[glibc] benchtests: Add pthread-mutex-locks bench

[glibc] benchtests: Add pthread-mutex-locks bench

[H.J. Lu]

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=9e5daa1f6a709dd42420f10971889dc1e47b6e92

commit 9e5daa1f6a709dd42420f10971889dc1e47b6e92
Author: Wangyang Guo <wangyang.guo@intel.com>
Date:   Thu Apr 21 03:28:41 2022 +0000

    benchtests: Add pthread-mutex-locks bench
    
    Benchmark for testing pthread mutex locks performance with different
    threads and critical sections.
    
    The test configuration consists of 3 parts:
    1. thread number
    2. critical-section length
    3. non-critical-section length
    
    Thread number starts from 1 and increased by 2x until num of CPU cores
    (nprocs). An additional over-saturation case (1.25 * nprocs) is also
    included.
    Critical-section is represented by a loop of shared do_filler(),
    length can be determined by the loop iters.
    Non-critical-section is similiar to the critical-section, except it's
    based on non-shared do_filler().
    
    Currently, adaptive pthread_mutex lock is tested.

Diff:
---
 benchtests/Makefile                    |   2 +
 benchtests/bench-pthread-mutex-locks.c | 288 +++++++++++++++++++++++++++++++++
 2 files changed, 290 insertions(+)

diff --git a/benchtests/Makefile b/benchtests/Makefile
index 8dfca592fd..b477042e6c 100644
--- a/benchtests/Makefile
+++ b/benchtests/Makefile
@@ -102,6 +102,7 @@ endif
 
 bench-pthread := \
   pthread-locks \
+  pthread-mutex-locks \
   pthread_once \
   thread_create \
 # bench-pthread
@@ -281,6 +282,7 @@ $(addprefix $(objpfx)bench-,$(math-benchset)): $(libm-benchtests)
 $(addprefix $(objpfx)bench-,$(bench-pthread)): $(thread-library-benchtests)
 $(addprefix $(objpfx)bench-,$(bench-malloc)): $(thread-library-benchtests)
 $(addprefix $(objpfx)bench-,pthread-locks): $(libm-benchtests)
+$(addprefix $(objpfx)bench-,pthread-mutex-locks): $(libm-benchtests)
 
 \f
 
diff --git a/benchtests/bench-pthread-mutex-locks.c b/benchtests/bench-pthread-mutex-locks.c
new file mode 100644
index 0000000000..1685b9dd1f
--- /dev/null
+++ b/benchtests/bench-pthread-mutex-locks.c
@@ -0,0 +1,288 @@
+/* Measure mutex_lock for different threads and critical sections.
+   Copyright (C) 2022 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 "pthread-mutex-locks"
+#define TIMEOUT (20 * 60)
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <math.h>
+#include <pthread.h>
+#include <sys/time.h>
+#include <sys/sysinfo.h>
+#include "bench-timing.h"
+#include "json-lib.h"
+
+static pthread_mutex_t lock;
+static pthread_mutexattr_t attr;
+static pthread_barrier_t barrier;
+
+#define START_ITERS 1000
+
+#pragma GCC push_options
+#pragma GCC optimize(1)
+
+static int __attribute__ ((noinline)) fibonacci (int i)
+{
+  asm("");
+  if (i > 2)
+    return fibonacci (i - 1) + fibonacci (i - 2);
+  return 10 + i;
+}
+
+static void
+do_filler (void)
+{
+  char buf1[512], buf2[512];
+  int f = fibonacci (4);
+  memcpy (buf1, buf2, f);
+}
+
+static void
+do_filler_shared (void)
+{
+  static char buf1[512], buf2[512];
+  int f = fibonacci (4);
+  memcpy (buf1, buf2, f);
+}
+
+#pragma GCC pop_options
+
+#define UNIT_WORK_CRT do_filler_shared ()
+#define UNIT_WORK_NON_CRT do_filler ()
+
+static inline void
+critical_section (int length)
+{
+  for (int i = length; i >= 0; i--)
+    UNIT_WORK_CRT;
+}
+
+static inline void
+non_critical_section (int length)
+{
+  for (int i = length; i >= 0; i--)
+    UNIT_WORK_NON_CRT;
+}
+
+typedef struct Worker_Params
+{
+  long iters;
+  int crt_len;
+  int non_crt_len;
+  timing_t duration;
+} Worker_Params;
+
+static void *
+worker (void *v)
+{
+  timing_t start, stop;
+  Worker_Params *p = (Worker_Params *) v;
+  long iters = p->iters;
+  int crt_len = p->crt_len;
+  int non_crt_len = p->non_crt_len;
+
+  pthread_barrier_wait (&barrier);
+  TIMING_NOW (start);
+  while (iters--)
+    {
+      pthread_mutex_lock (&lock);
+      critical_section (crt_len);
+      pthread_mutex_unlock (&lock);
+      non_critical_section (non_crt_len);
+    }
+  TIMING_NOW (stop);
+
+  TIMING_DIFF (p->duration, start, stop);
+  return NULL;
+}
+
+static double
+do_one_test (int num_threads, int crt_len, int non_crt_len, long iters)
+{
+  int i;
+  timing_t mean;
+  Worker_Params *p, params[num_threads];
+  pthread_t threads[num_threads];
+
+  pthread_mutex_init (&lock, &attr);
+  pthread_barrier_init (&barrier, NULL, num_threads);
+
+  for (i = 0; i < num_threads; i++)
+    {
+      p = &params[i];
+      p->iters = iters;
+      p->crt_len = crt_len;
+      p->non_crt_len = non_crt_len;
+      pthread_create (&threads[i], NULL, worker, (void *) p);
+    }
+  for (i = 0; i < num_threads; i++)
+    pthread_join (threads[i], NULL);
+
+  pthread_mutex_destroy (&lock);
+  pthread_barrier_destroy (&barrier);
+
+  mean = 0;
+  for (i = 0; i < num_threads; i++)
+    mean += params[i].duration;
+  mean /= num_threads;
+  return mean;
+}
+
+#define RUN_COUNT 10
+#define MIN_TEST_SEC 0.01
+
+static void
+do_bench_one (const char *name, int num_threads, int crt_len, int non_crt_len,
+	      json_ctx_t *js)
+{
+  timing_t cur;
+  struct timeval ts, te;
+  double tsd, ted, td;
+  long iters, iters_limit, total_iters;
+  timing_t curs[RUN_COUNT + 2];
+  int i, j;
+  double mean, stdev;
+
+  iters = START_ITERS;
+  iters_limit = LONG_MAX / 100;
+
+  while (1)
+    {
+      gettimeofday (&ts, NULL);
+      cur = do_one_test (num_threads, crt_len, non_crt_len, iters);
+      gettimeofday (&te, NULL);
+      /* Make sure the test to run at least MIN_TEST_SEC.  */
+      tsd = ts.tv_sec + ts.tv_usec / 1000000.0;
+      ted = te.tv_sec + te.tv_usec / 1000000.0;
+      td = ted - tsd;
+      if (td >= MIN_TEST_SEC || iters >= iters_limit)
+	break;
+
+      iters *= 10;
+    }
+
+  curs[0] = cur;
+  for (i = 1; i < RUN_COUNT + 2; i++)
+    curs[i] = do_one_test (num_threads, crt_len, non_crt_len, iters);
+
+  /* Sort the results so we can discard the fastest and slowest
+     times as outliers.  */
+  for (i = 0; i < RUN_COUNT + 1; i++)
+    for (j = i + 1; j < RUN_COUNT + 2; j++)
+      if (curs[i] > curs[j])
+	{
+	  timing_t temp = curs[i];
+	  curs[i] = curs[j];
+	  curs[j] = temp;
+	}
+
+  /* Calculate mean and standard deviation.  */
+  mean = 0.0;
+  total_iters = iters * num_threads;
+  for (i = 1; i < RUN_COUNT + 1; i++)
+    mean += (double) curs[i] / (double) total_iters;
+  mean /= RUN_COUNT;
+
+  stdev = 0.0;
+  for (i = 1; i < RUN_COUNT + 1; i++)
+    {
+      double s = (double) curs[i] / (double) total_iters - mean;
+      stdev += s * s;
+    }
+  stdev = sqrt (stdev / (RUN_COUNT - 1));
+
+  char buf[256];
+  snprintf (buf, sizeof buf, "%s,non_crt_len=%d,crt_len=%d,threads=%d", name,
+	    non_crt_len, crt_len, num_threads);
+
+  json_attr_object_begin (js, buf);
+
+  json_attr_double (js, "duration", (double) cur);
+  json_attr_double (js, "iterations", (double) total_iters);
+  json_attr_double (js, "mean", mean);
+  json_attr_double (js, "stdev", stdev);
+  json_attr_double (js, "min-outlier",
+		    (double) curs[0] / (double) total_iters);
+  json_attr_double (js, "min", (double) curs[1] / (double) total_iters);
+  json_attr_double (js, "max",
+		    (double) curs[RUN_COUNT] / (double) total_iters);
+  json_attr_double (js, "max-outlier",
+		    (double) curs[RUN_COUNT + 1] / (double) total_iters);
+
+  json_attr_object_end (js);
+}
+
+#define TH_CONF_MAX 10
+
+int
+do_bench (void)
+{
+  int rv = 0;
+  json_ctx_t json_ctx;
+  int i, j, k;
+  int th_num, th_conf, nprocs;
+  int threads[TH_CONF_MAX];
+  int crt_lens[] = { 0, 1, 2, 4, 8, 16, 32, 64, 128 };
+  int non_crt_lens[] = { 1, 32, 128 };
+  char name[128];
+
+  json_init (&json_ctx, 2, stdout);
+  json_attr_object_begin (&json_ctx, "pthread_mutex_locks");
+
+  /* The thread config begins from 1, and increases by 2x until nprocs.
+     We also wants to test over-saturation case (1.25*nprocs).  */
+  nprocs = get_nprocs ();
+  th_num = 1;
+  for (th_conf = 0; th_conf < (TH_CONF_MAX - 2) && th_num < nprocs; th_conf++)
+    {
+      threads[th_conf] = th_num;
+      th_num <<= 1;
+    }
+  threads[th_conf++] = nprocs;
+  threads[th_conf++] = nprocs + nprocs / 4;
+
+  pthread_mutexattr_init (&attr);
+  pthread_mutexattr_settype (&attr, PTHREAD_MUTEX_ADAPTIVE_NP);
+  snprintf (name, sizeof name, "type=adaptive");
+
+  for (k = 0; k < (sizeof (non_crt_lens) / sizeof (int)); k++)
+    {
+      int non_crt_len = non_crt_lens[k];
+      for (j = 0; j < (sizeof (crt_lens) / sizeof (int)); j++)
+	{
+	  int crt_len = crt_lens[j];
+	  for (i = 0; i < th_conf; i++)
+	    {
+	      th_num = threads[i];
+	      do_bench_one (name, th_num, crt_len, non_crt_len, &json_ctx);
+	    }
+	}
+    }
+
+  json_attr_object_end (&json_ctx);
+
+  return rv;
+}
+
+#define TEST_FUNCTION do_bench ()
+
+#include "../test-skeleton.c"