From: Arnaldo Carvalho de Melo <acme@kernel.org>
To: Ingo Molnar <mingo@kernel.org>
Cc: linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org,
Davidlohr Bueso <dave@stgolabs.net>,
Davidlohr Bueso <dbueso@suse.de>,
Kim Phillips <kim.phillips@arm.com>,
Arnaldo Carvalho de Melo <acme@redhat.com>
Subject: [PATCH 10/36] perf bench futex: Use cpumaps
Date: Wed, 6 Dec 2017 11:40:49 -0300 [thread overview]
Message-ID: <20171206144115.15097-11-acme@kernel.org> (raw)
In-Reply-To: <20171206144115.15097-1-acme@kernel.org>
From: Davidlohr Bueso <dave@stgolabs.net>
It was reported that the whole futex bench breaks when dealing with
non-contiguously numbered cpus.
$ echo 0 | sudo tee /sys/devices/system/cpu/cpu3/online
$ ./perf bench futex all
perf: pthread_create: Operation not permitted
Run summary [PID 14934]: 7 threads, each ....
James had implemented an approach with cpumaps that use an in house
flavor. Instead of re-inventing the wheel, I've redone the patch such
that we use the perf's util/cpumap.c interface instead.
Applies to all futex benchmarks.
Suggested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Originally-from: James Yang <james.yang@arm.com>
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Cc: Davidlohr Bueso <dbueso@suse.de>
Cc: Kim Phillips <kim.phillips@arm.com>
Link: http://lkml.kernel.org/r/20171127042101.3659-2-dave@stgolabs.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
tools/perf/bench/futex-hash.c | 19 ++++++++++++-------
tools/perf/bench/futex-lock-pi.c | 23 ++++++++++++++---------
tools/perf/bench/futex-requeue.c | 22 +++++++++++++---------
tools/perf/bench/futex-wake-parallel.c | 24 +++++++++++++++---------
tools/perf/bench/futex-wake.c | 18 +++++++++++-------
5 files changed, 65 insertions(+), 41 deletions(-)
diff --git a/tools/perf/bench/futex-hash.c b/tools/perf/bench/futex-hash.c
index 58ae6ed8f38b..2defb6df7fd0 100644
--- a/tools/perf/bench/futex-hash.c
+++ b/tools/perf/bench/futex-hash.c
@@ -24,6 +24,7 @@
#include <subcmd/parse-options.h>
#include "bench.h"
#include "futex.h"
+#include "cpumap.h"
#include <err.h>
#include <sys/time.h>
@@ -118,11 +119,12 @@ static void print_summary(void)
int bench_futex_hash(int argc, const char **argv)
{
int ret = 0;
- cpu_set_t cpu;
+ cpu_set_t cpuset;
struct sigaction act;
- unsigned int i, ncpus;
+ unsigned int i;
pthread_attr_t thread_attr;
struct worker *worker = NULL;
+ struct cpu_map *cpu;
argc = parse_options(argc, argv, options, bench_futex_hash_usage, 0);
if (argc) {
@@ -130,14 +132,16 @@ int bench_futex_hash(int argc, const char **argv)
exit(EXIT_FAILURE);
}
- ncpus = sysconf(_SC_NPROCESSORS_ONLN);
+ cpu = cpu_map__new(NULL);
+ if (!cpu)
+ goto errmem;
sigfillset(&act.sa_mask);
act.sa_sigaction = toggle_done;
sigaction(SIGINT, &act, NULL);
if (!nthreads) /* default to the number of CPUs */
- nthreads = ncpus;
+ nthreads = cpu->nr;
worker = calloc(nthreads, sizeof(*worker));
if (!worker)
@@ -163,10 +167,10 @@ int bench_futex_hash(int argc, const char **argv)
if (!worker[i].futex)
goto errmem;
- CPU_ZERO(&cpu);
- CPU_SET(i % ncpus, &cpu);
+ CPU_ZERO(&cpuset);
+ CPU_SET(cpu->map[i % cpu->nr], &cpuset);
- ret = pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpu);
+ ret = pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset);
if (ret)
err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
@@ -217,6 +221,7 @@ int bench_futex_hash(int argc, const char **argv)
print_summary();
free(worker);
+ free(cpu);
return ret;
errmem:
err(EXIT_FAILURE, "calloc");
diff --git a/tools/perf/bench/futex-lock-pi.c b/tools/perf/bench/futex-lock-pi.c
index 08653ae8a8c4..8e9c4753e304 100644
--- a/tools/perf/bench/futex-lock-pi.c
+++ b/tools/perf/bench/futex-lock-pi.c
@@ -15,6 +15,7 @@
#include <errno.h>
#include "bench.h"
#include "futex.h"
+#include "cpumap.h"
#include <err.h>
#include <stdlib.h>
@@ -32,7 +33,7 @@ static struct worker *worker;
static unsigned int nsecs = 10;
static bool silent = false, multi = false;
static bool done = false, fshared = false;
-static unsigned int ncpus, nthreads = 0;
+static unsigned int nthreads = 0;
static int futex_flag = 0;
struct timeval start, end, runtime;
static pthread_mutex_t thread_lock;
@@ -113,9 +114,10 @@ static void *workerfn(void *arg)
return NULL;
}
-static void create_threads(struct worker *w, pthread_attr_t thread_attr)
+static void create_threads(struct worker *w, pthread_attr_t thread_attr,
+ struct cpu_map *cpu)
{
- cpu_set_t cpu;
+ cpu_set_t cpuset;
unsigned int i;
threads_starting = nthreads;
@@ -130,10 +132,10 @@ static void create_threads(struct worker *w, pthread_attr_t thread_attr)
} else
worker[i].futex = &global_futex;
- CPU_ZERO(&cpu);
- CPU_SET(i % ncpus, &cpu);
+ CPU_ZERO(&cpuset);
+ CPU_SET(cpu->map[i % cpu->nr], &cpuset);
- if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpu))
+ if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset))
err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
if (pthread_create(&w[i].thread, &thread_attr, workerfn, &worker[i]))
@@ -147,19 +149,22 @@ int bench_futex_lock_pi(int argc, const char **argv)
unsigned int i;
struct sigaction act;
pthread_attr_t thread_attr;
+ struct cpu_map *cpu;
argc = parse_options(argc, argv, options, bench_futex_lock_pi_usage, 0);
if (argc)
goto err;
- ncpus = sysconf(_SC_NPROCESSORS_ONLN);
+ cpu = cpu_map__new(NULL);
+ if (!cpu)
+ err(EXIT_FAILURE, "calloc");
sigfillset(&act.sa_mask);
act.sa_sigaction = toggle_done;
sigaction(SIGINT, &act, NULL);
if (!nthreads)
- nthreads = ncpus;
+ nthreads = cpu->nr;
worker = calloc(nthreads, sizeof(*worker));
if (!worker)
@@ -180,7 +185,7 @@ int bench_futex_lock_pi(int argc, const char **argv)
pthread_attr_init(&thread_attr);
gettimeofday(&start, NULL);
- create_threads(worker, thread_attr);
+ create_threads(worker, thread_attr, cpu);
pthread_attr_destroy(&thread_attr);
pthread_mutex_lock(&thread_lock);
diff --git a/tools/perf/bench/futex-requeue.c b/tools/perf/bench/futex-requeue.c
index 1058c194608a..fc692efa0c05 100644
--- a/tools/perf/bench/futex-requeue.c
+++ b/tools/perf/bench/futex-requeue.c
@@ -22,6 +22,7 @@
#include <errno.h>
#include "bench.h"
#include "futex.h"
+#include "cpumap.h"
#include <err.h>
#include <stdlib.h>
@@ -40,7 +41,7 @@ static bool done = false, silent = false, fshared = false;
static pthread_mutex_t thread_lock;
static pthread_cond_t thread_parent, thread_worker;
static struct stats requeuetime_stats, requeued_stats;
-static unsigned int ncpus, threads_starting, nthreads = 0;
+static unsigned int threads_starting, nthreads = 0;
static int futex_flag = 0;
static const struct option options[] = {
@@ -83,19 +84,19 @@ static void *workerfn(void *arg __maybe_unused)
}
static void block_threads(pthread_t *w,
- pthread_attr_t thread_attr)
+ pthread_attr_t thread_attr, struct cpu_map *cpu)
{
- cpu_set_t cpu;
+ cpu_set_t cpuset;
unsigned int i;
threads_starting = nthreads;
/* create and block all threads */
for (i = 0; i < nthreads; i++) {
- CPU_ZERO(&cpu);
- CPU_SET(i % ncpus, &cpu);
+ CPU_ZERO(&cpuset);
+ CPU_SET(cpu->map[i % cpu->nr], &cpuset);
- if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpu))
+ if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset))
err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
if (pthread_create(&w[i], &thread_attr, workerfn, NULL))
@@ -116,19 +117,22 @@ int bench_futex_requeue(int argc, const char **argv)
unsigned int i, j;
struct sigaction act;
pthread_attr_t thread_attr;
+ struct cpu_map *cpu;
argc = parse_options(argc, argv, options, bench_futex_requeue_usage, 0);
if (argc)
goto err;
- ncpus = sysconf(_SC_NPROCESSORS_ONLN);
+ cpu = cpu_map__new(NULL);
+ if (!cpu)
+ err(EXIT_FAILURE, "cpu_map__new");
sigfillset(&act.sa_mask);
act.sa_sigaction = toggle_done;
sigaction(SIGINT, &act, NULL);
if (!nthreads)
- nthreads = ncpus;
+ nthreads = cpu->nr;
worker = calloc(nthreads, sizeof(*worker));
if (!worker)
@@ -156,7 +160,7 @@ int bench_futex_requeue(int argc, const char **argv)
struct timeval start, end, runtime;
/* create, launch & block all threads */
- block_threads(worker, thread_attr);
+ block_threads(worker, thread_attr, cpu);
/* make sure all threads are already blocked */
pthread_mutex_lock(&thread_lock);
diff --git a/tools/perf/bench/futex-wake-parallel.c b/tools/perf/bench/futex-wake-parallel.c
index b4732dad9f89..4488c27e8a43 100644
--- a/tools/perf/bench/futex-wake-parallel.c
+++ b/tools/perf/bench/futex-wake-parallel.c
@@ -21,6 +21,7 @@
#include <errno.h>
#include "bench.h"
#include "futex.h"
+#include "cpumap.h"
#include <err.h>
#include <stdlib.h>
@@ -43,7 +44,7 @@ static unsigned int nblocked_threads = 0, nwaking_threads = 0;
static pthread_mutex_t thread_lock;
static pthread_cond_t thread_parent, thread_worker;
static struct stats waketime_stats, wakeup_stats;
-static unsigned int ncpus, threads_starting;
+static unsigned int threads_starting;
static int futex_flag = 0;
static const struct option options[] = {
@@ -119,19 +120,20 @@ static void *blocked_workerfn(void *arg __maybe_unused)
return NULL;
}
-static void block_threads(pthread_t *w, pthread_attr_t thread_attr)
+static void block_threads(pthread_t *w, pthread_attr_t thread_attr,
+ struct cpu_map *cpu)
{
- cpu_set_t cpu;
+ cpu_set_t cpuset;
unsigned int i;
threads_starting = nblocked_threads;
/* create and block all threads */
for (i = 0; i < nblocked_threads; i++) {
- CPU_ZERO(&cpu);
- CPU_SET(i % ncpus, &cpu);
+ CPU_ZERO(&cpuset);
+ CPU_SET(cpu->map[i % cpu->nr], &cpuset);
- if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpu))
+ if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset))
err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
if (pthread_create(&w[i], &thread_attr, blocked_workerfn, NULL))
@@ -205,6 +207,7 @@ int bench_futex_wake_parallel(int argc, const char **argv)
struct sigaction act;
pthread_attr_t thread_attr;
struct thread_data *waking_worker;
+ struct cpu_map *cpu;
argc = parse_options(argc, argv, options,
bench_futex_wake_parallel_usage, 0);
@@ -217,9 +220,12 @@ int bench_futex_wake_parallel(int argc, const char **argv)
act.sa_sigaction = toggle_done;
sigaction(SIGINT, &act, NULL);
- ncpus = sysconf(_SC_NPROCESSORS_ONLN);
+ cpu = cpu_map__new(NULL);
+ if (!cpu)
+ err(EXIT_FAILURE, "calloc");
+
if (!nblocked_threads)
- nblocked_threads = ncpus;
+ nblocked_threads = cpu->nr;
/* some sanity checks */
if (nwaking_threads > nblocked_threads || !nwaking_threads)
@@ -259,7 +265,7 @@ int bench_futex_wake_parallel(int argc, const char **argv)
err(EXIT_FAILURE, "calloc");
/* create, launch & block all threads */
- block_threads(blocked_worker, thread_attr);
+ block_threads(blocked_worker, thread_attr, cpu);
/* make sure all threads are already blocked */
pthread_mutex_lock(&thread_lock);
diff --git a/tools/perf/bench/futex-wake.c b/tools/perf/bench/futex-wake.c
index 8c5c0b6b5c97..e8181ad7d088 100644
--- a/tools/perf/bench/futex-wake.c
+++ b/tools/perf/bench/futex-wake.c
@@ -22,6 +22,7 @@
#include <errno.h>
#include "bench.h"
#include "futex.h"
+#include "cpumap.h"
#include <err.h>
#include <stdlib.h>
@@ -89,19 +90,19 @@ static void print_summary(void)
}
static void block_threads(pthread_t *w,
- pthread_attr_t thread_attr)
+ pthread_attr_t thread_attr, struct cpu_map *cpu)
{
- cpu_set_t cpu;
+ cpu_set_t cpuset;
unsigned int i;
threads_starting = nthreads;
/* create and block all threads */
for (i = 0; i < nthreads; i++) {
- CPU_ZERO(&cpu);
- CPU_SET(i % ncpus, &cpu);
+ CPU_ZERO(&cpuset);
+ CPU_SET(cpu->map[i % cpu->nr], &cpuset);
- if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpu))
+ if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset))
err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
if (pthread_create(&w[i], &thread_attr, workerfn, NULL))
@@ -122,6 +123,7 @@ int bench_futex_wake(int argc, const char **argv)
unsigned int i, j;
struct sigaction act;
pthread_attr_t thread_attr;
+ struct cpu_map *cpu;
argc = parse_options(argc, argv, options, bench_futex_wake_usage, 0);
if (argc) {
@@ -129,7 +131,9 @@ int bench_futex_wake(int argc, const char **argv)
exit(EXIT_FAILURE);
}
- ncpus = sysconf(_SC_NPROCESSORS_ONLN);
+ cpu = cpu_map__new(NULL);
+ if (!cpu)
+ err(EXIT_FAILURE, "calloc");
sigfillset(&act.sa_mask);
act.sa_sigaction = toggle_done;
@@ -161,7 +165,7 @@ int bench_futex_wake(int argc, const char **argv)
struct timeval start, end, runtime;
/* create, launch & block all threads */
- block_threads(worker, thread_attr);
+ block_threads(worker, thread_attr, cpu);
/* make sure all threads are already blocked */
pthread_mutex_lock(&thread_lock);
--
2.13.6
next prev parent reply other threads:[~2017-12-06 14:40 UTC|newest]
Thread overview: 46+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20171206144115.15097-1-acme@kernel.org>
2017-12-06 14:40 ` [PATCH 01/36] tools headers: Follow the upstream UAPI header version 100% differ from the kernel Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 02/36] perf test: Disable test cases 19 and 20 on s390x Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 03/36] perf record: Synthesize unit/scale/... in event update Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 04/36] perf record: Synthesize thread map and cpu map Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 05/36] perf script: Allow computing 'perf stat' style metrics Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 06/36] perf buildid-cache: Document for Node.js USDT Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 07/36] perf report: Fix -D output for user metadata events Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 08/36] perf intel-pt: Improve build messages for files that differ from the kernel Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 09/36] Documentation: Add Arnaldo Melo to list of enforcement statement endorsers Arnaldo Carvalho de Melo
2017-12-06 14:40 ` Arnaldo Carvalho de Melo [this message]
2017-12-06 14:40 ` [PATCH 11/36] tools build feature: Check if pthread_barrier_t is available Arnaldo Carvalho de Melo
2017-12-06 21:31 ` Philippe Ombredanne
2017-12-07 11:24 ` Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 12/36] perf bench futex: Sync waker threads Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 13/36] perf annotate: Fix unnecessary memory allocation for s390x Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 14/36] perf annotate: Fix objdump comment parsing for Intel mov dissassembly Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 15/36] perf rblist: Create rblist__exit() function Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 16/36] perf stat: Add rbtree node_delete op Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 17/36] perf thread_map: Add method to map all threads in the system Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 18/36] perf s390: Always build with -fPIC Arnaldo Carvalho de Melo
2017-12-07 8:09 ` Hendrik Brueckner
2017-12-28 15:34 ` [tip:perf/core] " tip-bot for Hendrik Brueckner
2017-12-06 14:40 ` [PATCH 19/36] perf pmu: Pass pmu as a parameter to get_cpuid_str() Arnaldo Carvalho de Melo
2017-12-06 14:40 ` Arnaldo Carvalho de Melo
2017-12-06 14:40 ` [PATCH 20/36] perf tools arm64: Add support for get_cpuid_str function Arnaldo Carvalho de Melo
2017-12-06 14:40 ` Arnaldo Carvalho de Melo
2017-12-06 14:40 ` Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 21/36] perf pmu: Add helper function is_pmu_core to detect PMU CORE devices Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 22/36] perf vendor events arm64: Add ThunderX2 implementation defined pmu core events Arnaldo Carvalho de Melo
2017-12-06 14:41 ` Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 23/36] perf pmu: Add check for valid cpuid in perf_pmu__find_map() Arnaldo Carvalho de Melo
2017-12-06 14:41 ` Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 24/36] perf tools: Fix up build in hardnened environments Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 25/36] perf evlist: Remove 'overwrite' parameter from perf_evlist__mmap Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 26/36] perf evlist: Remove 'overwrite' parameter from perf_evlist__mmap_ex Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 27/36] perf evlist: Remove evlist->overwrite Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 28/36] perf mmap: Remove overwrite from arguments list of perf_mmap__push Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 29/36] perf mmap: Remove overwrite and check_messup from mmap read Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 30/36] perf c2c: Add a tip about cacheline events Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 31/36] perf vendor events: Use more flexible pattern matching for CPU identification for mapfile.csv Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 32/36] x86/asm: Allow again using asm.h when building for the 'bpf' clang target Arnaldo Carvalho de Melo
2017-12-06 14:41 ` Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 33/36] perf report: Set browser mode right before setup_browser() Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 34/36] perf mmap: Fix perf backward recording Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 35/36] perf mmap: Don't discard prev in backward mode Arnaldo Carvalho de Melo
2017-12-06 14:41 ` [PATCH 36/36] perf tools: Rename 'backward' to 'overwrite' in evlist, mmap and record Arnaldo Carvalho de Melo
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
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20171206144115.15097-11-acme@kernel.org \
--to=acme@kernel.org \
--cc=acme@redhat.com \
--cc=dave@stgolabs.net \
--cc=dbueso@suse.de \
--cc=kim.phillips@arm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-perf-users@vger.kernel.org \
--cc=mingo@kernel.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.