Use tsd offset_state instead of atomic

While working on #852, I noticed the prng state is atomic.  This is the only
atomic use of prng in all of jemalloc.  Instead, use a threadlocal prng
state if possible to avoid unnecessary cache line contention.

1 files changed, 10 insertions, 0 deletions

diff --git a/src/tsd.c b/src/tsd.c
index f968992..c143068 100644
--- a/src/tsd.c
+++ b/src/tsd.c
@@ -71,6 +71,16 @@ tsd_data_init(tsd_t *tsd) {
 	 */
 	rtree_ctx_data_init(tsd_rtree_ctxp_get_unsafe(tsd));
 
+	/*
+	 * A nondeterministic seed based on the address of tsd reduces
+	 * the likelihood of lockstep non-uniform cache index
+	 * utilization among identical concurrent processes, but at the
+	 * cost of test repeatability.  For debug builds, instead use a
+	 * deterministic seed.
+	 */
+	*tsd_offset_statep_get(tsd) = config_debug ? 0 :
+	    (uint64_t)(uintptr_t)tsd;
+
 	return tsd_tcache_enabled_data_init(tsd);
 }