1);
     }
-#ifndef CONFIG_ATOMIC64
-    seqlock_write_end(&e->sequence);
-#endif
 }
 
 static inline void qsp_entry_record(QSPEntry *e, int64_t delta)
@@ -550,7 +512,12 @@ static void qsp_aggregate(struct qht *global_ht, void *p, uint32_t h, void *up)
 
     hash = qsp_entry_no_thread_hash(e);
     agg = qsp_entry_find(ht, e, hash);
-    qsp_entry_aggregate(agg, e);
+    /*
+     * The entry is in the global hash table; read from it atomically (as in
+     * "read once").
+     */
+    agg->ns += atomic_read_u64(&e->ns);
+    agg->n_acqs += atomic_read_u64(&e->n_acqs);
 }
 
 static void qsp_iter_diff(struct qht *orig, void *p, uint32_t hash, void *htp)
-- 
2.17.1