public inbox for libc-alpha@sourceware.org
 help / color / mirror / Atom feed
* [PATCH 1/2] Add random benchmark
@ 2023-11-28 17:35 Wilco Dijkstra
  2023-12-04 14:38 ` Carlos O'Donell
  0 siblings, 1 reply; 4+ messages in thread
From: Wilco Dijkstra @ 2023-11-28 17:35 UTC (permalink / raw)
  To: 'GNU C Library'


Add a simple benchmark to measure the overhead of internal libc locks in
the random() implementation on both single- and multi-threaded cases.

---

diff --git a/benchtests/Makefile b/benchtests/Makefile
index e1346bbda125be9fc2b216f9e8be3f2ee7cb0c4d..1415d20e25798871936f3e2b87f8690517239b21 100644
--- a/benchtests/Makefile
+++ b/benchtests/Makefile
@@ -247,6 +247,7 @@ hash-benchset := \
 stdlib-benchset := \
   arc4random \
   strtod \
+  random-lock \
   # stdlib-benchset
 
 stdio-common-benchset := sprintf
diff --git a/benchtests/bench-random-lock.c b/benchtests/bench-random-lock.c
new file mode 100644
index 0000000000000000000000000000000000000000..ff11f807359c05043366c855751a97a383b1b132
--- /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 40000000
+
+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"


^ permalink raw reply	[flat|nested] 4+ messages in thread
* [PATCH 1/2] Add random benchmark
@ 2024-03-18 15:17 Wilco Dijkstra
  2024-04-16 16:55 ` Adhemerval Zanella Netto
  0 siblings, 1 reply; 4+ messages in thread
From: Wilco Dijkstra @ 2024-03-18 15:17 UTC (permalink / raw)
  To: 'GNU C Library'


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"


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2024-04-16 16:55 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-11-28 17:35 [PATCH 1/2] Add random benchmark Wilco Dijkstra
2023-12-04 14:38 ` Carlos O'Donell
2024-03-18 15:17 Wilco Dijkstra
2024-04-16 16:55 ` Adhemerval Zanella Netto

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).