*
* Sample output:
- $ perf stat ~/hackbench 10
- Time: 0.104
+ $ perf stat ./hackbench 10
- Performance counter stats for '/home/mingo/hackbench':
+ Time: 0.118
- 1255.538611 task clock ticks # 10.143 CPU utilization factor
- 54011 context switches # 0.043 M/sec
- 385 CPU migrations # 0.000 M/sec
- 17755 pagefaults # 0.014 M/sec
- 3808323185 CPU cycles # 3033.219 M/sec
- 1575111190 instructions # 1254.530 M/sec
- 17367895 cache references # 13.833 M/sec
- 7674421 cache misses # 6.112 M/sec
+ Performance counter stats for './hackbench 10':
- Wall-clock time elapsed: 123.786620 msecs
+ 1708.761321 task-clock # 11.037 CPUs utilized
+ 41,190 context-switches # 0.024 M/sec
+ 6,735 CPU-migrations # 0.004 M/sec
+ 17,318 page-faults # 0.010 M/sec
+ 5,205,202,243 cycles # 3.046 GHz
+ 3,856,436,920 stalled-cycles-frontend # 74.09% frontend cycles idle
+ 1,600,790,871 stalled-cycles-backend # 30.75% backend cycles idle
+ 2,603,501,247 instructions # 0.50 insns per cycle
+ # 1.48 stalled cycles per insn
+ 484,357,498 branches # 283.455 M/sec
+ 6,388,934 branch-misses # 1.32% of all branches
+
+ 0.154822978 seconds time elapsed
*
- * Copyright (C) 2008, Red Hat Inc, Ingo Molnar <mingo@redhat.com>
+ * Copyright (C) 2008-2011, Red Hat Inc, Ingo Molnar <mingo@redhat.com>
*
* Improvements and fixes by:
*
#include "util/parse-options.h"
#include "util/parse-events.h"
#include "util/event.h"
+#include "util/evlist.h"
+#include "util/evsel.h"
#include "util/debug.h"
+#include "util/color.h"
+#include "util/header.h"
+#include "util/cpumap.h"
+#include "util/thread.h"
+#include "util/thread_map.h"
#include <sys/prctl.h>
#include <math.h>
+#include <locale.h>
+
+#define DEFAULT_SEPARATOR " "
+#define CNTR_NOT_SUPPORTED "<not supported>"
+#define CNTR_NOT_COUNTED "<not counted>"
-static struct perf_counter_attr default_attrs[] = {
+static struct perf_event_attr default_attrs[] = {
- { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK },
- { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CONTEXT_SWITCHES},
- { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CPU_MIGRATIONS },
- { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS },
+ { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK },
+ { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CONTEXT_SWITCHES },
+ { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CPU_MIGRATIONS },
+ { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS },
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES },
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS },
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES},
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES },
+ { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES },
+ { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_FRONTEND },
+ { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_BACKEND },
+ { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS },
+ { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
+ { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES },
};
-#define MAX_RUN 100
+/*
+ * Detailed stats (-d), covering the L1 and last level data caches:
+ */
+static struct perf_event_attr detailed_attrs[] = {
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_L1D << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_L1D << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_LL << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_LL << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
+};
+
+/*
+ * Very detailed stats (-d -d), covering the instruction cache and the TLB caches:
+ */
+static struct perf_event_attr very_detailed_attrs[] = {
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_L1I << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_L1I << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_DTLB << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_DTLB << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_ITLB << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_ITLB << 0 |
+ (PERF_COUNT_HW_CACHE_OP_READ << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
+
+};
+
+/*
+ * Very, very detailed stats (-d -d -d), adding prefetch events:
+ */
+static struct perf_event_attr very_very_detailed_attrs[] = {
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_L1D << 0 |
+ (PERF_COUNT_HW_CACHE_OP_PREFETCH << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
+
+ { .type = PERF_TYPE_HW_CACHE,
+ .config =
+ PERF_COUNT_HW_CACHE_L1D << 0 |
+ (PERF_COUNT_HW_CACHE_OP_PREFETCH << 8) |
+ (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
+};
+
+
+
+struct perf_evlist *evsel_list;
-static int system_wide = 0;
-static unsigned int nr_cpus = 0;
+static bool system_wide = false;
static int run_idx = 0;
static int run_count = 1;
-static int inherit = 1;
-static int scale = 1;
-static int target_pid = -1;
-static int null_run = 0;
+static bool no_inherit = false;
+static bool scale = true;
+static bool no_aggr = false;
+static const char *target_pid;
+static const char *target_tid;
+static pid_t child_pid = -1;
+static bool null_run = false;
+static int detailed_run = 0;
+static bool sync_run = false;
+static bool big_num = true;
+static int big_num_opt = -1;
+static const char *cpu_list;
+static const char *csv_sep = NULL;
+static bool csv_output = false;
+static bool group = false;
+static const char *output_name = NULL;
+static FILE *output = NULL;
+static int output_fd;
+
+static volatile int done = 0;
+
+struct stats
+{
+ double n, mean, M2;
+};
-static int fd[MAX_NR_CPUS][MAX_COUNTERS];
+struct perf_stat {
+ struct stats res_stats[3];
+};
-static u64 runtime_nsecs[MAX_RUN];
-static u64 walltime_nsecs[MAX_RUN];
-static u64 runtime_cycles[MAX_RUN];
+static int perf_evsel__alloc_stat_priv(struct perf_evsel *evsel)
+{
+ evsel->priv = zalloc(sizeof(struct perf_stat));
+ return evsel->priv == NULL ? -ENOMEM : 0;
+}
-static u64 event_res[MAX_RUN][MAX_COUNTERS][3];
-static u64 event_scaled[MAX_RUN][MAX_COUNTERS];
+static void perf_evsel__free_stat_priv(struct perf_evsel *evsel)
+{
+ free(evsel->priv);
+ evsel->priv = NULL;
+}
-static u64 event_res_avg[MAX_COUNTERS][3];
-static u64 event_res_noise[MAX_COUNTERS][3];
+static void update_stats(struct stats *stats, u64 val)
+{
+ double delta;
-static u64 event_scaled_avg[MAX_COUNTERS];
+ stats->n++;
+ delta = val - stats->mean;
+ stats->mean += delta / stats->n;
+ stats->M2 += delta*(val - stats->mean);
+}
-static u64 runtime_nsecs_avg;
-static u64 runtime_nsecs_noise;
+static double avg_stats(struct stats *stats)
+{
+ return stats->mean;
+}
-static u64 walltime_nsecs_avg;
-static u64 walltime_nsecs_noise;
+/*
+ * http://en.wikipedia.org/wiki/Algorithms_for_calculating_variance
+ *
+ * (\Sum n_i^2) - ((\Sum n_i)^2)/n
+ * s^2 = -------------------------------
+ * n - 1
+ *
+ * http://en.wikipedia.org/wiki/Stddev
+ *
+ * The std dev of the mean is related to the std dev by:
+ *
+ * s
+ * s_mean = -------
+ * sqrt(n)
+ *
+ */
+static double stddev_stats(struct stats *stats)
+{
+ double variance, variance_mean;
-static u64 runtime_cycles_avg;
-static u64 runtime_cycles_noise;
+ if (!stats->n)
+ return 0.0;
-#define MATCH_EVENT(t, c, counter) \
- (attrs[counter].type == PERF_TYPE_##t && \
- attrs[counter].config == PERF_COUNT_##c)
+ variance = stats->M2 / (stats->n - 1);
+ variance_mean = variance / stats->n;
-#define ERR_PERF_OPEN \
-"Error: counter %d, sys_perf_counter_open() syscall returned with %d (%s)\n"
+ return sqrt(variance_mean);
+}
-static void create_perf_stat_counter(int counter, int pid)
+struct stats runtime_nsecs_stats[MAX_NR_CPUS];
+struct stats runtime_cycles_stats[MAX_NR_CPUS];
+struct stats runtime_stalled_cycles_front_stats[MAX_NR_CPUS];
+struct stats runtime_stalled_cycles_back_stats[MAX_NR_CPUS];
+struct stats runtime_branches_stats[MAX_NR_CPUS];
+struct stats runtime_cacherefs_stats[MAX_NR_CPUS];
+struct stats runtime_l1_dcache_stats[MAX_NR_CPUS];
+struct stats runtime_l1_icache_stats[MAX_NR_CPUS];
+struct stats runtime_ll_cache_stats[MAX_NR_CPUS];
+struct stats runtime_itlb_cache_stats[MAX_NR_CPUS];
+struct stats runtime_dtlb_cache_stats[MAX_NR_CPUS];
+struct stats walltime_nsecs_stats;
+
+static int create_perf_stat_counter(struct perf_evsel *evsel,
+ struct perf_evsel *first)
{
- struct perf_counter_attr *attr = attrs + counter;
+ struct perf_event_attr *attr = &evsel->attr;
+ struct xyarray *group_fd = NULL;
+ bool exclude_guest_missing = false;
+ int ret;
+
+ if (group && evsel != first)
+ group_fd = first->fd;
if (scale)
attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED |
PERF_FORMAT_TOTAL_TIME_RUNNING;
+ attr->inherit = !no_inherit;
+
+retry:
+ if (exclude_guest_missing)
+ evsel->attr.exclude_guest = evsel->attr.exclude_host = 0;
+
if (system_wide) {
- unsigned int cpu;
+ ret = perf_evsel__open_per_cpu(evsel, evsel_list->cpus,
+ group, group_fd);
+ if (ret)
+ goto check_ret;
+ return 0;
+ }
- for (cpu = 0; cpu < nr_cpus; cpu++) {
- fd[cpu][counter] = sys_perf_counter_open(attr, -1, cpu, -1, 0);
- if (fd[cpu][counter] < 0 && verbose)
- fprintf(stderr, ERR_PERF_OPEN, counter,
- fd[cpu][counter], strerror(errno));
- }
- } else {
- attr->inherit = inherit;
- attr->disabled = 1;
+ if (!target_pid && !target_tid && (!group || evsel == first)) {
+ attr->disabled = 1;
attr->enable_on_exec = 1;
+ }
- fd[0][counter] = sys_perf_counter_open(attr, pid, -1, -1, 0);
- if (fd[0][counter] < 0 && verbose)
- fprintf(stderr, ERR_PERF_OPEN, counter,
- fd[0][counter], strerror(errno));
+ ret = perf_evsel__open_per_thread(evsel, evsel_list->threads,
+ group, group_fd);
+ if (!ret)
+ return 0;
+ /* fall through */
+check_ret:
+ if (ret && errno == EINVAL) {
+ if (!exclude_guest_missing &&
+ (evsel->attr.exclude_guest || evsel->attr.exclude_host)) {
+ pr_debug("Old kernel, cannot exclude "
+ "guest or host samples.\n");
+ exclude_guest_missing = true;
+ goto retry;
+ }
}
+ return ret;
}
/*
* Does the counter have nsecs as a unit?
*/
-static inline int nsec_counter(int counter)
+static inline int nsec_counter(struct perf_evsel *evsel)
{
- if (MATCH_EVENT(SOFTWARE, SW_CPU_CLOCK, counter) ||
- MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter))
+ if (perf_evsel__match(evsel, SOFTWARE, SW_CPU_CLOCK) ||
+ perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK))
return 1;
return 0;
}
/*
+ * Update various tracking values we maintain to print
+ * more semantic information such as miss/hit ratios,
+ * instruction rates, etc:
+ */
+static void update_shadow_stats(struct perf_evsel *counter, u64 *count)
+{
+ if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
+ update_stats(&runtime_nsecs_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
+ update_stats(&runtime_cycles_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
+ update_stats(&runtime_stalled_cycles_front_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
+ update_stats(&runtime_stalled_cycles_back_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
+ update_stats(&runtime_branches_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
+ update_stats(&runtime_cacherefs_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
+ update_stats(&runtime_l1_dcache_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1I))
+ update_stats(&runtime_l1_icache_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_LL))
+ update_stats(&runtime_ll_cache_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_DTLB))
+ update_stats(&runtime_dtlb_cache_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_ITLB))
+ update_stats(&runtime_itlb_cache_stats[0], count[0]);
+}
+
+/*
* Read out the results of a single counter:
+ * aggregate counts across CPUs in system-wide mode
*/
-static void read_counter(int counter)
+static int read_counter_aggr(struct perf_evsel *counter)
{
- u64 *count, single_count[3];
- unsigned int cpu;
- size_t res, nv;
- int scaled;
+ struct perf_stat *ps = counter->priv;
+ u64 *count = counter->counts->aggr.values;
+ int i;
- count = event_res[run_idx][counter];
+ if (__perf_evsel__read(counter, evsel_list->cpus->nr,
+ evsel_list->threads->nr, scale) < 0)
+ return -1;
- count[0] = count[1] = count[2] = 0;
+ for (i = 0; i < 3; i++)
+ update_stats(&ps->res_stats[i], count[i]);
- nv = scale ? 3 : 1;
- for (cpu = 0; cpu < nr_cpus; cpu++) {
- if (fd[cpu][counter] < 0)
- continue;
+ if (verbose) {
+ fprintf(output, "%s: %" PRIu64 " %" PRIu64 " %" PRIu64 "\n",
+ event_name(counter), count[0], count[1], count[2]);
+ }
- res = read(fd[cpu][counter], single_count, nv * sizeof(u64));
- assert(res == nv * sizeof(u64));
+ /*
+ * Save the full runtime - to allow normalization during printout:
+ */
+ update_shadow_stats(counter, count);
- close(fd[cpu][counter]);
- fd[cpu][counter] = -1;
+ return 0;
+}
- count[0] += single_count[0];
- if (scale) {
- count[1] += single_count[1];
- count[2] += single_count[2];
- }
- }
+/*
+ * Read out the results of a single counter:
+ * do not aggregate counts across CPUs in system-wide mode
+ */
+static int read_counter(struct perf_evsel *counter)
+{
+ u64 *count;
+ int cpu;
- scaled = 0;
- if (scale) {
- if (count[2] == 0) {
- event_scaled[run_idx][counter] = -1;
- count[0] = 0;
- return;
- }
+ for (cpu = 0; cpu < evsel_list->cpus->nr; cpu++) {
+ if (__perf_evsel__read_on_cpu(counter, cpu, 0, scale) < 0)
+ return -1;
- if (count[2] < count[1]) {
- event_scaled[run_idx][counter] = 1;
- count[0] = (unsigned long long)
- ((double)count[0] * count[1] / count[2] + 0.5);
- }
+ count = counter->counts->cpu[cpu].values;
+
+ update_shadow_stats(counter, count);
}
- /*
- * Save the full runtime - to allow normalization during printout:
- */
- if (MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter))
- runtime_nsecs[run_idx] = count[0];
- if (MATCH_EVENT(HARDWARE, HW_CPU_CYCLES, counter))
- runtime_cycles[run_idx] = count[0];
+
+ return 0;
}
static int run_perf_stat(int argc __used, const char **argv)
{
unsigned long long t0, t1;
+ struct perf_evsel *counter, *first;
int status = 0;
- int counter;
- int pid;
int child_ready_pipe[2], go_pipe[2];
+ const bool forks = (argc > 0);
char buf;
- if (!system_wide)
- nr_cpus = 1;
-
- if (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0) {
+ if (forks && (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0)) {
perror("failed to create pipes");
exit(1);
}
- if ((pid = fork()) < 0)
- perror("failed to fork");
+ if (forks) {
+ if ((child_pid = fork()) < 0)
+ perror("failed to fork");
+
+ if (!child_pid) {
+ close(child_ready_pipe[0]);
+ close(go_pipe[1]);
+ fcntl(go_pipe[0], F_SETFD, FD_CLOEXEC);
+
+ /*
+ * Do a dummy execvp to get the PLT entry resolved,
+ * so we avoid the resolver overhead on the real
+ * execvp call.
+ */
+ execvp("", (char **)argv);
+
+ /*
+ * Tell the parent we're ready to go
+ */
+ close(child_ready_pipe[1]);
+
+ /*
+ * Wait until the parent tells us to go.
+ */
+ if (read(go_pipe[0], &buf, 1) == -1)
+ perror("unable to read pipe");
+
+ execvp(argv[0], (char **)argv);
+
+ perror(argv[0]);
+ exit(-1);
+ }
- if (!pid) {
- close(child_ready_pipe[0]);
- close(go_pipe[1]);
- fcntl(go_pipe[0], F_SETFD, FD_CLOEXEC);
+ if (!target_tid && !target_pid && !system_wide)
+ evsel_list->threads->map[0] = child_pid;
/*
- * Do a dummy execvp to get the PLT entry resolved,
- * so we avoid the resolver overhead on the real
- * execvp call.
- */
- execvp("", (char **)argv);
-
- /*
- * Tell the parent we're ready to go
+ * Wait for the child to be ready to exec.
*/
close(child_ready_pipe[1]);
-
- /*
- * Wait until the parent tells us to go.
- */
- if (read(go_pipe[0], &buf, 1) == -1)
+ close(go_pipe[0]);
+ if (read(child_ready_pipe[0], &buf, 1) == -1)
perror("unable to read pipe");
-
- execvp(argv[0], (char **)argv);
-
- perror(argv[0]);
- exit(-1);
+ close(child_ready_pipe[0]);
}
- /*
- * Wait for the child to be ready to exec.
- */
- close(child_ready_pipe[1]);
- close(go_pipe[0]);
- if (read(child_ready_pipe[0], &buf, 1) == -1)
- perror("unable to read pipe");
- close(child_ready_pipe[0]);
+ first = list_entry(evsel_list->entries.next, struct perf_evsel, node);
+
+ list_for_each_entry(counter, &evsel_list->entries, node) {
+ if (create_perf_stat_counter(counter, first) < 0) {
+ /*
+ * PPC returns ENXIO for HW counters until 2.6.37
+ * (behavior changed with commit b0a873e).
+ */
+ if (errno == EINVAL || errno == ENOSYS ||
+ errno == ENOENT || errno == EOPNOTSUPP ||
+ errno == ENXIO) {
+ if (verbose)
+ ui__warning("%s event is not supported by the kernel.\n",
+ event_name(counter));
+ counter->supported = false;
+ continue;
+ }
+
+ if (errno == EPERM || errno == EACCES) {
+ error("You may not have permission to collect %sstats.\n"
+ "\t Consider tweaking"
+ " /proc/sys/kernel/perf_event_paranoid or running as root.",
+ system_wide ? "system-wide " : "");
+ } else {
+ error("open_counter returned with %d (%s). "
+ "/bin/dmesg may provide additional information.\n",
+ errno, strerror(errno));
+ }
+ if (child_pid != -1)
+ kill(child_pid, SIGTERM);
+ die("Not all events could be opened.\n");
+ return -1;
+ }
+ counter->supported = true;
+ }
- for (counter = 0; counter < nr_counters; counter++)
- create_perf_stat_counter(counter, pid);
+ if (perf_evlist__set_filters(evsel_list)) {
+ error("failed to set filter with %d (%s)\n", errno,
+ strerror(errno));
+ return -1;
+ }
/*
* Enable counters and exec the command:
*/
t0 = rdclock();
- close(go_pipe[1]);
- wait(&status);
+ if (forks) {
+ close(go_pipe[1]);
+ wait(&status);
+ if (WIFSIGNALED(status))
+ psignal(WTERMSIG(status), argv[0]);
+ } else {
+ while(!done) sleep(1);
+ }
t1 = rdclock();
- walltime_nsecs[run_idx] = t1 - t0;
+ update_stats(&walltime_nsecs_stats, t1 - t0);
- for (counter = 0; counter < nr_counters; counter++)
- read_counter(counter);
+ if (no_aggr) {
+ list_for_each_entry(counter, &evsel_list->entries, node) {
+ read_counter(counter);
+ perf_evsel__close_fd(counter, evsel_list->cpus->nr, 1);
+ }
+ } else {
+ list_for_each_entry(counter, &evsel_list->entries, node) {
+ read_counter_aggr(counter);
+ perf_evsel__close_fd(counter, evsel_list->cpus->nr,
+ evsel_list->threads->nr);
+ }
+ }
return WEXITSTATUS(status);
}
-static void print_noise(u64 *count, u64 *noise)
+static void print_noise_pct(double total, double avg)
{
- if (run_count > 1)
- fprintf(stderr, " ( +- %7.3f%% )",
- (double)noise[0]/(count[0]+1)*100.0);
+ double pct = 0.0;
+
+ if (avg)
+ pct = 100.0*total/avg;
+
+ if (csv_output)
+ fprintf(output, "%s%.2f%%", csv_sep, pct);
+ else if (pct)
+ fprintf(output, " ( +-%6.2f%% )", pct);
}
-static void nsec_printout(int counter, u64 *count, u64 *noise)
+static void print_noise(struct perf_evsel *evsel, double avg)
{
- double msecs = (double)count[0] / 1000000;
+ struct perf_stat *ps;
- fprintf(stderr, " %14.6f %-24s", msecs, event_name(counter));
+ if (run_count == 1)
+ return;
- if (MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter)) {
- if (walltime_nsecs_avg)
- fprintf(stderr, " # %10.3f CPUs ",
- (double)count[0] / (double)walltime_nsecs_avg);
- }
- print_noise(count, noise);
+ ps = evsel->priv;
+ print_noise_pct(stddev_stats(&ps->res_stats[0]), avg);
}
-static void abs_printout(int counter, u64 *count, u64 *noise)
+static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg)
{
- fprintf(stderr, " %14Ld %-24s", count[0], event_name(counter));
+ double msecs = avg / 1e6;
+ char cpustr[16] = { '\0', };
+ const char *fmt = csv_output ? "%s%.6f%s%s" : "%s%18.6f%s%-25s";
- if (runtime_cycles_avg &&
- MATCH_EVENT(HARDWARE, HW_INSTRUCTIONS, counter)) {
- fprintf(stderr, " # %10.3f IPC ",
- (double)count[0] / (double)runtime_cycles_avg);
- } else {
- if (runtime_nsecs_avg) {
- fprintf(stderr, " # %10.3f M/sec",
- (double)count[0]/runtime_nsecs_avg*1000.0);
+ if (no_aggr)
+ sprintf(cpustr, "CPU%*d%s",
+ csv_output ? 0 : -4,
+ evsel_list->cpus->map[cpu], csv_sep);
+
+ fprintf(output, fmt, cpustr, msecs, csv_sep, event_name(evsel));
+
+ if (evsel->cgrp)
+ fprintf(output, "%s%s", csv_sep, evsel->cgrp->name);
+
+ if (csv_output)
+ return;
+
+ if (perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK))
+ fprintf(output, " # %8.3f CPUs utilized ",
+ avg / avg_stats(&walltime_nsecs_stats));
+ else
+ fprintf(output, " ");
+}
+
+/* used for get_ratio_color() */
+enum grc_type {
+ GRC_STALLED_CYCLES_FE,
+ GRC_STALLED_CYCLES_BE,
+ GRC_CACHE_MISSES,
+ GRC_MAX_NR
+};
+
+static const char *get_ratio_color(enum grc_type type, double ratio)
+{
+ static const double grc_table[GRC_MAX_NR][3] = {
+ [GRC_STALLED_CYCLES_FE] = { 50.0, 30.0, 10.0 },
+ [GRC_STALLED_CYCLES_BE] = { 75.0, 50.0, 20.0 },
+ [GRC_CACHE_MISSES] = { 20.0, 10.0, 5.0 },
+ };
+ const char *color = PERF_COLOR_NORMAL;
+
+ if (ratio > grc_table[type][0])
+ color = PERF_COLOR_RED;
+ else if (ratio > grc_table[type][1])
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > grc_table[type][2])
+ color = PERF_COLOR_YELLOW;
+
+ return color;
+}
+
+static void print_stalled_cycles_frontend(int cpu, struct perf_evsel *evsel __used, double avg)
+{
+ double total, ratio = 0.0;
+ const char *color;
+
+ total = avg_stats(&runtime_cycles_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);
+
+ fprintf(output, " # ");
+ color_fprintf(output, color, "%6.2f%%", ratio);
+ fprintf(output, " frontend cycles idle ");
+}
+
+static void print_stalled_cycles_backend(int cpu, struct perf_evsel *evsel __used, double avg)
+{
+ double total, ratio = 0.0;
+ const char *color;
+
+ total = avg_stats(&runtime_cycles_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);
+
+ fprintf(output, " # ");
+ color_fprintf(output, color, "%6.2f%%", ratio);
+ fprintf(output, " backend cycles idle ");
+}
+
+static void print_branch_misses(int cpu, struct perf_evsel *evsel __used, double avg)
+{
+ double total, ratio = 0.0;
+ const char *color;
+
+ total = avg_stats(&runtime_branches_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+ fprintf(output, " # ");
+ color_fprintf(output, color, "%6.2f%%", ratio);
+ fprintf(output, " of all branches ");
+}
+
+static void print_l1_dcache_misses(int cpu, struct perf_evsel *evsel __used, double avg)
+{
+ double total, ratio = 0.0;
+ const char *color;
+
+ total = avg_stats(&runtime_l1_dcache_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+ fprintf(output, " # ");
+ color_fprintf(output, color, "%6.2f%%", ratio);
+ fprintf(output, " of all L1-dcache hits ");
+}
+
+static void print_l1_icache_misses(int cpu, struct perf_evsel *evsel __used, double avg)
+{
+ double total, ratio = 0.0;
+ const char *color;
+
+ total = avg_stats(&runtime_l1_icache_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+ fprintf(output, " # ");
+ color_fprintf(output, color, "%6.2f%%", ratio);
+ fprintf(output, " of all L1-icache hits ");
+}
+
+static void print_dtlb_cache_misses(int cpu, struct perf_evsel *evsel __used, double avg)
+{
+ double total, ratio = 0.0;
+ const char *color;
+
+ total = avg_stats(&runtime_dtlb_cache_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+ fprintf(output, " # ");
+ color_fprintf(output, color, "%6.2f%%", ratio);
+ fprintf(output, " of all dTLB cache hits ");
+}
+
+static void print_itlb_cache_misses(int cpu, struct perf_evsel *evsel __used, double avg)
+{
+ double total, ratio = 0.0;
+ const char *color;
+
+ total = avg_stats(&runtime_itlb_cache_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+ fprintf(output, " # ");
+ color_fprintf(output, color, "%6.2f%%", ratio);
+ fprintf(output, " of all iTLB cache hits ");
+}
+
+static void print_ll_cache_misses(int cpu, struct perf_evsel *evsel __used, double avg)
+{
+ double total, ratio = 0.0;
+ const char *color;
+
+ total = avg_stats(&runtime_ll_cache_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+ fprintf(output, " # ");
+ color_fprintf(output, color, "%6.2f%%", ratio);
+ fprintf(output, " of all LL-cache hits ");
+}
+
+static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
+{
+ double total, ratio = 0.0;
+ char cpustr[16] = { '\0', };
+ const char *fmt;
+
+ if (csv_output)
+ fmt = "%s%.0f%s%s";
+ else if (big_num)
+ fmt = "%s%'18.0f%s%-25s";
+ else
+ fmt = "%s%18.0f%s%-25s";
+
+ if (no_aggr)
+ sprintf(cpustr, "CPU%*d%s",
+ csv_output ? 0 : -4,
+ evsel_list->cpus->map[cpu], csv_sep);
+ else
+ cpu = 0;
+
+ fprintf(output, fmt, cpustr, avg, csv_sep, event_name(evsel));
+
+ if (evsel->cgrp)
+ fprintf(output, "%s%s", csv_sep, evsel->cgrp->name);
+
+ if (csv_output)
+ return;
+
+ if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
+ total = avg_stats(&runtime_cycles_stats[cpu]);
+
+ if (total)
+ ratio = avg / total;
+
+ fprintf(output, " # %5.2f insns per cycle ", ratio);
+
+ total = avg_stats(&runtime_stalled_cycles_front_stats[cpu]);
+ total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[cpu]));
+
+ if (total && avg) {
+ ratio = total / avg;
+ fprintf(output, "\n # %5.2f stalled cycles per insn", ratio);
}
+
+ } else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
+ runtime_branches_stats[cpu].n != 0) {
+ print_branch_misses(cpu, evsel, avg);
+ } else if (
+ evsel->attr.type == PERF_TYPE_HW_CACHE &&
+ evsel->attr.config == ( PERF_COUNT_HW_CACHE_L1D |
+ ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
+ ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16)) &&
+ runtime_l1_dcache_stats[cpu].n != 0) {
+ print_l1_dcache_misses(cpu, evsel, avg);
+ } else if (
+ evsel->attr.type == PERF_TYPE_HW_CACHE &&
+ evsel->attr.config == ( PERF_COUNT_HW_CACHE_L1I |
+ ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
+ ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16)) &&
+ runtime_l1_icache_stats[cpu].n != 0) {
+ print_l1_icache_misses(cpu, evsel, avg);
+ } else if (
+ evsel->attr.type == PERF_TYPE_HW_CACHE &&
+ evsel->attr.config == ( PERF_COUNT_HW_CACHE_DTLB |
+ ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
+ ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16)) &&
+ runtime_dtlb_cache_stats[cpu].n != 0) {
+ print_dtlb_cache_misses(cpu, evsel, avg);
+ } else if (
+ evsel->attr.type == PERF_TYPE_HW_CACHE &&
+ evsel->attr.config == ( PERF_COUNT_HW_CACHE_ITLB |
+ ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
+ ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16)) &&
+ runtime_itlb_cache_stats[cpu].n != 0) {
+ print_itlb_cache_misses(cpu, evsel, avg);
+ } else if (
+ evsel->attr.type == PERF_TYPE_HW_CACHE &&
+ evsel->attr.config == ( PERF_COUNT_HW_CACHE_LL |
+ ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
+ ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16)) &&
+ runtime_ll_cache_stats[cpu].n != 0) {
+ print_ll_cache_misses(cpu, evsel, avg);
+ } else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
+ runtime_cacherefs_stats[cpu].n != 0) {
+ total = avg_stats(&runtime_cacherefs_stats[cpu]);
+
+ if (total)
+ ratio = avg * 100 / total;
+
+ fprintf(output, " # %8.3f %% of all cache refs ", ratio);
+
+ } else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
+ print_stalled_cycles_frontend(cpu, evsel, avg);
+ } else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
+ print_stalled_cycles_backend(cpu, evsel, avg);
+ } else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
+ total = avg_stats(&runtime_nsecs_stats[cpu]);
+
+ if (total)
+ ratio = 1.0 * avg / total;
+
+ fprintf(output, " # %8.3f GHz ", ratio);
+ } else if (runtime_nsecs_stats[cpu].n != 0) {
+ char unit = 'M';
+
+ total = avg_stats(&runtime_nsecs_stats[cpu]);
+
+ if (total)
+ ratio = 1000.0 * avg / total;
+ if (ratio < 0.001) {
+ ratio *= 1000;
+ unit = 'K';
+ }
+
+ fprintf(output, " # %8.3f %c/sec ", ratio, unit);
+ } else {
+ fprintf(output, " ");
}
- print_noise(count, noise);
}
/*
* Print out the results of a single counter:
+ * aggregated counts in system-wide mode
*/
-static void print_counter(int counter)
+static void print_counter_aggr(struct perf_evsel *counter)
{
- u64 *count, *noise;
- int scaled;
-
- count = event_res_avg[counter];
- noise = event_res_noise[counter];
- scaled = event_scaled_avg[counter];
+ struct perf_stat *ps = counter->priv;
+ double avg = avg_stats(&ps->res_stats[0]);
+ int scaled = counter->counts->scaled;
if (scaled == -1) {
- fprintf(stderr, " %14s %-24s\n",
- "<not counted>", event_name(counter));
+ fprintf(output, "%*s%s%*s",
+ csv_output ? 0 : 18,
+ counter->supported ? CNTR_NOT_COUNTED : CNTR_NOT_SUPPORTED,
+ csv_sep,
+ csv_output ? 0 : -24,
+ event_name(counter));
+
+ if (counter->cgrp)
+ fprintf(output, "%s%s", csv_sep, counter->cgrp->name);
+
+ fputc('\n', output);
return;
}
if (nsec_counter(counter))
- nsec_printout(counter, count, noise);
+ nsec_printout(-1, counter, avg);
else
- abs_printout(counter, count, noise);
+ abs_printout(-1, counter, avg);
- if (scaled)
- fprintf(stderr, " (scaled from %.2f%%)",
- (double) count[2] / count[1] * 100);
+ print_noise(counter, avg);
- fprintf(stderr, "\n");
-}
+ if (csv_output) {
+ fputc('\n', output);
+ return;
+ }
-/*
- * normalize_noise noise values down to stddev:
- */
-static void normalize_noise(u64 *val)
-{
- double res;
+ if (scaled) {
+ double avg_enabled, avg_running;
- res = (double)*val / (run_count * sqrt((double)run_count));
+ avg_enabled = avg_stats(&ps->res_stats[1]);
+ avg_running = avg_stats(&ps->res_stats[2]);
- *val = (u64)res;
+ fprintf(output, " [%5.2f%%]", 100 * avg_running / avg_enabled);
+ }
+ fprintf(output, "\n");
}
-static void update_avg(const char *name, int idx, u64 *avg, u64 *val)
-{
- *avg += *val;
-
- if (verbose > 1)
- fprintf(stderr, "debug: %20s[%d]: %Ld\n", name, idx, *val);
-}
/*
- * Calculate the averages and noises:
+ * Print out the results of a single counter:
+ * does not use aggregated count in system-wide
*/
-static void calc_avg(void)
-{
- int i, j;
-
- if (verbose > 1)
- fprintf(stderr, "\n");
-
- for (i = 0; i < run_count; i++) {
- update_avg("runtime", 0, &runtime_nsecs_avg, runtime_nsecs + i);
- update_avg("walltime", 0, &walltime_nsecs_avg, walltime_nsecs + i);
- update_avg("runtime_cycles", 0, &runtime_cycles_avg, runtime_cycles + i);
-
- for (j = 0; j < nr_counters; j++) {
- update_avg("counter/0", j,
- event_res_avg[j]+0, event_res[i][j]+0);
- update_avg("counter/1", j,
- event_res_avg[j]+1, event_res[i][j]+1);
- update_avg("counter/2", j,
- event_res_avg[j]+2, event_res[i][j]+2);
- if (event_scaled[i][j] != (u64)-1)
- update_avg("scaled", j,
- event_scaled_avg + j, event_scaled[i]+j);
- else
- event_scaled_avg[j] = -1;
+static void print_counter(struct perf_evsel *counter)
+{
+ u64 ena, run, val;
+ int cpu;
+
+ for (cpu = 0; cpu < evsel_list->cpus->nr; cpu++) {
+ val = counter->counts->cpu[cpu].val;
+ ena = counter->counts->cpu[cpu].ena;
+ run = counter->counts->cpu[cpu].run;
+ if (run == 0 || ena == 0) {
+ fprintf(output, "CPU%*d%s%*s%s%*s",
+ csv_output ? 0 : -4,
+ evsel_list->cpus->map[cpu], csv_sep,
+ csv_output ? 0 : 18,
+ counter->supported ? CNTR_NOT_COUNTED : CNTR_NOT_SUPPORTED,
+ csv_sep,
+ csv_output ? 0 : -24,
+ event_name(counter));
+
+ if (counter->cgrp)
+ fprintf(output, "%s%s",
+ csv_sep, counter->cgrp->name);
+
+ fputc('\n', output);
+ continue;
}
- }
- runtime_nsecs_avg /= run_count;
- walltime_nsecs_avg /= run_count;
- runtime_cycles_avg /= run_count;
-
- for (j = 0; j < nr_counters; j++) {
- event_res_avg[j][0] /= run_count;
- event_res_avg[j][1] /= run_count;
- event_res_avg[j][2] /= run_count;
- }
- for (i = 0; i < run_count; i++) {
- runtime_nsecs_noise +=
- abs((s64)(runtime_nsecs[i] - runtime_nsecs_avg));
- walltime_nsecs_noise +=
- abs((s64)(walltime_nsecs[i] - walltime_nsecs_avg));
- runtime_cycles_noise +=
- abs((s64)(runtime_cycles[i] - runtime_cycles_avg));
-
- for (j = 0; j < nr_counters; j++) {
- event_res_noise[j][0] +=
- abs((s64)(event_res[i][j][0] - event_res_avg[j][0]));
- event_res_noise[j][1] +=
- abs((s64)(event_res[i][j][1] - event_res_avg[j][1]));
- event_res_noise[j][2] +=
- abs((s64)(event_res[i][j][2] - event_res_avg[j][2]));
- }
- }
+ if (nsec_counter(counter))
+ nsec_printout(cpu, counter, val);
+ else
+ abs_printout(cpu, counter, val);
- normalize_noise(&runtime_nsecs_noise);
- normalize_noise(&walltime_nsecs_noise);
- normalize_noise(&runtime_cycles_noise);
+ if (!csv_output) {
+ print_noise(counter, 1.0);
- for (j = 0; j < nr_counters; j++) {
- normalize_noise(&event_res_noise[j][0]);
- normalize_noise(&event_res_noise[j][1]);
- normalize_noise(&event_res_noise[j][2]);
+ if (run != ena)
+ fprintf(output, " (%.2f%%)",
+ 100.0 * run / ena);
+ }
+ fputc('\n', output);
}
}
static void print_stat(int argc, const char **argv)
{
- int i, counter;
-
- calc_avg();
+ struct perf_evsel *counter;
+ int i;
fflush(stdout);
- fprintf(stderr, "\n");
- fprintf(stderr, " Performance counter stats for \'%s", argv[0]);
-
- for (i = 1; i < argc; i++)
- fprintf(stderr, " %s", argv[i]);
-
- fprintf(stderr, "\'");
- if (run_count > 1)
- fprintf(stderr, " (%d runs)", run_count);
- fprintf(stderr, ":\n\n");
+ if (!csv_output) {
+ fprintf(output, "\n");
+ fprintf(output, " Performance counter stats for ");
+ if (!target_pid && !target_tid) {
+ fprintf(output, "\'%s", argv[0]);
+ for (i = 1; i < argc; i++)
+ fprintf(output, " %s", argv[i]);
+ } else if (target_pid)
+ fprintf(output, "process id \'%s", target_pid);
+ else
+ fprintf(output, "thread id \'%s", target_tid);
+
+ fprintf(output, "\'");
+ if (run_count > 1)
+ fprintf(output, " (%d runs)", run_count);
+ fprintf(output, ":\n\n");
+ }
- for (counter = 0; counter < nr_counters; counter++)
- print_counter(counter);
+ if (no_aggr) {
+ list_for_each_entry(counter, &evsel_list->entries, node)
+ print_counter(counter);
+ } else {
+ list_for_each_entry(counter, &evsel_list->entries, node)
+ print_counter_aggr(counter);
+ }
- fprintf(stderr, "\n");
- fprintf(stderr, " %14.9f seconds time elapsed",
- (double)walltime_nsecs_avg/1e9);
- if (run_count > 1) {
- fprintf(stderr, " ( +- %7.3f%% )",
- 100.0*(double)walltime_nsecs_noise/(double)walltime_nsecs_avg);
+ if (!csv_output) {
+ if (!null_run)
+ fprintf(output, "\n");
+ fprintf(output, " %17.9f seconds time elapsed",
+ avg_stats(&walltime_nsecs_stats)/1e9);
+ if (run_count > 1) {
+ fprintf(output, " ");
+ print_noise_pct(stddev_stats(&walltime_nsecs_stats),
+ avg_stats(&walltime_nsecs_stats));
+ }
+ fprintf(output, "\n\n");
}
- fprintf(stderr, "\n\n");
}
static volatile int signr = -1;
static void skip_signal(int signo)
{
+ if(child_pid == -1)
+ done = 1;
+
signr = signo;
}
static void sig_atexit(void)
{
+ if (child_pid != -1)
+ kill(child_pid, SIGTERM);
+
if (signr == -1)
return;
}
static const char * const stat_usage[] = {
- "perf stat [<options>] <command>",
+ "perf stat [<options>] [<command>]",
NULL
};
+static int stat__set_big_num(const struct option *opt __used,
+ const char *s __used, int unset)
+{
+ big_num_opt = unset ? 0 : 1;
+ return 0;
+}
+
+static bool append_file;
+
static const struct option options[] = {
- OPT_CALLBACK('e', "event", NULL, "event",
+ OPT_CALLBACK('e', "event", &evsel_list, "event",
"event selector. use 'perf list' to list available events",
- parse_events),
- OPT_BOOLEAN('i', "inherit", &inherit,
- "child tasks inherit counters"),
- OPT_INTEGER('p', "pid", &target_pid,
- "stat events on existing pid"),
+ parse_events_option),
+ OPT_CALLBACK(0, "filter", &evsel_list, "filter",
+ "event filter", parse_filter),
+ OPT_BOOLEAN('i', "no-inherit", &no_inherit,
+ "child tasks do not inherit counters"),
+ OPT_STRING('p', "pid", &target_pid, "pid",
+ "stat events on existing process id"),
+ OPT_STRING('t', "tid", &target_tid, "tid",
+ "stat events on existing thread id"),
OPT_BOOLEAN('a', "all-cpus", &system_wide,
"system-wide collection from all CPUs"),
+ OPT_BOOLEAN('g', "group", &group,
+ "put the counters into a counter group"),
OPT_BOOLEAN('c', "scale", &scale,
"scale/normalize counters"),
- OPT_BOOLEAN('v', "verbose", &verbose,
+ OPT_INCR('v', "verbose", &verbose,
"be more verbose (show counter open errors, etc)"),
OPT_INTEGER('r', "repeat", &run_count,
"repeat command and print average + stddev (max: 100)"),
OPT_BOOLEAN('n', "null", &null_run,
"null run - dont start any counters"),
+ OPT_INCR('d', "detailed", &detailed_run,
+ "detailed run - start a lot of events"),
+ OPT_BOOLEAN('S', "sync", &sync_run,
+ "call sync() before starting a run"),
+ OPT_CALLBACK_NOOPT('B', "big-num", NULL, NULL,
+ "print large numbers with thousands\' separators",
+ stat__set_big_num),
+ OPT_STRING('C', "cpu", &cpu_list, "cpu",
+ "list of cpus to monitor in system-wide"),
+ OPT_BOOLEAN('A', "no-aggr", &no_aggr,
+ "disable CPU count aggregation"),
+ OPT_STRING('x', "field-separator", &csv_sep, "separator",
+ "print counts with custom separator"),
+ OPT_CALLBACK('G', "cgroup", &evsel_list, "name",
+ "monitor event in cgroup name only",
+ parse_cgroups),
+ OPT_STRING('o', "output", &output_name, "file",
+ "output file name"),
+ 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_END()
};
+/*
+ * Add default attributes, if there were no attributes specified or
+ * if -d/--detailed, -d -d or -d -d -d is used:
+ */
+static int add_default_attributes(void)
+{
+ /* Set attrs if no event is selected and !null_run: */
+ if (null_run)
+ return 0;
+
+ if (!evsel_list->nr_entries) {
+ if (perf_evlist__add_attrs_array(evsel_list, default_attrs) < 0)
+ return -1;
+ }
+
+ /* Detailed events get appended to the event list: */
+
+ if (detailed_run < 1)
+ return 0;
+
+ /* Append detailed run extra attributes: */
+ if (perf_evlist__add_attrs_array(evsel_list, detailed_attrs) < 0)
+ return -1;
+
+ if (detailed_run < 2)
+ return 0;
+
+ /* Append very detailed run extra attributes: */
+ if (perf_evlist__add_attrs_array(evsel_list, very_detailed_attrs) < 0)
+ return -1;
+
+ if (detailed_run < 3)
+ return 0;
+
+ /* Append very, very detailed run extra attributes: */
+ return perf_evlist__add_attrs_array(evsel_list, very_very_detailed_attrs);
+}
+
int cmd_stat(int argc, const char **argv, const char *prefix __used)
{
- int status;
+ struct perf_evsel *pos;
+ int status = -ENOMEM;
+ const char *mode;
+
+ setlocale(LC_ALL, "");
+
+ evsel_list = perf_evlist__new(NULL, NULL);
+ if (evsel_list == NULL)
+ return -ENOMEM;
argc = parse_options(argc, argv, options, stat_usage,
PARSE_OPT_STOP_AT_NON_OPTION);
- if (!argc)
+
+ output = stderr;
+ if (output_name && strcmp(output_name, "-"))
+ output = NULL;
+
+ if (output_name && output_fd) {
+ fprintf(stderr, "cannot use both --output and --log-fd\n");
+ usage_with_options(stat_usage, options);
+ }
+ if (!output) {
+ struct timespec tm;
+ mode = append_file ? "a" : "w";
+
+ output = fopen(output_name, mode);
+ if (!output) {
+ perror("failed to create output file");
+ exit(-1);
+ }
+ clock_gettime(CLOCK_REALTIME, &tm);
+ fprintf(output, "# started on %s\n", ctime(&tm.tv_sec));
+ } else if (output_fd != 2) {
+ mode = append_file ? "a" : "w";
+ output = fdopen(output_fd, mode);
+ if (!output) {
+ perror("Failed opening logfd");
+ return -errno;
+ }
+ }
+
+ if (csv_sep) {
+ csv_output = true;
+ if (!strcmp(csv_sep, "\\t"))
+ csv_sep = "\t";
+ } else
+ csv_sep = DEFAULT_SEPARATOR;
+
+ /*
+ * let the spreadsheet do the pretty-printing
+ */
+ if (csv_output) {
+ /* User explicitly passed -B? */
+ if (big_num_opt == 1) {
+ fprintf(stderr, "-B option not supported with -x\n");
+ usage_with_options(stat_usage, options);
+ } else /* Nope, so disable big number formatting */
+ big_num = false;
+ } else if (big_num_opt == 0) /* User passed --no-big-num */
+ big_num = false;
+
+ if (!argc && !target_pid && !target_tid)
usage_with_options(stat_usage, options);
- if (run_count <= 0 || run_count > MAX_RUN)
+ if (run_count <= 0)
usage_with_options(stat_usage, options);
- /* Set attrs and nr_counters if no event is selected and !null_run */
- if (!null_run && !nr_counters) {
- memcpy(attrs, default_attrs, sizeof(default_attrs));
- nr_counters = ARRAY_SIZE(default_attrs);
+ /* no_aggr, cgroup are for system-wide only */
+ if ((no_aggr || nr_cgroups) && !system_wide) {
+ fprintf(stderr, "both cgroup and no-aggregation "
+ "modes only available in system-wide mode\n");
+
+ usage_with_options(stat_usage, options);
+ }
+
+ if (add_default_attributes())
+ goto out;
+
+ if (target_pid)
+ target_tid = target_pid;
+
+ evsel_list->threads = thread_map__new_str(target_pid,
+ target_tid, UINT_MAX);
+ if (evsel_list->threads == NULL) {
+ pr_err("Problems finding threads of monitor\n");
+ usage_with_options(stat_usage, options);
+ }
+
+ if (system_wide)
+ evsel_list->cpus = cpu_map__new(cpu_list);
+ else
+ evsel_list->cpus = cpu_map__dummy_new();
+
+ if (evsel_list->cpus == NULL) {
+ perror("failed to parse CPUs map");
+ usage_with_options(stat_usage, options);
+ return -1;
}
- nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
- assert(nr_cpus <= MAX_NR_CPUS);
- assert((int)nr_cpus >= 0);
+ list_for_each_entry(pos, &evsel_list->entries, node) {
+ if (perf_evsel__alloc_stat_priv(pos) < 0 ||
+ perf_evsel__alloc_counts(pos, evsel_list->cpus->nr) < 0)
+ goto out_free_fd;
+ }
/*
* We dont want to block the signals - that would cause
status = 0;
for (run_idx = 0; run_idx < run_count; run_idx++) {
if (run_count != 1 && verbose)
- fprintf(stderr, "[ perf stat: executing run #%d ... ]\n", run_idx + 1);
+ fprintf(output, "[ perf stat: executing run #%d ... ]\n",
+ run_idx + 1);
+
+ if (sync_run)
+ sync();
+
status = run_perf_stat(argc, argv);
}
- print_stat(argc, argv);
-
+ if (status != -1)
+ print_stat(argc, argv);
+out_free_fd:
+ list_for_each_entry(pos, &evsel_list->entries, node)
+ perf_evsel__free_stat_priv(pos);
+ perf_evlist__delete_maps(evsel_list);
+out:
+ perf_evlist__delete(evsel_list);
return status;
}
projects / linux-flexiantxendom0-3.2.10.git / blobdiff