From: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
To: 'GNU C Library' <libc-alpha@sourceware.org>
Subject: [PATCH 1/2] Add random benchmark
Date: Mon, 18 Mar 2024 15:17:01 +0000 [thread overview]
Message-ID: <PAWPR08MB89826762EA450DD0FD209AC9832D2@PAWPR08MB8982.eurprd08.prod.outlook.com> (raw)
Add a simple benchmark to measure the overhead of internal libc locks in
the random() implementation on both single- and multi-threaded cases.
This relies on the implementation of random using internal locks to
access shared global data, and that the runtime uses multi-threaded
locking once a thread has been created (even after it finishes).
OK for commit?
---
diff --git a/benchtests/Makefile b/benchtests/Makefile
index e1346bbda125be9fc2b216f9e8be3f2ee7cb0c4d..1ec14078ab73d7c1c0fa1d4d870a075a66543a5c 100644
--- a/benchtests/Makefile
+++ b/benchtests/Makefile
@@ -246,6 +246,7 @@ hash-benchset := \
stdlib-benchset := \
arc4random \
+ random-lock \
strtod \
# stdlib-benchset
diff --git a/benchtests/bench-random-lock.c b/benchtests/bench-random-lock.c
new file mode 100644
index 0000000000000000000000000000000000000000..29a02ae9ff3a81114e8dd7e1dddcb3309b92df6c
--- /dev/null
+++ b/benchtests/bench-random-lock.c
@@ -0,0 +1,108 @@
+/* Benchmark internal libc locking functions used in random.
+ Copyright (C) 2022-2023 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
+
+#define TEST_MAIN
+#define TEST_NAME "random-lock"
+
+#include <pthread.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include "bench-timing.h"
+#include "json-lib.h"
+
+/* Modern cores run 20M iterations in about 1 second. */
+#define NUM_ITERS 50000000
+
+json_ctx_t json_ctx;
+
+
+/* Measure the overhead of __libc_lock_lock and __libc_lock_unlock by
+ calling random (). */
+static void
+bench_random_lock (size_t iters)
+{
+ timing_t start, stop, total;
+
+ srandom (0);
+
+ /* Warmup to reduce variations due to frequency scaling. */
+ for (int i = 0; i < iters / 4; i++)
+ (void) random ();
+
+ TIMING_NOW (start);
+
+ for (int i = 0; i < iters; i++)
+ (void) random ();
+
+ TIMING_NOW (stop);
+
+ TIMING_DIFF (total, start, stop);
+
+ json_element_double (&json_ctx, (double) total / (double) iters);
+}
+
+static void *
+thread_start (void *p)
+{
+ return p;
+}
+
+int
+do_bench (void)
+{
+ json_init (&json_ctx, 0, stdout);
+
+ json_document_begin (&json_ctx);
+
+ json_attr_string (&json_ctx, "timing_type", TIMING_TYPE);
+ json_attr_object_begin (&json_ctx, "functions");
+ json_attr_object_begin (&json_ctx, "random");
+ json_attr_string (&json_ctx, "bench-variant", "single-threaded");
+ json_array_begin (&json_ctx, "results");
+
+ /* Run benchmark single threaded. */
+ bench_random_lock (NUM_ITERS);
+
+ json_array_end (&json_ctx);
+ json_attr_object_end (&json_ctx);
+
+ json_attr_object_begin (&json_ctx, "random");
+ json_attr_string (&json_ctx, "bench-variant", "multi-threaded");
+ json_array_begin (&json_ctx, "results");
+
+ /* Start a short thread to force SINGLE_THREAD_P == false. This relies on
+ the runtime disabling single-threaded optimizations when multiple
+ threads are used, even after they finish. */
+
+ pthread_t t;
+ pthread_create (&t, NULL, thread_start, NULL);
+ pthread_join (t, NULL);
+
+ /* Repeat benchmark with single-threaded optimizations disabled. */
+ bench_random_lock (NUM_ITERS);
+
+ json_array_end (&json_ctx);
+ json_attr_object_end (&json_ctx);
+ json_attr_object_end (&json_ctx);
+ json_document_end (&json_ctx);
+ return 0;
+}
+
+#define TEST_FUNCTION do_bench ()
+
+#include "../test-skeleton.c"
next reply other threads:[~2024-03-18 15:17 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-03-18 15:17 Wilco Dijkstra [this message]
2024-04-16 16:55 ` [PATCH 1/2] Add random benchmark Adhemerval Zanella Netto
-- strict thread matches above, loose matches on Subject: below --
2023-11-28 17:35 Wilco Dijkstra
2023-12-04 14:38 ` Carlos O'Donell
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://www.gnu.org/software/libc/involved.html
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=PAWPR08MB89826762EA450DD0FD209AC9832D2@PAWPR08MB8982.eurprd08.prod.outlook.com \
--to=wilco.dijkstra@arm.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).