To get consistent benchmarking results, sometimes we need to set the
sched_policy/priority/mask of the workload to reduce system noise.
For example, CPU binding is required on big.little system.
$ perf stat -- taskset -c 0 ls
Nevertheless, the 'taskset' is also counted here.
To get away of the middleman, this adds a new option '--workload-attr' to
do the same jobs for stat and record commands.
$ sudo perf stat --workload-attr fifo,40,0-3:7 -- ls
Above will make 'ls' run on CPU #0-#3 and #7 with fifo scheduler and
realtime priority is 40.
Signed-off-by: Changbin Du <[email protected]>
---
v2: Use cpu list spec instead of cpu mask number.
---
tools/perf/Documentation/perf-record.txt | 7 ++
tools/perf/Documentation/perf-stat.txt | 6 ++
tools/perf/builtin-record.c | 26 +++++
tools/perf/builtin-stat.c | 18 ++++
tools/perf/util/evlist.c | 117 +++++++++++++++++++++++
tools/perf/util/evlist.h | 3 +
tools/perf/util/target.h | 9 ++
7 files changed, 186 insertions(+)
diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
index 680396c56bd1..9c01076f6c89 100644
--- a/tools/perf/Documentation/perf-record.txt
+++ b/tools/perf/Documentation/perf-record.txt
@@ -838,6 +838,13 @@ filtered through the mask provided by -C option.
only, as of now. So the applications built without the frame
pointer might see bogus addresses.
+--workload-attr <sched_policy[,priority][,cpu-list]>::
+ setup target workload (the <command>) attributes:
+
+ sched_policy: other|fifo|rr|batch|idle
+ priority: scheduling priority for fifo|rr, nice value for other
+ cpu-list: CPU affinity. e.g. 1-2:4 is processors #1, #2, and #4
+
include::intel-hybrid.txt[]
SEE ALSO
diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index 8f789fa1242e..18d0f73458b1 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -262,6 +262,12 @@ disable events during measurements:
wait -n ${perf_pid}
exit $?
+--workload-attr <sched_policy[,priority][,cpu-list]>::
+ setup target workload (the <command>) attributes:
+
+ sched_policy: other|fifo|rr|batch|idle
+ priority: scheduling priority for fifo|rr, nice value for other
+ cpu-list: CPU affinity. e.g. 1-2:4 is processors #1, #2, and #4
--pre::
--post::
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index aec18db7ff23..c16e230afcd5 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -3289,6 +3289,17 @@ static int parse_record_synth_option(const struct option *opt,
return 0;
}
+static int record_parse_workload_attr_opt(const struct option *opt,
+ const char *arg,
+ int unset __maybe_unused)
+{
+ struct record_opts *opts = opt->value;
+
+ return evlist__parse_workload_attr(arg, &opts->target.workload.sched_policy,
+ &opts->target.workload.sched_priority,
+ &opts->target.workload.cpu_map);
+}
+
/*
* XXX Ideally would be local to cmd_record() and passed to a record__new
* because we need to have access to it in record__exit, that is called
@@ -3309,6 +3320,8 @@ static struct record record = {
.target = {
.uses_mmap = true,
.default_per_cpu = true,
+ .workload.sched_policy = -1,
+ .workload.sched_priority = 0,
},
.mmap_flush = MMAP_FLUSH_DEFAULT,
.nr_threads_synthesize = 1,
@@ -3333,6 +3346,12 @@ static struct record record = {
const char record_callchain_help[] = CALLCHAIN_RECORD_HELP
"\n\t\t\t\tDefault: fp";
+const char record_workload_attr_help[] =
+ "setup target workload (the <command>) attributes:\n\n"
+ HELP_PAD "sched_policy: other|fifo|rr|batch|idle\n"
+ HELP_PAD "priority: scheduling priority for fifo|rr, nice value for other\n"
+ HELP_PAD "cpu-list: CPU affinity. e.g. 1-2:4 is processors #1, #2, and #4";
+
static bool dry_run;
static struct parse_events_option_args parse_events_option_args = {
@@ -3551,6 +3570,9 @@ static struct option __record_options[] = {
"write collected trace data into several data files using parallel threads",
record__parse_threads),
OPT_BOOLEAN(0, "off-cpu", &record.off_cpu, "Enable off-cpu analysis"),
+ OPT_CALLBACK(0, "workload-attr", &record.opts,
+ "sched_policy[,priority][,cpu-list]", record_workload_attr_help,
+ &record_parse_workload_attr_opt),
OPT_END()
};
@@ -4266,6 +4288,10 @@ int cmd_record(int argc, const char **argv)
record__free_thread_masks(rec, rec->nr_threads);
rec->nr_threads = 0;
evlist__close_control(rec->opts.ctl_fd, rec->opts.ctl_fd_ack, &rec->opts.ctl_fd_close);
+ if (rec->opts.target.workload.cpu_map) {
+ perf_cpu_map__put(rec->opts.target.workload.cpu_map);
+ rec->opts.target.workload.cpu_map = NULL;
+ }
return err;
}
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 07b48f6df48e..bb755cb44103 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -108,6 +108,8 @@ static bool all_counters_use_bpf = true;
static struct target target = {
.uid = UINT_MAX,
+ .workload.sched_policy = -1,
+ .workload.sched_priority = 0,
};
#define METRIC_ONLY_LEN 20
@@ -1160,6 +1162,14 @@ static int parse_cache_level(const struct option *opt,
return 0;
}
+static int parse_workload_attr_opt(const struct option *opt __maybe_unused, const char *arg,
+ int unset __maybe_unused)
+{
+ return evlist__parse_workload_attr(arg, &target.workload.sched_policy,
+ &target.workload.sched_priority,
+ &target.workload.cpu_map);
+}
+
static struct option stat_options[] = {
OPT_BOOLEAN('T', "transaction", &transaction_run,
"hardware transaction statistics"),
@@ -1220,6 +1230,9 @@ static struct option stat_options[] = {
OPT_BOOLEAN(0, "append", &append_file, "append to the output file"),
OPT_INTEGER(0, "log-fd", &output_fd,
"log output to fd, instead of stderr"),
+ OPT_CALLBACK(0, "workload-attr", &stat_config,
+ "sched_policy[,priority][,cpu-list]", record_workload_attr_help,
+ &parse_workload_attr_opt),
OPT_STRING(0, "pre", &pre_cmd, "command",
"command to run prior to the measured command"),
OPT_STRING(0, "post", &post_cmd, "command",
@@ -2893,5 +2906,10 @@ int cmd_stat(int argc, const char **argv)
metricgroup__rblist_exit(&stat_config.metric_events);
evlist__close_control(stat_config.ctl_fd, stat_config.ctl_fd_ack, &stat_config.ctl_fd_close);
+ if (target.workload.cpu_map) {
+ perf_cpu_map__put(target.workload.cpu_map);
+ target.workload.cpu_map = NULL;
+ }
+
return status;
}
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 7ef43f72098e..cbdc2184f8a1 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -46,6 +46,7 @@
#include <sys/mman.h>
#include <sys/prctl.h>
#include <sys/timerfd.h>
+#include <sys/resource.h>
#include <linux/bitops.h>
#include <linux/hash.h>
@@ -58,6 +59,7 @@
#include <perf/evsel.h>
#include <perf/cpumap.h>
#include <perf/mmap.h>
+#include <perf/cpumap.h>
#include <internal/xyarray.h>
@@ -1398,6 +1400,118 @@ int evlist__open(struct evlist *evlist)
return err;
}
+int evlist__parse_workload_attr(const char *str, int *sched_policy, int *sched_priority,
+ struct perf_cpu_map **cpu_map)
+{
+ char *tok, *saveptr = NULL;
+ char *buf;
+ int ret = -1;
+
+ /* We need buffer that we know we can write to. */
+ buf = strdup(str);
+ if (!buf)
+ return -ENOMEM;
+
+ tok = strtok_r((char *)buf, ",", &saveptr);
+ do {
+ /* sched policy */
+ if (!strncmp(tok, "other", sizeof("other")))
+ *sched_policy = SCHED_OTHER;
+ else if (!strncmp(tok, "fifo", sizeof("fifo"))) {
+ *sched_policy = SCHED_FIFO;
+ /* default to lowest priority */
+ *sched_priority = 99;
+ } else if (!strncmp(tok, "rr", sizeof("rr"))) {
+ *sched_policy = SCHED_RR;
+ *sched_priority = 99;
+ } else if (!strncmp(tok, "batch", sizeof("batch")))
+ *sched_policy = SCHED_BATCH;
+ else if (!strncmp(tok, "idle", sizeof("idle")))
+ *sched_policy = SCHED_IDLE;
+ else {
+ pr_err("workload_attr: unknown sched policy %s\n", tok);
+ break;
+ }
+
+ /* sched priority */
+ tok = strtok_r(NULL, ",", &saveptr);
+ if (tok) {
+ int priority;
+ char *endptr;
+
+ priority = strtol(tok, &endptr, 0);
+ if (*endptr) {
+ pr_err("workload_attr: invalid sched priority %s\n", tok);
+ break;
+ }
+
+ if (*sched_policy == SCHED_FIFO || *sched_policy == SCHED_RR) {
+ if (priority < 1 || priority > 99) {
+ pr_err("workload_attr: invalid priority %d for fifo and rr, allowed [1,99]\n",
+ priority);
+ break;
+ }
+ }
+ *sched_priority = priority;
+ }
+
+ /* cpu list */
+ tok = strtok_r(NULL, ",", &saveptr);
+ if (tok) {
+ *cpu_map = __perf_cpu_map__new(tok, ':');
+ if (!*cpu_map) {
+ pr_err("workload_attr: failed to get cpus map from %s\n", tok);
+ break;
+ }
+ }
+ ret = 0;
+ } while (0);
+
+ free(buf);
+ return ret;
+}
+
+static int setup_workload_attr(struct target *target)
+{
+ struct sched_param param;
+ int policy = target->workload.sched_policy;
+ int priority = target->workload.sched_priority;
+
+ if (policy >= 0) {
+ param.sched_priority = (policy == SCHED_FIFO || policy == SCHED_RR) ?
+ priority : 0;
+ if (sched_setscheduler(0, policy, ¶m) != 0) {
+ perror("failed to set the sched policy");
+ return -1;
+ }
+
+ if (policy == SCHED_OTHER) {
+ if (setpriority(PRIO_PROCESS, 0, priority) != 0) {
+ perror("failed to set the nice value");
+ return -1;
+ }
+ }
+ }
+
+ if (target->workload.cpu_map) {
+ size_t cpuset_size = -1;
+ cpu_set_t *cpu_set;
+
+ cpu_set = perf_cpu_map__2_cpuset(target->workload.cpu_map, &cpuset_size);
+ if (!cpu_set)
+ return -1;
+
+ if (sched_setaffinity(0, cpuset_size, cpu_set) != 0) {
+ perror("failed to set the sched affinity");
+ CPU_FREE(cpu_set);
+ return -1;
+ }
+ CPU_FREE(cpu_set);
+ }
+
+ return 0;
+}
+
int evlist__prepare_workload(struct evlist *evlist, struct target *target, const char *argv[],
bool pipe_output, void (*exec_error)(int signo, siginfo_t *info, void *ucontext))
{
@@ -1464,6 +1578,9 @@ int evlist__prepare_workload(struct evlist *evlist, struct target *target, const
exit(ret);
}
+ if (setup_workload_attr(target) != 0)
+ exit(-1);
+
execvp(argv[0], (char **)argv);
if (exec_error) {
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index 664c6bf7b3e0..761ae102dda6 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -15,6 +15,7 @@
#include <pthread.h>
#include <signal.h>
#include <unistd.h>
+#include <sched.h>
struct pollfd;
struct thread_map;
@@ -180,6 +181,8 @@ void evlist__set_id_pos(struct evlist *evlist);
void evlist__config(struct evlist *evlist, struct record_opts *opts, struct callchain_param *callchain);
int record_opts__config(struct record_opts *opts);
+int evlist__parse_workload_attr(const char *str, int *sched_policy, int *sched_priority,
+ struct perf_cpu_map **cpu_set);
int evlist__prepare_workload(struct evlist *evlist, struct target *target,
const char *argv[], bool pipe_output,
void (*exec_error)(int signo, siginfo_t *info, void *ucontext));
diff --git a/tools/perf/util/target.h b/tools/perf/util/target.h
index d582cae8e105..16fe33dda2ef 100644
--- a/tools/perf/util/target.h
+++ b/tools/perf/util/target.h
@@ -4,6 +4,7 @@
#include <stdbool.h>
#include <sys/types.h>
+#include <sched.h>
struct target {
const char *pid;
@@ -19,6 +20,12 @@ struct target {
bool use_bpf;
int initial_delay;
const char *attr_map;
+
+ struct {
+ int sched_policy;
+ int sched_priority;
+ struct perf_cpu_map *cpu_map;
+ } workload;
};
enum target_errno {
@@ -103,4 +110,6 @@ static inline bool target__uses_dummy_map(struct target *target)
return use_dummy;
}
+extern const char record_workload_attr_help[];
+
#endif /* _PERF_TARGET_H */
--
2.25.1