From: Willy Tarreau <w@1wt.eu>
Date: Mon, 11 Aug 2025 15:43:51 +0000 (+0200)
Subject: OPTIM: server: start to use aligned allocs in server
X-Git-Tag: v3.3-dev7~31
X-Git-Url: http://git.ipfire.org/gitweb/?a=commitdiff_plain;h=refs%2Fheads%2Fquic-interop;p=thirdparty%2Fhaproxy.git

OPTIM: server: start to use aligned allocs in server

This is currently for per-thread arrays like idle conns etc. We're
now cache-aligning the per-thread arrays so as to put an end to false
sharing. A comparative test between no alignment and alignment on a
simple config with round robin between 4 servers showed an average
rate of 1.75M/s vs 1.72M/s before for 100M requests. The gain seems
to be more commonly less than 1% however. This should mostly help
make measurements more reproducible across multiple runs.
---

diff --git a/src/server.c b/src/server.c
index e53d4847c..c299d202b 100644
--- a/src/server.c
+++ b/src/server.c
@@ -3122,9 +3122,9 @@ void srv_free_params(struct server *srv)
 	free(srv->hostname);
 	free(srv->hostname_dn);
 	free((char*)srv->conf.file);
-	free(srv->per_thr);
-	free(srv->per_tgrp);
-	free(srv->curr_idle_thr);
+	ha_aligned_free(srv->per_thr);
+	ha_aligned_free(srv->per_tgrp);
+	ha_aligned_free(srv->curr_idle_thr);
 	free(srv->pool_conn_name);
 	release_sample_expr(srv->pool_conn_name_expr);
 	free(srv->resolvers_id);
@@ -3482,7 +3482,7 @@ int srv_init(struct server *srv)
 
 	/* initialize idle conns lists */
 	if (srv->max_idle_conns != 0) {
-		srv->curr_idle_thr = calloc(global.nbthread, sizeof(*srv->curr_idle_thr));
+		srv->curr_idle_thr = ha_aligned_zalloc(64, global.nbthread * sizeof(*srv->curr_idle_thr));
 		if (!srv->curr_idle_thr) {
 			ha_alert("memory error during idle conn list init for %s/%s server\n",
 			         srv->proxy->id, srv->id);
@@ -5918,8 +5918,8 @@ static int srv_init_per_thr(struct server *srv)
 {
 	int i;
 
-	srv->per_thr = calloc(global.nbthread, sizeof(*srv->per_thr));
-	srv->per_tgrp = calloc(global.nbtgroups, sizeof(*srv->per_tgrp));
+	srv->per_thr = ha_aligned_zalloc(64, global.nbthread * sizeof(*srv->per_thr));
+	srv->per_tgrp = ha_aligned_zalloc(64, global.nbtgroups * sizeof(*srv->per_tgrp));
 	if (!srv->per_thr || !srv->per_tgrp)
 		return -1;