summaryrefslogtreecommitdiff
path: root/tools/perf/builtin-stat.c
diff options
context:
space:
mode:
Diffstat (limited to 'tools/perf/builtin-stat.c')
-rw-r--r--tools/perf/builtin-stat.c580
1 files changed, 514 insertions, 66 deletions
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index e2109f9b43eb..a9f06715e44d 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -6,24 +6,28 @@
*
* 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:
*
@@ -46,6 +50,7 @@
#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"
@@ -65,14 +70,107 @@ static struct perf_event_attr default_attrs[] = {
{ .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_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 },
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_REFERENCES },
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CACHE_MISSES },
};
+/*
+ * 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 bool system_wide = false;
@@ -86,6 +184,8 @@ static pid_t target_pid = -1;
static pid_t target_tid = -1;
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;
@@ -156,7 +256,15 @@ static double stddev_stats(struct stats *stats)
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)
@@ -167,16 +275,17 @@ static int create_perf_stat_counter(struct perf_evsel *evsel)
attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED |
PERF_FORMAT_TOTAL_TIME_RUNNING;
+ attr->inherit = !no_inherit;
+
if (system_wide)
- return perf_evsel__open_per_cpu(evsel, evsel_list->cpus, false, false);
+ return perf_evsel__open_per_cpu(evsel, evsel_list->cpus, false);
- attr->inherit = !no_inherit;
if (target_pid == -1 && target_tid == -1) {
attr->disabled = 1;
attr->enable_on_exec = 1;
}
- return perf_evsel__open_per_thread(evsel, evsel_list->threads, false, false);
+ return perf_evsel__open_per_thread(evsel, evsel_list->threads, false);
}
/*
@@ -192,6 +301,37 @@ static inline int nsec_counter(struct perf_evsel *evsel)
}
/*
+ * 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
*/
@@ -216,12 +356,7 @@ static int read_counter_aggr(struct perf_evsel *counter)
/*
* Save the full runtime - to allow normalization during printout:
*/
- if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
- update_stats(&runtime_nsecs_stats[0], count[0]);
- if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
- update_stats(&runtime_cycles_stats[0], count[0]);
- if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
- update_stats(&runtime_branches_stats[0], count[0]);
+ update_shadow_stats(counter, count);
return 0;
}
@@ -241,12 +376,7 @@ static int read_counter(struct perf_evsel *counter)
count = counter->counts->cpu[cpu].values;
- if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
- update_stats(&runtime_nsecs_stats[cpu], count[0]);
- if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
- update_stats(&runtime_cycles_stats[cpu], count[0]);
- if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
- update_stats(&runtime_branches_stats[cpu], count[0]);
+ update_shadow_stats(counter, count);
}
return 0;
@@ -314,13 +444,18 @@ static int run_perf_stat(int argc __used, const char **argv)
list_for_each_entry(counter, &evsel_list->entries, node) {
if (create_perf_stat_counter(counter) < 0) {
- if (errno == -EPERM || errno == -EACCES) {
+ if (errno == EINVAL || errno == ENOSYS || errno == ENOENT) {
+ if (verbose)
+ ui__warning("%s event is not supported by the kernel.\n",
+ event_name(counter));
+ 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 if (errno == ENOENT) {
- error("%s event is not supported. ", event_name(counter));
} else {
error("open_counter returned with %d (%s). "
"/bin/dmesg may provide additional information.\n",
@@ -371,6 +506,16 @@ static int run_perf_stat(int argc __used, const char **argv)
return WEXITSTATUS(status);
}
+static void print_noise_pct(double total, double avg)
+{
+ double pct = 0.0;
+
+ if (avg)
+ pct = 100.0*total/avg;
+
+ fprintf(stderr, " ( +-%6.2f%% )", pct);
+}
+
static void print_noise(struct perf_evsel *evsel, double avg)
{
struct perf_stat *ps;
@@ -379,15 +524,14 @@ static void print_noise(struct perf_evsel *evsel, double avg)
return;
ps = evsel->priv;
- fprintf(stderr, " ( +- %7.3f%% )",
- 100 * stddev_stats(&ps->res_stats[0]) / avg);
+ print_noise_pct(stddev_stats(&ps->res_stats[0]), avg);
}
static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg)
{
double msecs = avg / 1e6;
char cpustr[16] = { '\0', };
- const char *fmt = csv_output ? "%s%.6f%s%s" : "%s%18.6f%s%-24s";
+ const char *fmt = csv_output ? "%s%.6f%s%s" : "%s%18.6f%s%-25s";
if (no_aggr)
sprintf(cpustr, "CPU%*d%s",
@@ -403,8 +547,191 @@ static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg)
return;
if (perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK))
- fprintf(stderr, " # %10.3f CPUs ",
- avg / avg_stats(&walltime_nsecs_stats));
+ fprintf(stderr, " # %8.3f CPUs utilized ", avg / avg_stats(&walltime_nsecs_stats));
+}
+
+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 = PERF_COLOR_NORMAL;
+ if (ratio > 50.0)
+ color = PERF_COLOR_RED;
+ else if (ratio > 30.0)
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > 10.0)
+ color = PERF_COLOR_YELLOW;
+
+ fprintf(stderr, " # ");
+ color_fprintf(stderr, color, "%6.2f%%", ratio);
+ fprintf(stderr, " 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 = PERF_COLOR_NORMAL;
+ if (ratio > 75.0)
+ color = PERF_COLOR_RED;
+ else if (ratio > 50.0)
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > 20.0)
+ color = PERF_COLOR_YELLOW;
+
+ fprintf(stderr, " # ");
+ color_fprintf(stderr, color, "%6.2f%%", ratio);
+ fprintf(stderr, " 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 = PERF_COLOR_NORMAL;
+ if (ratio > 20.0)
+ color = PERF_COLOR_RED;
+ else if (ratio > 10.0)
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > 5.0)
+ color = PERF_COLOR_YELLOW;
+
+ fprintf(stderr, " # ");
+ color_fprintf(stderr, color, "%6.2f%%", ratio);
+ fprintf(stderr, " 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 = PERF_COLOR_NORMAL;
+ if (ratio > 20.0)
+ color = PERF_COLOR_RED;
+ else if (ratio > 10.0)
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > 5.0)
+ color = PERF_COLOR_YELLOW;
+
+ fprintf(stderr, " # ");
+ color_fprintf(stderr, color, "%6.2f%%", ratio);
+ fprintf(stderr, " 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 = PERF_COLOR_NORMAL;
+ if (ratio > 20.0)
+ color = PERF_COLOR_RED;
+ else if (ratio > 10.0)
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > 5.0)
+ color = PERF_COLOR_YELLOW;
+
+ fprintf(stderr, " # ");
+ color_fprintf(stderr, color, "%6.2f%%", ratio);
+ fprintf(stderr, " 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 = PERF_COLOR_NORMAL;
+ if (ratio > 20.0)
+ color = PERF_COLOR_RED;
+ else if (ratio > 10.0)
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > 5.0)
+ color = PERF_COLOR_YELLOW;
+
+ fprintf(stderr, " # ");
+ color_fprintf(stderr, color, "%6.2f%%", ratio);
+ fprintf(stderr, " 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 = PERF_COLOR_NORMAL;
+ if (ratio > 20.0)
+ color = PERF_COLOR_RED;
+ else if (ratio > 10.0)
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > 5.0)
+ color = PERF_COLOR_YELLOW;
+
+ fprintf(stderr, " # ");
+ color_fprintf(stderr, color, "%6.2f%%", ratio);
+ fprintf(stderr, " 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 = PERF_COLOR_NORMAL;
+ if (ratio > 20.0)
+ color = PERF_COLOR_RED;
+ else if (ratio > 10.0)
+ color = PERF_COLOR_MAGENTA;
+ else if (ratio > 5.0)
+ color = PERF_COLOR_YELLOW;
+
+ fprintf(stderr, " # ");
+ color_fprintf(stderr, color, "%6.2f%%", ratio);
+ fprintf(stderr, " of all LL-cache hits ");
}
static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
@@ -416,9 +743,9 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
if (csv_output)
fmt = "%s%.0f%s%s";
else if (big_num)
- fmt = "%s%'18.0f%s%-24s";
+ fmt = "%s%'18.0f%s%-25s";
else
- fmt = "%s%18.0f%s%-24s";
+ fmt = "%s%18.0f%s%-25s";
if (no_aggr)
sprintf(cpustr, "CPU%*d%s",
@@ -441,23 +768,83 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
if (total)
ratio = avg / total;
- fprintf(stderr, " # %10.3f IPC ", ratio);
+ fprintf(stderr, " # %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(stderr, "\n # %5.2f stalled cycles per insn", ratio);
+ }
+
} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
runtime_branches_stats[cpu].n != 0) {
- total = avg_stats(&runtime_branches_stats[cpu]);
+ 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(stderr, " # %10.3f %% ", ratio);
+ fprintf(stderr, " # %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(stderr, " # %8.3f GHz ", ratio);
} else if (runtime_nsecs_stats[cpu].n != 0) {
total = avg_stats(&runtime_nsecs_stats[cpu]);
if (total)
ratio = 1000.0 * avg / total;
- fprintf(stderr, " # %10.3f M/sec", ratio);
+ fprintf(stderr, " # %8.3f M/sec ", ratio);
+ } else {
+ fprintf(stderr, " ");
}
}
@@ -504,8 +891,7 @@ static void print_counter_aggr(struct perf_evsel *counter)
avg_enabled = avg_stats(&ps->res_stats[1]);
avg_running = avg_stats(&ps->res_stats[2]);
- fprintf(stderr, " (scaled from %.2f%%)",
- 100 * avg_running / avg_enabled);
+ fprintf(stderr, " [%5.2f%%]", 100 * avg_running / avg_enabled);
}
fprintf(stderr, "\n");
}
@@ -547,10 +933,8 @@ static void print_counter(struct perf_evsel *counter)
if (!csv_output) {
print_noise(counter, 1.0);
- if (run != ena) {
- fprintf(stderr, " (scaled from %.2f%%)",
- 100.0 * run / ena);
- }
+ if (run != ena)
+ fprintf(stderr, " (%.2f%%)", 100.0 * run / ena);
}
fputc('\n', stderr);
}
@@ -590,13 +974,14 @@ static void print_stat(int argc, const char **argv)
}
if (!csv_output) {
- fprintf(stderr, "\n");
- fprintf(stderr, " %18.9f seconds time elapsed",
+ if (!null_run)
+ fprintf(stderr, "\n");
+ fprintf(stderr, " %17.9f seconds time elapsed",
avg_stats(&walltime_nsecs_stats)/1e9);
if (run_count > 1) {
- fprintf(stderr, " ( +- %7.3f%% )",
- 100*stddev_stats(&walltime_nsecs_stats) /
- avg_stats(&walltime_nsecs_stats));
+ fprintf(stderr, " ");
+ print_noise_pct(stddev_stats(&walltime_nsecs_stats),
+ avg_stats(&walltime_nsecs_stats));
}
fprintf(stderr, "\n\n");
}
@@ -658,6 +1043,10 @@ static const struct option options[] = {
"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),
@@ -673,6 +1062,70 @@ static const struct option options[] = {
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)
+{
+ struct perf_evsel *pos;
+ size_t attr_nr = 0;
+ size_t c;
+
+ /* Set attrs if no event is selected and !null_run: */
+ if (null_run)
+ return 0;
+
+ if (!evsel_list->nr_entries) {
+ for (c = 0; c < ARRAY_SIZE(default_attrs); c++) {
+ pos = perf_evsel__new(default_attrs + c, c + attr_nr);
+ if (pos == NULL)
+ return -1;
+ perf_evlist__add(evsel_list, pos);
+ }
+ attr_nr += c;
+ }
+
+ /* Detailed events get appended to the event list: */
+
+ if (detailed_run < 1)
+ return 0;
+
+ /* Append detailed run extra attributes: */
+ for (c = 0; c < ARRAY_SIZE(detailed_attrs); c++) {
+ pos = perf_evsel__new(detailed_attrs + c, c + attr_nr);
+ if (pos == NULL)
+ return -1;
+ perf_evlist__add(evsel_list, pos);
+ }
+ attr_nr += c;
+
+ if (detailed_run < 2)
+ return 0;
+
+ /* Append very detailed run extra attributes: */
+ for (c = 0; c < ARRAY_SIZE(very_detailed_attrs); c++) {
+ pos = perf_evsel__new(very_detailed_attrs + c, c + attr_nr);
+ if (pos == NULL)
+ return -1;
+ perf_evlist__add(evsel_list, pos);
+ }
+
+ if (detailed_run < 3)
+ return 0;
+
+ /* Append very, very detailed run extra attributes: */
+ for (c = 0; c < ARRAY_SIZE(very_very_detailed_attrs); c++) {
+ pos = perf_evsel__new(very_very_detailed_attrs + c, c + attr_nr);
+ if (pos == NULL)
+ return -1;
+ perf_evlist__add(evsel_list, pos);
+ }
+
+
+ return 0;
+}
+
int cmd_stat(int argc, const char **argv, const char *prefix __used)
{
struct perf_evsel *pos;
@@ -718,17 +1171,8 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used)
usage_with_options(stat_usage, options);
}
- /* Set attrs and nr_counters if no event is selected and !null_run */
- if (!null_run && !evsel_list->nr_entries) {
- size_t c;
-
- for (c = 0; c < ARRAY_SIZE(default_attrs); ++c) {
- pos = perf_evsel__new(&default_attrs[c], c);
- if (pos == NULL)
- goto out;
- perf_evlist__add(evsel_list, pos);
- }
- }
+ if (add_default_attributes())
+ goto out;
if (target_pid != -1)
target_tid = target_pid;
@@ -772,6 +1216,10 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used)
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);
+
+ if (sync_run)
+ sync();
+
status = run_perf_stat(argc, argv);
}
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-22 16:01:22 +0000