[PATCHv2 0/9] perf stat: Separate shadow counters code

[PATCHv2 0/9] perf stat: Separate shadow counters code

[Jiri Olsa]

hi,
sending cleanup/move of shadow counters. I need it
for my other changes of stat scripting, but I think
it can stand alone as a cleanup.

v2 changes:
  - move evsel id into 'struct perf_stat'
    and change the id functions namespace

Also in:
  git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
  perf/shadow

I tested on Haswel with limited set of transaction events,
but all seemed ok.

thanks,
jirka
---
Jiri Olsa (9):
      perf stat: Add id into perf_stat struct
      perf stat: Replace transaction event possition check with id check
      perf stat: Remove setup_events function
      perf stat: Remove transaction_run from shadow update/print code
      perf stat: Introduce reset_shadow_stats function
      perf stat: Introduce print_shadow_stats function
      perf stat: Add output file argument to print_shadow_stats function
      perf stat: Add aggr_mode argument to print_shadow_stats function
      perf stat: Move shadow stat counters into separate object

 tools/perf/builtin-stat.c     | 506 ++------------------------------------------------------------------------
 tools/perf/util/Build         |   1 +
 tools/perf/util/stat-shadow.c | 434 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/stat.c        |  35 +++++-
 tools/perf/util/stat.h        |  40 ++++++
 5 files changed, 522 insertions(+), 494 deletions(-)
 create mode 100644 tools/perf/util/stat-shadow.c

[PATCH 1/9] perf stat: Add id into perf_stat struct

[Jiri Olsa]

We need fast way to identify evsel as transaction event
for shadow counters computation. Currently we are using
possition (in evlist) based way.

Adding 'id' into 'struct perf_stat' so it can carry transaction
event ID and we can use it for shadow counters computations.

Link: http://lkml.kernel.org/n/tip-9lnvjksibrs0flhkpix2qkm9@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c |  6 ++----
 tools/perf/util/stat.c    | 31 ++++++++++++++++++++++++++++++-
 tools/perf/util/stat.h    | 20 ++++++++++++++++++++
 3 files changed, 52 insertions(+), 5 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index fd577f725d23..a6ae1007f1f9 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -147,10 +147,6 @@ static int			(*aggr_get_id)(struct cpu_map *m, int cpu);
 
 static volatile int done = 0;
 
-struct perf_stat {
-	struct stats	  res_stats[3];
-};
-
 static inline void diff_timespec(struct timespec *r, struct timespec *a,
 				 struct timespec *b)
 {
@@ -180,6 +176,8 @@ static void perf_evsel__reset_stat_priv(struct perf_evsel *evsel)
 
 	for (i = 0; i < 3; i++)
 		init_stats(&ps->res_stats[i]);
+
+	perf_stat_evsel_id_init(evsel);
 }
 
 static int perf_evsel__alloc_stat_priv(struct perf_evsel *evsel)
diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
index 6506b3dfb605..873d07607937 100644
--- a/tools/perf/util/stat.c
+++ b/tools/perf/util/stat.c
@@ -1,6 +1,6 @@
 #include <math.h>
-
 #include "stat.h"
+#include "evsel.h"
 
 void update_stats(struct stats *stats, u64 val)
 {
@@ -61,3 +61,32 @@ double rel_stddev_stats(double stddev, double avg)
 
 	return pct;
 }
+
+bool __perf_evsel_stat__is(struct perf_evsel *evsel,
+			   enum perf_stat_evsel_id id)
+{
+	struct perf_stat *ps = evsel->priv;
+
+	return ps->id == id;
+}
+
+#define ID(id, name) [PERF_STAT_EVSEL_ID__##id] = #name
+static const char *id_str[PERF_STAT_EVSEL_ID__MAX] = {
+	ID(NONE, x),
+};
+#undef ID
+
+void perf_stat_evsel_id_init(struct perf_evsel *evsel)
+{
+	struct perf_stat *ps = evsel->priv;
+	int i;
+
+	/* ps->id is 0 hence PERF_STAT_EVSEL_ID__NONE by default */
+
+	for (i = 0; i < PERF_STAT_EVSEL_ID__MAX; i++) {
+		if (!strcmp(evsel->name, id_str[i])) {
+			ps->id = i;
+			break;
+		}
+	}
+}
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index 5667fc3e39cf..f4136cfd3cc9 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -9,6 +9,16 @@ struct stats
 	u64 max, min;
 };
 
+enum perf_stat_evsel_id {
+	PERF_STAT_EVSEL_ID__NONE = 0,
+	PERF_STAT_EVSEL_ID__MAX,
+};
+
+struct perf_stat {
+	struct stats		res_stats[3];
+	enum perf_stat_evsel_id	id;
+};
+
 void update_stats(struct stats *stats, u64 val);
 double avg_stats(struct stats *stats);
 double stddev_stats(struct stats *stats);
@@ -22,4 +32,14 @@ static inline void init_stats(struct stats *stats)
 	stats->min  = (u64) -1;
 	stats->max  = 0;
 }
+
+struct perf_evsel;
+bool __perf_evsel_stat__is(struct perf_evsel *evsel,
+			   enum perf_stat_evsel_id id);
+
+#define perf_stat_evsel__is(evsel, id) \
+	__perf_evsel_stat__is(evsel, PERF_STAT_EVSEL_ID__ ## id)
+
+void perf_stat_evsel_id_init(struct perf_evsel *evsel);
+
 #endif
-- 
1.9.3

[PATCH 2/9] perf stat: Replace transaction event possition check with id check

[Jiri Olsa]

Using perf_stat::id to check for transaction events,
instead of current possition based way.

Link: http://lkml.kernel.org/n/tip-ewc81hslqwhaljvlock51zyv@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c | 55 ++++++-----------------------------------------
 tools/perf/util/stat.c    |  6 +++++-
 tools/perf/util/stat.h    |  4 ++++
 3 files changed, 16 insertions(+), 49 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index a6ae1007f1f9..514493d703da 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -96,17 +96,6 @@ static const char * const transaction_limited_attrs[] = {
 	"}"
 };
 
-/* must match transaction_attrs and the beginning limited_attrs */
-enum {
-	T_TASK_CLOCK,
-	T_INSTRUCTIONS,
-	T_CYCLES,
-	T_CYCLES_IN_TX,
-	T_TRANSACTION_START,
-	T_ELISION_START,
-	T_CYCLES_IN_TX_CP,
-};
-
 static struct perf_evlist	*evsel_list;
 
 static struct target target = {
@@ -352,29 +341,6 @@ static inline int nsec_counter(struct perf_evsel *evsel)
 	return 0;
 }
 
-static struct perf_evsel *nth_evsel(int n)
-{
-	static struct perf_evsel **array;
-	static int array_len;
-	struct perf_evsel *ev;
-	int j;
-
-	/* Assumes this only called when evsel_list does not change anymore. */
-	if (!array) {
-		evlist__for_each(evsel_list, ev)
-			array_len++;
-		array = malloc(array_len * sizeof(void *));
-		if (!array)
-			exit(ENOMEM);
-		j = 0;
-		evlist__for_each(evsel_list, ev)
-			array[j++] = ev;
-	}
-	if (n < array_len)
-		return array[n];
-	return NULL;
-}
-
 /*
  * Update various tracking values we maintain to print
  * more semantic information such as miss/hit ratios,
@@ -389,14 +355,11 @@ static void update_shadow_stats(struct perf_evsel *counter, u64 *count,
 		update_stats(&runtime_nsecs_stats[cpu], count[0]);
 	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
 		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
-	else if (transaction_run &&
-		 perf_evsel__cmp(counter, nth_evsel(T_CYCLES_IN_TX)))
+	else if (transaction_run && perf_stat_evsel__is(counter, CYCLES_IN_TX))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (transaction_run &&
-		 perf_evsel__cmp(counter, nth_evsel(T_TRANSACTION_START)))
+	else if (transaction_run && perf_stat_evsel__is(counter, TRANSACTION_START))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (transaction_run &&
-		 perf_evsel__cmp(counter, nth_evsel(T_ELISION_START)))
+	else if (transaction_run && perf_stat_evsel__is(counter, ELISION_START))
 		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
 	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
 		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
@@ -1207,15 +1170,13 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 		} else {
 			fprintf(output, "                                   ");
 		}
-	} else if (transaction_run &&
-		   perf_evsel__cmp(evsel, nth_evsel(T_CYCLES_IN_TX))) {
+	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total)
 			fprintf(output,
 				" #   %5.2f%% transactional cycles   ",
 				100.0 * (avg / total));
-	} else if (transaction_run &&
-		   perf_evsel__cmp(evsel, nth_evsel(T_CYCLES_IN_TX_CP))) {
+	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
 		if (total2 < avg)
@@ -1224,8 +1185,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 			fprintf(output,
 				" #   %5.2f%% aborted cycles         ",
 				100.0 * ((total2-avg) / total));
-	} else if (transaction_run &&
-		   perf_evsel__cmp(evsel, nth_evsel(T_TRANSACTION_START)) &&
+	} else if (transaction_run && perf_stat_evsel__is(evsel, TRANSACTION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
@@ -1234,8 +1194,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 			ratio = total / avg;
 
 		fprintf(output, " # %8.0f cycles / transaction   ", ratio);
-	} else if (transaction_run &&
-		   perf_evsel__cmp(evsel, nth_evsel(T_ELISION_START)) &&
+	} else if (transaction_run && perf_stat_evsel__is(evsel, ELISION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
index 873d07607937..d22791d02db0 100644
--- a/tools/perf/util/stat.c
+++ b/tools/perf/util/stat.c
@@ -72,7 +72,11 @@ bool __perf_evsel_stat__is(struct perf_evsel *evsel,
 
 #define ID(id, name) [PERF_STAT_EVSEL_ID__##id] = #name
 static const char *id_str[PERF_STAT_EVSEL_ID__MAX] = {
-	ID(NONE, x),
+	ID(NONE,		x),
+	ID(CYCLES_IN_TX,	cpu/cycles-t/),
+	ID(TRANSACTION_START,	cpu/tx-start/),
+	ID(ELISION_START,	cpu/el-start/),
+	ID(CYCLES_IN_TX_CP,	cpu/cycles-ct/),
 };
 #undef ID
 
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index f4136cfd3cc9..3df529bd0774 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -11,6 +11,10 @@ struct stats
 
 enum perf_stat_evsel_id {
 	PERF_STAT_EVSEL_ID__NONE = 0,
+	PERF_STAT_EVSEL_ID__CYCLES_IN_TX,
+	PERF_STAT_EVSEL_ID__TRANSACTION_START,
+	PERF_STAT_EVSEL_ID__ELISION_START,
+	PERF_STAT_EVSEL_ID__CYCLES_IN_TX_CP,
 	PERF_STAT_EVSEL_ID__MAX,
 };
 
-- 
1.9.3

[PATCH 3/9] perf stat: Remove setup_events function

[Jiri Olsa]

We can use already existing parse_events interface.

Both transaction_attrs and transaction_limited_attrs are
changed to be single strings.

Link: http://lkml.kernel.org/n/tip-9r37hrfb81v16xthafgxtppk@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c | 27 +++++++--------------------
 1 file changed, 7 insertions(+), 20 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 514493d703da..0c0071cf4fba 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -73,8 +73,8 @@ static void print_counter(struct perf_evsel *counter, char *prefix);
 static void print_aggr(char *prefix);
 
 /* Default events used for perf stat -T */
-static const char * const transaction_attrs[] = {
-	"task-clock",
+static const char *transaction_attrs = {
+	"task-clock,"
 	"{"
 	"instructions,"
 	"cycles,"
@@ -86,8 +86,8 @@ static const char * const transaction_attrs[] = {
 };
 
 /* More limited version when the CPU does not have all events. */
-static const char * const transaction_limited_attrs[] = {
-	"task-clock",
+static const char * transaction_limited_attrs = {
+	"task-clock,"
 	"{"
 	"instructions,"
 	"cycles,"
@@ -1533,17 +1533,6 @@ static int perf_stat_init_aggr_mode(void)
 	return 0;
 }
 
-static int setup_events(const char * const *attrs, unsigned len)
-{
-	unsigned i;
-
-	for (i = 0; i < len; i++) {
-		if (parse_events(evsel_list, attrs[i], NULL))
-			return -1;
-	}
-	return 0;
-}
-
 /*
  * Add default attributes, if there were no attributes specified or
  * if -d/--detailed, -d -d or -d -d -d is used:
@@ -1665,12 +1654,10 @@ static int add_default_attributes(void)
 		int err;
 		if (pmu_have_event("cpu", "cycles-ct") &&
 		    pmu_have_event("cpu", "el-start"))
-			err = setup_events(transaction_attrs,
-					ARRAY_SIZE(transaction_attrs));
+			err = parse_events(evsel_list, transaction_attrs, NULL);
 		else
-			err = setup_events(transaction_limited_attrs,
-				 ARRAY_SIZE(transaction_limited_attrs));
-		if (err < 0) {
+			err = parse_events(evsel_list, transaction_limited_attrs, NULL);
+		if (err) {
 			fprintf(stderr, "Cannot set up transaction events\n");
 			return -1;
 		}
-- 
1.9.3

[PATCH 4/9] perf stat: Remove transaction_run from shadow update/print code

[Jiri Olsa]

It's no longer needed, because we use nameid to recognize
transaction events.

Keeping it only in stat code to initialize transaction events.

Link: http://lkml.kernel.org/n/tip-eo93tek7cjy8zqdvfd6f5a4x@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 0c0071cf4fba..b3e08ce2c564 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -355,11 +355,11 @@ static void update_shadow_stats(struct perf_evsel *counter, u64 *count,
 		update_stats(&runtime_nsecs_stats[cpu], count[0]);
 	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
 		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
-	else if (transaction_run && perf_stat_evsel__is(counter, CYCLES_IN_TX))
+	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (transaction_run && perf_stat_evsel__is(counter, TRANSACTION_START))
+	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (transaction_run && perf_stat_evsel__is(counter, ELISION_START))
+	else if (perf_stat_evsel__is(counter, ELISION_START))
 		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
 	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
 		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
@@ -1170,13 +1170,13 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 		} else {
 			fprintf(output, "                                   ");
 		}
-	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
+	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total)
 			fprintf(output,
 				" #   %5.2f%% transactional cycles   ",
 				100.0 * (avg / total));
-	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
+	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
 		if (total2 < avg)
@@ -1185,7 +1185,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 			fprintf(output,
 				" #   %5.2f%% aborted cycles         ",
 				100.0 * ((total2-avg) / total));
-	} else if (transaction_run && perf_stat_evsel__is(evsel, TRANSACTION_START) &&
+	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
@@ -1194,7 +1194,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 			ratio = total / avg;
 
 		fprintf(output, " # %8.0f cycles / transaction   ", ratio);
-	} else if (transaction_run && perf_stat_evsel__is(evsel, ELISION_START) &&
+	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
-- 
1.9.3

[PATCH 5/9] perf stat: Introduce reset_shadow_stats function

[Jiri Olsa]

Move shadow counters reset code into separate function
as preparation for moving it into its own object.

Link: http://lkml.kernel.org/n/tip-0i8ssz8kj0rpp5r699srqity@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c | 21 +++++++++++++--------
 1 file changed, 13 insertions(+), 8 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index b3e08ce2c564..fc85e6b9bd13 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -279,15 +279,8 @@ static int evsel_context(struct perf_evsel *evsel)
 	return ctx;
 }
 
-static void perf_stat__reset_stats(struct perf_evlist *evlist)
+static void reset_shadow_stats(void)
 {
-	struct perf_evsel *evsel;
-
-	evlist__for_each(evlist, evsel) {
-		perf_evsel__reset_stat_priv(evsel);
-		perf_evsel__reset_counts(evsel, perf_evsel__nr_cpus(evsel));
-	}
-
 	memset(runtime_nsecs_stats, 0, sizeof(runtime_nsecs_stats));
 	memset(runtime_cycles_stats, 0, sizeof(runtime_cycles_stats));
 	memset(runtime_stalled_cycles_front_stats, 0, sizeof(runtime_stalled_cycles_front_stats));
@@ -307,6 +300,18 @@ static void perf_stat__reset_stats(struct perf_evlist *evlist)
 	memset(&walltime_nsecs_stats, 0, sizeof(walltime_nsecs_stats));
 }
 
+static void perf_stat__reset_stats(struct perf_evlist *evlist)
+{
+	struct perf_evsel *evsel;
+
+	evlist__for_each(evlist, evsel) {
+		perf_evsel__reset_stat_priv(evsel);
+		perf_evsel__reset_counts(evsel, perf_evsel__nr_cpus(evsel));
+	}
+
+	reset_shadow_stats();
+}
+
 static int create_perf_stat_counter(struct perf_evsel *evsel)
 {
 	struct perf_event_attr *attr = &evsel->attr;
-- 
1.9.3

[PATCH 6/9] perf stat: Introduce print_shadow_stats function

[Jiri Olsa]

Move shadow counters display code into separate function
as preparation for moving it into its own object.

Link: http://lkml.kernel.org/n/tip-loph8qgefq6664yqko3jg2tc@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c | 72 +++++++++++++++++++++++++----------------------
 1 file changed, 39 insertions(+), 33 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index fc85e6b9bd13..2ff2e22aa140 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1059,43 +1059,11 @@ static void print_ll_cache_misses(int cpu,
 	fprintf(output, " of all LL-cache hits   ");
 }
 
-static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
+static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 {
 	double total, ratio = 0.0, total2;
-	double sc =  evsel->scale;
-	const char *fmt;
-	int cpu = cpu_map__id_to_cpu(id);
 	int ctx = evsel_context(evsel);
 
-	if (csv_output) {
-		fmt = sc != 1.0 ?  "%.2f%s" : "%.0f%s";
-	} else {
-		if (big_num)
-			fmt = sc != 1.0 ? "%'18.2f%s" : "%'18.0f%s";
-		else
-			fmt = sc != 1.0 ? "%18.2f%s" : "%18.0f%s";
-	}
-
-	aggr_printout(evsel, id, nr);
-
-	if (aggr_mode == AGGR_GLOBAL)
-		cpu = 0;
-
-	fprintf(output, fmt, avg, csv_sep);
-
-	if (evsel->unit)
-		fprintf(output, "%-*s%s",
-			csv_output ? 0 : unit_width,
-			evsel->unit, csv_sep);
-
-	fprintf(output, "%-*s", csv_output ? 0 : 25, perf_evsel__name(evsel));
-
-	if (evsel->cgrp)
-		fprintf(output, "%s%s", csv_sep, evsel->cgrp->name);
-
-	if (csv_output || interval)
-		return;
-
 	if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total) {
@@ -1226,6 +1194,44 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 	}
 }
 
+static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
+{
+	double sc =  evsel->scale;
+	const char *fmt;
+	int cpu = cpu_map__id_to_cpu(id);
+
+	if (csv_output) {
+		fmt = sc != 1.0 ?  "%.2f%s" : "%.0f%s";
+	} else {
+		if (big_num)
+			fmt = sc != 1.0 ? "%'18.2f%s" : "%'18.0f%s";
+		else
+			fmt = sc != 1.0 ? "%18.2f%s" : "%18.0f%s";
+	}
+
+	aggr_printout(evsel, id, nr);
+
+	if (aggr_mode == AGGR_GLOBAL)
+		cpu = 0;
+
+	fprintf(output, fmt, avg, csv_sep);
+
+	if (evsel->unit)
+		fprintf(output, "%-*s%s",
+			csv_output ? 0 : unit_width,
+			evsel->unit, csv_sep);
+
+	fprintf(output, "%-*s", csv_output ? 0 : 25, perf_evsel__name(evsel));
+
+	if (evsel->cgrp)
+		fprintf(output, "%s%s", csv_sep, evsel->cgrp->name);
+
+	if (csv_output || interval)
+		return;
+
+	print_shadow_stats(evsel, avg, cpu);
+}
+
 static void print_aggr(char *prefix)
 {
 	struct perf_evsel *counter;
-- 
1.9.3

[PATCH 7/9] perf stat: Add output file argument to print_shadow_stats function

[Jiri Olsa]

As preparation for moving shadow counters code into its own object.

Link: http://lkml.kernel.org/n/tip-uhx2opot7qtd3xdq9pvx4z8v@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c | 112 +++++++++++++++++++++++-----------------------
 1 file changed, 56 insertions(+), 56 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 2ff2e22aa140..14a75ddb60c7 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -899,7 +899,7 @@ static const char *get_ratio_color(enum grc_type type, double ratio)
 	return color;
 }
 
-static void print_stalled_cycles_frontend(int cpu,
+static void print_stalled_cycles_frontend(FILE *out, int cpu,
 					  struct perf_evsel *evsel
 					  __maybe_unused, double avg)
 {
@@ -914,12 +914,12 @@ static void print_stalled_cycles_frontend(int cpu,
 
 	color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " frontend cycles idle   ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " frontend cycles idle   ");
 }
 
-static void print_stalled_cycles_backend(int cpu,
+static void print_stalled_cycles_backend(FILE *out, int cpu,
 					 struct perf_evsel *evsel
 					 __maybe_unused, double avg)
 {
@@ -934,12 +934,12 @@ static void print_stalled_cycles_backend(int cpu,
 
 	color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " backend  cycles idle   ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " backend  cycles idle   ");
 }
 
-static void print_branch_misses(int cpu,
+static void print_branch_misses(FILE *out, int cpu,
 				struct perf_evsel *evsel __maybe_unused,
 				double avg)
 {
@@ -954,12 +954,12 @@ static void print_branch_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all branches        ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all branches        ");
 }
 
-static void print_l1_dcache_misses(int cpu,
+static void print_l1_dcache_misses(FILE *out, int cpu,
 				   struct perf_evsel *evsel __maybe_unused,
 				   double avg)
 {
@@ -974,12 +974,12 @@ static void print_l1_dcache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all L1-dcache hits  ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all L1-dcache hits  ");
 }
 
-static void print_l1_icache_misses(int cpu,
+static void print_l1_icache_misses(FILE *out, int cpu,
 				   struct perf_evsel *evsel __maybe_unused,
 				   double avg)
 {
@@ -994,12 +994,12 @@ static void print_l1_icache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all L1-icache hits  ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all L1-icache hits  ");
 }
 
-static void print_dtlb_cache_misses(int cpu,
+static void print_dtlb_cache_misses(FILE *out, int cpu,
 				    struct perf_evsel *evsel __maybe_unused,
 				    double avg)
 {
@@ -1014,12 +1014,12 @@ static void print_dtlb_cache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all dTLB cache hits ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all dTLB cache hits ");
 }
 
-static void print_itlb_cache_misses(int cpu,
+static void print_itlb_cache_misses(FILE *out, int cpu,
 				    struct perf_evsel *evsel __maybe_unused,
 				    double avg)
 {
@@ -1034,12 +1034,12 @@ static void print_itlb_cache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all iTLB cache hits ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all iTLB cache hits ");
 }
 
-static void print_ll_cache_misses(int cpu,
+static void print_ll_cache_misses(FILE *out, int cpu,
 				  struct perf_evsel *evsel __maybe_unused,
 				  double avg)
 {
@@ -1054,12 +1054,12 @@ static void print_ll_cache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all LL-cache hits   ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all LL-cache hits   ");
 }
 
-static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
+static void print_shadow_stats(FILE *out, struct perf_evsel *evsel, double avg, int cpu)
 {
 	double total, ratio = 0.0, total2;
 	int ctx = evsel_context(evsel);
@@ -1068,59 +1068,59 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total) {
 			ratio = avg / total;
-			fprintf(output, " #   %5.2f  insns per cycle        ", ratio);
+			fprintf(out, " #   %5.2f  insns per cycle        ", ratio);
 		} else {
-			fprintf(output, "                                   ");
+			fprintf(out, "                                   ");
 		}
 		total = avg_stats(&runtime_stalled_cycles_front_stats[ctx][cpu]);
 		total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[ctx][cpu]));
 
 		if (total && avg) {
 			ratio = total / avg;
-			fprintf(output, "\n");
+			fprintf(out, "\n");
 			if (aggr_mode == AGGR_NONE)
-				fprintf(output, "        ");
-			fprintf(output, "                                                  #   %5.2f  stalled cycles per insn", ratio);
+				fprintf(out, "        ");
+			fprintf(out, "                                                  #   %5.2f  stalled cycles per insn", ratio);
 		}
 
 	} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
 			runtime_branches_stats[ctx][cpu].n != 0) {
-		print_branch_misses(cpu, evsel, avg);
+		print_branch_misses(out, 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[ctx][cpu].n != 0) {
-		print_l1_dcache_misses(cpu, evsel, avg);
+		print_l1_dcache_misses(out, 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[ctx][cpu].n != 0) {
-		print_l1_icache_misses(cpu, evsel, avg);
+		print_l1_icache_misses(out, 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[ctx][cpu].n != 0) {
-		print_dtlb_cache_misses(cpu, evsel, avg);
+		print_dtlb_cache_misses(out, 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[ctx][cpu].n != 0) {
-		print_itlb_cache_misses(cpu, evsel, avg);
+		print_itlb_cache_misses(out, 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[ctx][cpu].n != 0) {
-		print_ll_cache_misses(cpu, evsel, avg);
+		print_ll_cache_misses(out, cpu, evsel, avg);
 	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
 			runtime_cacherefs_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cacherefs_stats[ctx][cpu]);
@@ -1128,25 +1128,25 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		if (total)
 			ratio = avg * 100 / total;
 
-		fprintf(output, " # %8.3f %% of all cache refs    ", ratio);
+		fprintf(out, " # %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);
+		print_stalled_cycles_frontend(out, cpu, evsel, avg);
 	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
-		print_stalled_cycles_backend(cpu, evsel, avg);
+		print_stalled_cycles_backend(out, cpu, evsel, avg);
 	} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
 		total = avg_stats(&runtime_nsecs_stats[cpu]);
 
 		if (total) {
 			ratio = avg / total;
-			fprintf(output, " # %8.3f GHz                    ", ratio);
+			fprintf(out, " # %8.3f GHz                    ", ratio);
 		} else {
-			fprintf(output, "                                   ");
+			fprintf(out, "                                   ");
 		}
 	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total)
-			fprintf(output,
+			fprintf(out,
 				" #   %5.2f%% transactional cycles   ",
 				100.0 * (avg / total));
 	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
@@ -1155,7 +1155,7 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		if (total2 < avg)
 			total2 = avg;
 		if (total)
-			fprintf(output,
+			fprintf(out,
 				" #   %5.2f%% aborted cycles         ",
 				100.0 * ((total2-avg) / total));
 	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
@@ -1166,7 +1166,7 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		if (total)
 			ratio = total / avg;
 
-		fprintf(output, " # %8.0f cycles / transaction   ", ratio);
+		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
 	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
@@ -1175,7 +1175,7 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		if (total)
 			ratio = total / avg;
 
-		fprintf(output, " # %8.0f cycles / elision       ", ratio);
+		fprintf(out, " # %8.0f cycles / elision       ", ratio);
 	} else if (runtime_nsecs_stats[cpu].n != 0) {
 		char unit = 'M';
 
@@ -1188,9 +1188,9 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 			unit = 'K';
 		}
 
-		fprintf(output, " # %8.3f %c/sec                  ", ratio, unit);
+		fprintf(out, " # %8.3f %c/sec                  ", ratio, unit);
 	} else {
-		fprintf(output, "                                   ");
+		fprintf(out, "                                   ");
 	}
 }
 
@@ -1229,7 +1229,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 	if (csv_output || interval)
 		return;
 
-	print_shadow_stats(evsel, avg, cpu);
+	print_shadow_stats(output, evsel, avg, cpu);
 }
 
 static void print_aggr(char *prefix)
-- 
1.9.3

[PATCH 8/9] perf stat: Add aggr_mode argument to print_shadow_stats function

[Jiri Olsa]

As preparation for moving shadow counters code into its own object.

Link: http://lkml.kernel.org/n/tip-kdgryri2kahing584g5is35b@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 14a75ddb60c7..50918dc9fb31 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1059,7 +1059,8 @@ static void print_ll_cache_misses(FILE *out, int cpu,
 	fprintf(out, " of all LL-cache hits   ");
 }
 
-static void print_shadow_stats(FILE *out, struct perf_evsel *evsel, double avg, int cpu)
+static void print_shadow_stats(FILE *out, struct perf_evsel *evsel,
+			       double avg, int cpu, enum aggr_mode aggr)
 {
 	double total, ratio = 0.0, total2;
 	int ctx = evsel_context(evsel);
@@ -1078,7 +1079,7 @@ static void print_shadow_stats(FILE *out, struct perf_evsel *evsel, double avg,
 		if (total && avg) {
 			ratio = total / avg;
 			fprintf(out, "\n");
-			if (aggr_mode == AGGR_NONE)
+			if (aggr == AGGR_NONE)
 				fprintf(out, "        ");
 			fprintf(out, "                                                  #   %5.2f  stalled cycles per insn", ratio);
 		}
@@ -1229,7 +1230,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 	if (csv_output || interval)
 		return;
 
-	print_shadow_stats(output, evsel, avg, cpu);
+	print_shadow_stats(output, evsel, avg, cpu, aggr_mode);
 }
 
 static void print_aggr(char *prefix)
-- 
1.9.3

[PATCH 9/9] perf stat: Move shadow stat counters into separate object

[Jiri Olsa]

Separating shadow counters code into separate object
as a cleanup, but mainly for upcomming changes, so
could use it from script command context.

Link: http://lkml.kernel.org/n/tip-ic34l55jem02kx4h3wodn7ql@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c     | 444 +-----------------------------------------
 tools/perf/util/Build         |   1 +
 tools/perf/util/stat-shadow.c | 434 +++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/stat.h        |  16 ++
 4 files changed, 455 insertions(+), 440 deletions(-)
 create mode 100644 tools/perf/util/stat-shadow.c

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 50918dc9fb31..ff3d25803400 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -102,13 +102,6 @@ static struct target target = {
 	.uid	= UINT_MAX,
 };
 
-enum aggr_mode {
-	AGGR_NONE,
-	AGGR_GLOBAL,
-	AGGR_SOCKET,
-	AGGR_CORE,
-};
-
 static int			run_count			=  1;
 static bool			no_inherit			= false;
 static bool			scale				=  true;
@@ -234,72 +227,6 @@ out_free:
 	return -1;
 }
 
-enum {
-	CTX_BIT_USER	= 1 << 0,
-	CTX_BIT_KERNEL	= 1 << 1,
-	CTX_BIT_HV	= 1 << 2,
-	CTX_BIT_HOST	= 1 << 3,
-	CTX_BIT_IDLE	= 1 << 4,
-	CTX_BIT_MAX	= 1 << 5,
-};
-
-#define NUM_CTX CTX_BIT_MAX
-
-static struct stats runtime_nsecs_stats[MAX_NR_CPUS];
-static struct stats runtime_cycles_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_stalled_cycles_front_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_stalled_cycles_back_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_branches_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_cacherefs_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_l1_dcache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_l1_icache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_ll_cache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_itlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_dtlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_cycles_in_tx_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats walltime_nsecs_stats;
-static struct stats runtime_transaction_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_elision_stats[NUM_CTX][MAX_NR_CPUS];
-
-static int evsel_context(struct perf_evsel *evsel)
-{
-	int ctx = 0;
-
-	if (evsel->attr.exclude_kernel)
-		ctx |= CTX_BIT_KERNEL;
-	if (evsel->attr.exclude_user)
-		ctx |= CTX_BIT_USER;
-	if (evsel->attr.exclude_hv)
-		ctx |= CTX_BIT_HV;
-	if (evsel->attr.exclude_host)
-		ctx |= CTX_BIT_HOST;
-	if (evsel->attr.exclude_idle)
-		ctx |= CTX_BIT_IDLE;
-
-	return ctx;
-}
-
-static void reset_shadow_stats(void)
-{
-	memset(runtime_nsecs_stats, 0, sizeof(runtime_nsecs_stats));
-	memset(runtime_cycles_stats, 0, sizeof(runtime_cycles_stats));
-	memset(runtime_stalled_cycles_front_stats, 0, sizeof(runtime_stalled_cycles_front_stats));
-	memset(runtime_stalled_cycles_back_stats, 0, sizeof(runtime_stalled_cycles_back_stats));
-	memset(runtime_branches_stats, 0, sizeof(runtime_branches_stats));
-	memset(runtime_cacherefs_stats, 0, sizeof(runtime_cacherefs_stats));
-	memset(runtime_l1_dcache_stats, 0, sizeof(runtime_l1_dcache_stats));
-	memset(runtime_l1_icache_stats, 0, sizeof(runtime_l1_icache_stats));
-	memset(runtime_ll_cache_stats, 0, sizeof(runtime_ll_cache_stats));
-	memset(runtime_itlb_cache_stats, 0, sizeof(runtime_itlb_cache_stats));
-	memset(runtime_dtlb_cache_stats, 0, sizeof(runtime_dtlb_cache_stats));
-	memset(runtime_cycles_in_tx_stats, 0,
-			sizeof(runtime_cycles_in_tx_stats));
-	memset(runtime_transaction_stats, 0,
-		sizeof(runtime_transaction_stats));
-	memset(runtime_elision_stats, 0, sizeof(runtime_elision_stats));
-	memset(&walltime_nsecs_stats, 0, sizeof(walltime_nsecs_stats));
-}
-
 static void perf_stat__reset_stats(struct perf_evlist *evlist)
 {
 	struct perf_evsel *evsel;
@@ -309,7 +236,7 @@ static void perf_stat__reset_stats(struct perf_evlist *evlist)
 		perf_evsel__reset_counts(evsel, perf_evsel__nr_cpus(evsel));
 	}
 
-	reset_shadow_stats();
+	perf_stat__reset_shadow_stats();
 }
 
 static int create_perf_stat_counter(struct perf_evsel *evsel)
@@ -346,46 +273,6 @@ static inline int nsec_counter(struct perf_evsel *evsel)
 	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,
-				int cpu)
-{
-	int ctx = evsel_context(counter);
-
-	if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
-		update_stats(&runtime_nsecs_stats[cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
-		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
-	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
-		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
-		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (perf_stat_evsel__is(counter, ELISION_START))
-		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
-		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
-		update_stats(&runtime_stalled_cycles_back_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
-		update_stats(&runtime_branches_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
-		update_stats(&runtime_cacherefs_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
-		update_stats(&runtime_l1_dcache_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1I))
-		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_LL))
-		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_DTLB))
-		update_stats(&runtime_dtlb_cache_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_ITLB))
-		update_stats(&runtime_itlb_cache_stats[ctx][cpu], count[0]);
-}
-
 static void zero_per_pkg(struct perf_evsel *counter)
 {
 	if (counter->per_pkg_mask)
@@ -446,7 +333,7 @@ static int read_cb(struct perf_evsel *evsel, int cpu, int thread __maybe_unused,
 		perf_counts_values__scale(count, scale, NULL);
 		evsel->counts->cpu[cpu] = *count;
 		if (aggr_mode == AGGR_NONE)
-			update_shadow_stats(evsel, count->values, cpu);
+			perf_stat__update_shadow_stats(evsel, count->values, cpu);
 		break;
 	case AGGR_GLOBAL:
 		aggr->val += count->val;
@@ -494,7 +381,7 @@ static int read_counter_aggr(struct perf_evsel *counter)
 	/*
 	 * Save the full runtime - to allow normalization during printout:
 	 */
-	update_shadow_stats(counter, count, 0);
+	perf_stat__update_shadow_stats(counter, count, 0);
 
 	return 0;
 }
@@ -872,329 +759,6 @@ static void nsec_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 		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(FILE *out, int cpu,
-					  struct perf_evsel *evsel
-					  __maybe_unused, double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " frontend cycles idle   ");
-}
-
-static void print_stalled_cycles_backend(FILE *out, int cpu,
-					 struct perf_evsel *evsel
-					 __maybe_unused, double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " backend  cycles idle   ");
-}
-
-static void print_branch_misses(FILE *out, int cpu,
-				struct perf_evsel *evsel __maybe_unused,
-				double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_branches_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all branches        ");
-}
-
-static void print_l1_dcache_misses(FILE *out, int cpu,
-				   struct perf_evsel *evsel __maybe_unused,
-				   double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_l1_dcache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all L1-dcache hits  ");
-}
-
-static void print_l1_icache_misses(FILE *out, int cpu,
-				   struct perf_evsel *evsel __maybe_unused,
-				   double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_l1_icache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all L1-icache hits  ");
-}
-
-static void print_dtlb_cache_misses(FILE *out, int cpu,
-				    struct perf_evsel *evsel __maybe_unused,
-				    double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_dtlb_cache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all dTLB cache hits ");
-}
-
-static void print_itlb_cache_misses(FILE *out, int cpu,
-				    struct perf_evsel *evsel __maybe_unused,
-				    double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_itlb_cache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all iTLB cache hits ");
-}
-
-static void print_ll_cache_misses(FILE *out, int cpu,
-				  struct perf_evsel *evsel __maybe_unused,
-				  double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_ll_cache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all LL-cache hits   ");
-}
-
-static void print_shadow_stats(FILE *out, struct perf_evsel *evsel,
-			       double avg, int cpu, enum aggr_mode aggr)
-{
-	double total, ratio = 0.0, total2;
-	int ctx = evsel_context(evsel);
-
-	if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
-		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-		if (total) {
-			ratio = avg / total;
-			fprintf(out, " #   %5.2f  insns per cycle        ", ratio);
-		} else {
-			fprintf(out, "                                   ");
-		}
-		total = avg_stats(&runtime_stalled_cycles_front_stats[ctx][cpu]);
-		total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[ctx][cpu]));
-
-		if (total && avg) {
-			ratio = total / avg;
-			fprintf(out, "\n");
-			if (aggr == AGGR_NONE)
-				fprintf(out, "        ");
-			fprintf(out, "                                                  #   %5.2f  stalled cycles per insn", ratio);
-		}
-
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
-			runtime_branches_stats[ctx][cpu].n != 0) {
-		print_branch_misses(out, 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[ctx][cpu].n != 0) {
-		print_l1_dcache_misses(out, 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[ctx][cpu].n != 0) {
-		print_l1_icache_misses(out, 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[ctx][cpu].n != 0) {
-		print_dtlb_cache_misses(out, 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[ctx][cpu].n != 0) {
-		print_itlb_cache_misses(out, 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[ctx][cpu].n != 0) {
-		print_ll_cache_misses(out, cpu, evsel, avg);
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
-			runtime_cacherefs_stats[ctx][cpu].n != 0) {
-		total = avg_stats(&runtime_cacherefs_stats[ctx][cpu]);
-
-		if (total)
-			ratio = avg * 100 / total;
-
-		fprintf(out, " # %8.3f %% of all cache refs    ", ratio);
-
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
-		print_stalled_cycles_frontend(out, cpu, evsel, avg);
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
-		print_stalled_cycles_backend(out, cpu, evsel, avg);
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
-		total = avg_stats(&runtime_nsecs_stats[cpu]);
-
-		if (total) {
-			ratio = avg / total;
-			fprintf(out, " # %8.3f GHz                    ", ratio);
-		} else {
-			fprintf(out, "                                   ");
-		}
-	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
-		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-		if (total)
-			fprintf(out,
-				" #   %5.2f%% transactional cycles   ",
-				100.0 * (avg / total));
-	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
-		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
-		if (total2 < avg)
-			total2 = avg;
-		if (total)
-			fprintf(out,
-				" #   %5.2f%% aborted cycles         ",
-				100.0 * ((total2-avg) / total));
-	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
-		   avg > 0 &&
-		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
-		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
-
-		if (total)
-			ratio = total / avg;
-
-		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
-	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
-		   avg > 0 &&
-		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
-		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
-
-		if (total)
-			ratio = total / avg;
-
-		fprintf(out, " # %8.0f cycles / elision       ", 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(out, " # %8.3f %c/sec                  ", ratio, unit);
-	} else {
-		fprintf(out, "                                   ");
-	}
-}
-
 static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 {
 	double sc =  evsel->scale;
@@ -1230,7 +794,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 	if (csv_output || interval)
 		return;
 
-	print_shadow_stats(output, evsel, avg, cpu, aggr_mode);
+	perf_stat__print_shadow_stats(output, evsel, avg, cpu, aggr_mode);
 }
 
 static void print_aggr(char *prefix)
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index e4b676de2f64..586a59d46022 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -68,6 +68,7 @@ libperf-y += rblist.o
 libperf-y += intlist.o
 libperf-y += vdso.o
 libperf-y += stat.o
+libperf-y += stat-shadow.o
 libperf-y += record.o
 libperf-y += srcline.o
 libperf-y += data.o
diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
new file mode 100644
index 000000000000..53e8bb7bc852
--- /dev/null
+++ b/tools/perf/util/stat-shadow.c
@@ -0,0 +1,434 @@
+#include <stdio.h>
+#include "evsel.h"
+#include "stat.h"
+#include "color.h"
+
+enum {
+	CTX_BIT_USER	= 1 << 0,
+	CTX_BIT_KERNEL	= 1 << 1,
+	CTX_BIT_HV	= 1 << 2,
+	CTX_BIT_HOST	= 1 << 3,
+	CTX_BIT_IDLE	= 1 << 4,
+	CTX_BIT_MAX	= 1 << 5,
+};
+
+#define NUM_CTX CTX_BIT_MAX
+
+static struct stats runtime_nsecs_stats[MAX_NR_CPUS];
+static struct stats runtime_cycles_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_stalled_cycles_front_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_stalled_cycles_back_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_branches_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_cacherefs_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_l1_dcache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_l1_icache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_ll_cache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_itlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_dtlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_cycles_in_tx_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_transaction_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_elision_stats[NUM_CTX][MAX_NR_CPUS];
+
+struct stats walltime_nsecs_stats;
+
+static int evsel_context(struct perf_evsel *evsel)
+{
+	int ctx = 0;
+
+	if (evsel->attr.exclude_kernel)
+		ctx |= CTX_BIT_KERNEL;
+	if (evsel->attr.exclude_user)
+		ctx |= CTX_BIT_USER;
+	if (evsel->attr.exclude_hv)
+		ctx |= CTX_BIT_HV;
+	if (evsel->attr.exclude_host)
+		ctx |= CTX_BIT_HOST;
+	if (evsel->attr.exclude_idle)
+		ctx |= CTX_BIT_IDLE;
+
+	return ctx;
+}
+
+void perf_stat__reset_shadow_stats(void)
+{
+	memset(runtime_nsecs_stats, 0, sizeof(runtime_nsecs_stats));
+	memset(runtime_cycles_stats, 0, sizeof(runtime_cycles_stats));
+	memset(runtime_stalled_cycles_front_stats, 0, sizeof(runtime_stalled_cycles_front_stats));
+	memset(runtime_stalled_cycles_back_stats, 0, sizeof(runtime_stalled_cycles_back_stats));
+	memset(runtime_branches_stats, 0, sizeof(runtime_branches_stats));
+	memset(runtime_cacherefs_stats, 0, sizeof(runtime_cacherefs_stats));
+	memset(runtime_l1_dcache_stats, 0, sizeof(runtime_l1_dcache_stats));
+	memset(runtime_l1_icache_stats, 0, sizeof(runtime_l1_icache_stats));
+	memset(runtime_ll_cache_stats, 0, sizeof(runtime_ll_cache_stats));
+	memset(runtime_itlb_cache_stats, 0, sizeof(runtime_itlb_cache_stats));
+	memset(runtime_dtlb_cache_stats, 0, sizeof(runtime_dtlb_cache_stats));
+	memset(runtime_cycles_in_tx_stats, 0,
+			sizeof(runtime_cycles_in_tx_stats));
+	memset(runtime_transaction_stats, 0,
+		sizeof(runtime_transaction_stats));
+	memset(runtime_elision_stats, 0, sizeof(runtime_elision_stats));
+	memset(&walltime_nsecs_stats, 0, sizeof(walltime_nsecs_stats));
+}
+
+/*
+ * Update various tracking values we maintain to print
+ * more semantic information such as miss/hit ratios,
+ * instruction rates, etc:
+ */
+void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
+				    int cpu)
+{
+	int ctx = evsel_context(counter);
+
+	if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
+		update_stats(&runtime_nsecs_stats[cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
+		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
+	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
+		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
+	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
+		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
+	else if (perf_stat_evsel__is(counter, ELISION_START))
+		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
+		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
+		update_stats(&runtime_stalled_cycles_back_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
+		update_stats(&runtime_branches_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
+		update_stats(&runtime_cacherefs_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
+		update_stats(&runtime_l1_dcache_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1I))
+		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_LL))
+		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_DTLB))
+		update_stats(&runtime_dtlb_cache_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_ITLB))
+		update_stats(&runtime_itlb_cache_stats[ctx][cpu], count[0]);
+}
+
+/* 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(FILE *out, int cpu,
+					  struct perf_evsel *evsel
+					  __maybe_unused, double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " frontend cycles idle   ");
+}
+
+static void print_stalled_cycles_backend(FILE *out, int cpu,
+					 struct perf_evsel *evsel
+					 __maybe_unused, double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " backend  cycles idle   ");
+}
+
+static void print_branch_misses(FILE *out, int cpu,
+				struct perf_evsel *evsel __maybe_unused,
+				double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_branches_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all branches        ");
+}
+
+static void print_l1_dcache_misses(FILE *out, int cpu,
+				   struct perf_evsel *evsel __maybe_unused,
+				   double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_l1_dcache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all L1-dcache hits  ");
+}
+
+static void print_l1_icache_misses(FILE *out, int cpu,
+				   struct perf_evsel *evsel __maybe_unused,
+				   double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_l1_icache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all L1-icache hits  ");
+}
+
+static void print_dtlb_cache_misses(FILE *out, int cpu,
+				    struct perf_evsel *evsel __maybe_unused,
+				    double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_dtlb_cache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all dTLB cache hits ");
+}
+
+static void print_itlb_cache_misses(FILE *out, int cpu,
+				    struct perf_evsel *evsel __maybe_unused,
+				    double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_itlb_cache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all iTLB cache hits ");
+}
+
+static void print_ll_cache_misses(FILE *out, int cpu,
+				  struct perf_evsel *evsel __maybe_unused,
+				  double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_ll_cache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all LL-cache hits   ");
+}
+
+void perf_stat__print_shadow_stats(FILE *out, struct perf_evsel *evsel,
+				   double avg, int cpu, enum aggr_mode aggr)
+{
+	double total, ratio = 0.0, total2;
+	int ctx = evsel_context(evsel);
+
+	if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
+		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+		if (total) {
+			ratio = avg / total;
+			fprintf(out, " #   %5.2f  insns per cycle        ", ratio);
+		} else {
+			fprintf(out, "                                   ");
+		}
+		total = avg_stats(&runtime_stalled_cycles_front_stats[ctx][cpu]);
+		total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[ctx][cpu]));
+
+		if (total && avg) {
+			ratio = total / avg;
+			fprintf(out, "\n");
+			if (aggr == AGGR_NONE)
+				fprintf(out, "        ");
+			fprintf(out, "                                                  #   %5.2f  stalled cycles per insn", ratio);
+		}
+
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
+			runtime_branches_stats[ctx][cpu].n != 0) {
+		print_branch_misses(out, 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[ctx][cpu].n != 0) {
+		print_l1_dcache_misses(out, 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[ctx][cpu].n != 0) {
+		print_l1_icache_misses(out, 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[ctx][cpu].n != 0) {
+		print_dtlb_cache_misses(out, 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[ctx][cpu].n != 0) {
+		print_itlb_cache_misses(out, 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[ctx][cpu].n != 0) {
+		print_ll_cache_misses(out, cpu, evsel, avg);
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
+			runtime_cacherefs_stats[ctx][cpu].n != 0) {
+		total = avg_stats(&runtime_cacherefs_stats[ctx][cpu]);
+
+		if (total)
+			ratio = avg * 100 / total;
+
+		fprintf(out, " # %8.3f %% of all cache refs    ", ratio);
+
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
+		print_stalled_cycles_frontend(out, cpu, evsel, avg);
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
+		print_stalled_cycles_backend(out, cpu, evsel, avg);
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
+		total = avg_stats(&runtime_nsecs_stats[cpu]);
+
+		if (total) {
+			ratio = avg / total;
+			fprintf(out, " # %8.3f GHz                    ", ratio);
+		} else {
+			fprintf(out, "                                   ");
+		}
+	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
+		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+		if (total)
+			fprintf(out,
+				" #   %5.2f%% transactional cycles   ",
+				100.0 * (avg / total));
+	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
+		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
+		if (total2 < avg)
+			total2 = avg;
+		if (total)
+			fprintf(out,
+				" #   %5.2f%% aborted cycles         ",
+				100.0 * ((total2-avg) / total));
+	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
+		   avg > 0 &&
+		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
+		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
+
+		if (total)
+			ratio = total / avg;
+
+		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
+	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
+		   avg > 0 &&
+		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
+		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
+
+		if (total)
+			ratio = total / avg;
+
+		fprintf(out, " # %8.0f cycles / elision       ", 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(out, " # %8.3f %c/sec                  ", ratio, unit);
+	} else {
+		fprintf(out, "                                   ");
+	}
+}
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index 3df529bd0774..615c779eb42a 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -2,6 +2,7 @@
 #define __PERF_STATS_H
 
 #include <linux/types.h>
+#include <stdio.h>
 
 struct stats
 {
@@ -23,6 +24,13 @@ struct perf_stat {
 	enum perf_stat_evsel_id	id;
 };
 
+enum aggr_mode {
+	AGGR_NONE,
+	AGGR_GLOBAL,
+	AGGR_SOCKET,
+	AGGR_CORE,
+};
+
 void update_stats(struct stats *stats, u64 val);
 double avg_stats(struct stats *stats);
 double stddev_stats(struct stats *stats);
@@ -46,4 +54,12 @@ bool __perf_evsel_stat__is(struct perf_evsel *evsel,
 
 void perf_stat_evsel_id_init(struct perf_evsel *evsel);
 
+extern struct stats walltime_nsecs_stats;
+
+void perf_stat__reset_shadow_stats(void);
+void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
+				    int cpu);
+void perf_stat__print_shadow_stats(FILE *out, struct perf_evsel *evsel,
+				   double avg, int cpu, enum aggr_mode aggr);
+
 #endif
-- 
1.9.3

Re: [PATCH 4/9] perf stat: Remove transaction_run from shadow update/print code

[Arnaldo Carvalho de Melo]

Em Wed, Jun 03, 2015 at 04:25:54PM +0200, Jiri Olsa escreveu:
> It's no longer needed, because we use nameid to recognize
> transaction events.
> 
> Keeping it only in stat code to initialize transaction events.

Lemme see if I understand this correctly, this update_shadow_stats() is
called only for transaction runs?

It doesn't seem so, so now we will update those stats all the time?

- Arnaldo
 
> Link: http://lkml.kernel.org/n/tip-eo93tek7cjy8zqdvfd6f5a4x@git.kernel.org
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  tools/perf/builtin-stat.c | 14 +++++++-------
>  1 file changed, 7 insertions(+), 7 deletions(-)
> 
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index 0c0071cf4fba..b3e08ce2c564 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -355,11 +355,11 @@ static void update_shadow_stats(struct perf_evsel *counter, u64 *count,
>  		update_stats(&runtime_nsecs_stats[cpu], count[0]);
>  	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
>  		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
> -	else if (transaction_run && perf_stat_evsel__is(counter, CYCLES_IN_TX))
> +	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
>  		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
> -	else if (transaction_run && perf_stat_evsel__is(counter, TRANSACTION_START))
> +	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
>  		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
> -	else if (transaction_run && perf_stat_evsel__is(counter, ELISION_START))
> +	else if (perf_stat_evsel__is(counter, ELISION_START))
>  		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
>  	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
>  		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
> @@ -1170,13 +1170,13 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
>  		} else {
>  			fprintf(output, "                                   ");
>  		}
> -	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
> +	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
>  		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
>  		if (total)
>  			fprintf(output,
>  				" #   %5.2f%% transactional cycles   ",
>  				100.0 * (avg / total));
> -	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
> +	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
>  		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
>  		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
>  		if (total2 < avg)
> @@ -1185,7 +1185,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
>  			fprintf(output,
>  				" #   %5.2f%% aborted cycles         ",
>  				100.0 * ((total2-avg) / total));
> -	} else if (transaction_run && perf_stat_evsel__is(evsel, TRANSACTION_START) &&
> +	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
>  		   avg > 0 &&
>  		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
>  		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
> @@ -1194,7 +1194,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
>  			ratio = total / avg;
>  
>  		fprintf(output, " # %8.0f cycles / transaction   ", ratio);
> -	} else if (transaction_run && perf_stat_evsel__is(evsel, ELISION_START) &&
> +	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
>  		   avg > 0 &&
>  		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
>  		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
> -- 
> 1.9.3

Re: [PATCH 4/9] perf stat: Remove transaction_run from shadow update/print code

[Jiri Olsa]

On Wed, Jun 03, 2015 at 12:10:38PM -0300, Arnaldo Carvalho de Melo wrote:
> Em Wed, Jun 03, 2015 at 04:25:54PM +0200, Jiri Olsa escreveu:
> > It's no longer needed, because we use nameid to recognize
> > transaction events.
> > 
> > Keeping it only in stat code to initialize transaction events.
> 
> Lemme see if I understand this correctly, this update_shadow_stats() is
> called only for transaction runs?
> 
> It doesn't seem so, so now we will update those stats all the time?

update_shadow_stats is called on currently read/processed
counter and based on its 'type' we update various other
counters - shadow counters - which are the base for things
like IPC in the stat output

jirka

Re: [PATCH 4/9] perf stat: Remove transaction_run from shadow update/print code

[Arnaldo Carvalho de Melo]

Em Wed, Jun 03, 2015 at 05:14:50PM +0200, Jiri Olsa escreveu:
> On Wed, Jun 03, 2015 at 12:10:38PM -0300, Arnaldo Carvalho de Melo wrote:
> > Em Wed, Jun 03, 2015 at 04:25:54PM +0200, Jiri Olsa escreveu:
> > > It's no longer needed, because we use nameid to recognize
> > > transaction events.
> > > 
> > > Keeping it only in stat code to initialize transaction events.
> > 
> > Lemme see if I understand this correctly, this update_shadow_stats() is
> > called only for transaction runs?
> > 
> > It doesn't seem so, so now we will update those stats all the time?
> 
> update_shadow_stats is called on currently read/processed
> counter and based on its 'type' we update various other
> counters - shadow counters - which are the base for things
> like IPC in the stat output


So:

-       else if (transaction_run && perf_stat_evsel__is(counter, CYCLES_IN_TX))
+       else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
                update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);


Before we were going to check this _only_ when transaction_run was set,
now we are doing it always, or are you saying that
perf_stat_evsel__is(counter, CYCLES_IN_TX) will only return true for
transaction 'perf stat' usage?

Let me take a look:

Yes, in all cases we, start doing:

   perf_evlist__alloc_stats()
     perf_evsel__alloc_stat_priv()
       perf_evsel__reset_stat_priv()
         perf_stat_evsel_id_init();
           struct perf_stat *ps = evsel->priv;
           ps->id = one of the entries below:

static const char *id_str[PERF_STAT_EVSEL_ID__MAX] = {
        ID(NONE,                x),
        ID(CYCLES_IN_TX,        cpu/cycles-t/),
        ID(TRANSACTION_START,   cpu/tx-start/),
        ID(ELISION_START,       cpu/el-start/),
        ID(CYCLES_IN_TX_CP,     cpu/cycles-ct/),
};


Ok, guess I understood now, all the events above are transaction events,
so if one of them is present, then, yes, this is a transaction run, and
we don't need to look at the 'transaction_run' variable to update the
shadow stats, got it right?

I.e. we _never_ needed to look at 'transaction_id', just noticing that
ps->id is not zero, ok.

- Arnaldo

Re: [PATCHv2 0/9] perf stat: Separate shadow counters code

[Arnaldo Carvalho de Melo]

Em Wed, Jun 03, 2015 at 04:25:50PM +0200, Jiri Olsa escreveu:
> hi,
> sending cleanup/move of shadow counters. I need it
> for my other changes of stat scripting, but I think
> it can stand alone as a cleanup.
> 
> v2 changes:
>   - move evsel id into 'struct perf_stat'
>     and change the id functions namespace

Thanks, applied.

- Arnaldo

Re: [PATCHv2 0/9] perf stat: Separate shadow counters code

[Arnaldo Carvalho de Melo]

Em Wed, Jun 03, 2015 at 07:07:05PM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Wed, Jun 03, 2015 at 04:25:50PM +0200, Jiri Olsa escreveu:
> > hi,
> > sending cleanup/move of shadow counters. I need it
> > for my other changes of stat scripting, but I think
> > it can stand alone as a cleanup.
> > 
> > v2 changes:
> >   - move evsel id into 'struct perf_stat'
> >     and change the id functions namespace
> 
> Thanks, applied.

15: struct perf_event_attr setup                             :sh: line 1: 27101 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmptb9pWi /home/acme/bin/perf stat -o /tmp/tmptb9pWi/perf.data kill > /dev/null 2>&1
sh: line 1: 27104 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpWgRK9D /home/acme/bin/perf stat -o /tmp/tmpWgRK9D/perf.data -d kill > /dev/null 2>&1
sh: line 1: 27107 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpmCOyBe /home/acme/bin/perf stat -o /tmp/tmpmCOyBe/perf.data -dd kill > /dev/null 2>&1
sh: line 1: 27110 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpJP9x2e /home/acme/bin/perf stat -o /tmp/tmpJP9x2e/perf.data -ddd kill > /dev/null 2>&1
 Ok

Can you please check?

- Arnaldo

Re: [PATCHv2 0/9] perf stat: Separate shadow counters code

[Jiri Olsa]

On Wed, Jun 03, 2015 at 07:24:53PM -0300, Arnaldo Carvalho de Melo wrote:
> Em Wed, Jun 03, 2015 at 07:07:05PM -0300, Arnaldo Carvalho de Melo escreveu:
> > Em Wed, Jun 03, 2015 at 04:25:50PM +0200, Jiri Olsa escreveu:
> > > hi,
> > > sending cleanup/move of shadow counters. I need it
> > > for my other changes of stat scripting, but I think
> > > it can stand alone as a cleanup.
> > > 
> > > v2 changes:
> > >   - move evsel id into 'struct perf_stat'
> > >     and change the id functions namespace
> > 
> > Thanks, applied.
> 
> 15: struct perf_event_attr setup                             :sh: line 1: 27101 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmptb9pWi /home/acme/bin/perf stat -o /tmp/tmptb9pWi/perf.data kill > /dev/null 2>&1
> sh: line 1: 27104 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpWgRK9D /home/acme/bin/perf stat -o /tmp/tmpWgRK9D/perf.data -d kill > /dev/null 2>&1
> sh: line 1: 27107 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpmCOyBe /home/acme/bin/perf stat -o /tmp/tmpmCOyBe/perf.data -dd kill > /dev/null 2>&1
> sh: line 1: 27110 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpJP9x2e /home/acme/bin/perf stat -o /tmp/tmpJP9x2e/perf.data -ddd kill > /dev/null 2>&1
>  Ok
> 
> Can you please check?

arghh.. sry, I remember running tests :-\ really ;-)
it's the -d default events that do not set evsel->name

attached patch fixes it, but I'll send v3 for the first patch

thanks,
jirka


---
diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
index d22791d02db0..dba762da5679 100644
--- a/tools/perf/util/stat.c
+++ b/tools/perf/util/stat.c
@@ -88,7 +88,7 @@ void perf_stat_evsel_id_init(struct perf_evsel *evsel)
 	/* ps->id is 0 hence PERF_STAT_EVSEL_ID__NONE by default */
 
 	for (i = 0; i < PERF_STAT_EVSEL_ID__MAX; i++) {
-		if (!strcmp(evsel->name, id_str[i])) {
+		if (evsel->name && !strcmp(evsel->name, id_str[i])) {
 			ps->id = i;
 			break;
 		}

Re: [PATCHv2 0/9] perf stat: Separate shadow counters code

[Arnaldo Carvalho de Melo]

Em Thu, Jun 04, 2015 at 12:38:24AM +0200, Jiri Olsa escreveu:
> On Wed, Jun 03, 2015 at 07:24:53PM -0300, Arnaldo Carvalho de Melo wrote:
> > Em Wed, Jun 03, 2015 at 07:07:05PM -0300, Arnaldo Carvalho de Melo escreveu:
> > > Em Wed, Jun 03, 2015 at 04:25:50PM +0200, Jiri Olsa escreveu:
> > > > hi,
> > > > sending cleanup/move of shadow counters. I need it
> > > > for my other changes of stat scripting, but I think
> > > > it can stand alone as a cleanup.
> > > > 
> > > > v2 changes:
> > > >   - move evsel id into 'struct perf_stat'
> > > >     and change the id functions namespace
> > > 
> > > Thanks, applied.
> > 
> > 15: struct perf_event_attr setup                             :sh: line 1: 27101 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmptb9pWi /home/acme/bin/perf stat -o /tmp/tmptb9pWi/perf.data kill > /dev/null 2>&1
> > sh: line 1: 27104 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpWgRK9D /home/acme/bin/perf stat -o /tmp/tmpWgRK9D/perf.data -d kill > /dev/null 2>&1
> > sh: line 1: 27107 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpmCOyBe /home/acme/bin/perf stat -o /tmp/tmpmCOyBe/perf.data -dd kill > /dev/null 2>&1
> > sh: line 1: 27110 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpJP9x2e /home/acme/bin/perf stat -o /tmp/tmpJP9x2e/perf.data -ddd kill > /dev/null 2>&1
> >  Ok
> > 
> > Can you please check?
> 
> arghh.. sry, I remember running tests :-\ really ;-)
> it's the -d default events that do not set evsel->name

> attached patch fixes it, but I'll send v3 for the first patch
> 
> thanks,
> jirka
> 
> 
> ---
> diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
> index d22791d02db0..dba762da5679 100644
> --- a/tools/perf/util/stat.c
> +++ b/tools/perf/util/stat.c
> @@ -88,7 +88,7 @@ void perf_stat_evsel_id_init(struct perf_evsel *evsel)
>  	/* ps->id is 0 hence PERF_STAT_EVSEL_ID__NONE by default */
>  
>  	for (i = 0; i < PERF_STAT_EVSEL_ID__MAX; i++) {
> -		if (!strcmp(evsel->name, id_str[i])) {
> +		if (evsel->name && !strcmp(evsel->name, id_str[i])) {

Why not use perf_evsel__name() then?

>  			ps->id = i;
>  			break;
>  		}

Re: [PATCHv2 0/9] perf stat: Separate shadow counters code

[Jiri Olsa]

On Wed, Jun 03, 2015 at 07:42:00PM -0300, Arnaldo Carvalho de Melo wrote:
> Em Thu, Jun 04, 2015 at 12:38:24AM +0200, Jiri Olsa escreveu:
> > On Wed, Jun 03, 2015 at 07:24:53PM -0300, Arnaldo Carvalho de Melo wrote:
> > > Em Wed, Jun 03, 2015 at 07:07:05PM -0300, Arnaldo Carvalho de Melo escreveu:
> > > > Em Wed, Jun 03, 2015 at 04:25:50PM +0200, Jiri Olsa escreveu:
> > > > > hi,
> > > > > sending cleanup/move of shadow counters. I need it
> > > > > for my other changes of stat scripting, but I think
> > > > > it can stand alone as a cleanup.
> > > > > 
> > > > > v2 changes:
> > > > >   - move evsel id into 'struct perf_stat'
> > > > >     and change the id functions namespace
> > > > 
> > > > Thanks, applied.
> > > 
> > > 15: struct perf_event_attr setup                             :sh: line 1: 27101 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmptb9pWi /home/acme/bin/perf stat -o /tmp/tmptb9pWi/perf.data kill > /dev/null 2>&1
> > > sh: line 1: 27104 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpWgRK9D /home/acme/bin/perf stat -o /tmp/tmpWgRK9D/perf.data -d kill > /dev/null 2>&1
> > > sh: line 1: 27107 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpmCOyBe /home/acme/bin/perf stat -o /tmp/tmpmCOyBe/perf.data -dd kill > /dev/null 2>&1
> > > sh: line 1: 27110 Segmentation fault      (core dumped) PERF_TEST_ATTR=/tmp/tmpJP9x2e /home/acme/bin/perf stat -o /tmp/tmpJP9x2e/perf.data -ddd kill > /dev/null 2>&1
> > >  Ok
> > > 
> > > Can you please check?
> > 
> > arghh.. sry, I remember running tests :-\ really ;-)
> > it's the -d default events that do not set evsel->name
> 
> > attached patch fixes it, but I'll send v3 for the first patch
> > 
> > thanks,
> > jirka
> > 
> > 
> > ---
> > diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
> > index d22791d02db0..dba762da5679 100644
> > --- a/tools/perf/util/stat.c
> > +++ b/tools/perf/util/stat.c
> > @@ -88,7 +88,7 @@ void perf_stat_evsel_id_init(struct perf_evsel *evsel)
> >  	/* ps->id is 0 hence PERF_STAT_EVSEL_ID__NONE by default */
> >  
> >  	for (i = 0; i < PERF_STAT_EVSEL_ID__MAX; i++) {
> > -		if (!strcmp(evsel->name, id_str[i])) {
> > +		if (evsel->name && !strcmp(evsel->name, id_str[i])) {
> 
> Why not use perf_evsel__name() then?

right, I'll repost v3 for first patch in a few

jirka

[PATCHv3] perf stat: Add id into perf_stat struct

[Jiri Olsa]

On Wed, Jun 03, 2015 at 04:25:51PM +0200, Jiri Olsa wrote:
> We need fast way to identify evsel as transaction event
> for shadow counters computation. Currently we are using
> possition (in evlist) based way.
> 
> Adding 'id' into 'struct perf_stat' so it can carry transaction
> event ID and we can use it for shadow counters computations.
> 
> Link: http://lkml.kernel.org/n/tip-9lnvjksibrs0flhkpix2qkm9@git.kernel.org
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---

using perf_evsel__name instead of evsel->name pointer

jirka

---
We need fast way to identify evsel as transaction event
for shadow counters computation. Currently we are using
possition (in evlist) based way.

Adding 'id' into 'struct perf_stat' so it can carry transaction
event ID and we can use it for shadow counters computations.

Link: http://lkml.kernel.org/n/tip-9lnvjksibrs0flhkpix2qkm9@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-stat.c |  6 ++----
 tools/perf/util/stat.c    | 31 ++++++++++++++++++++++++++++++-
 tools/perf/util/stat.h    | 20 ++++++++++++++++++++
 3 files changed, 52 insertions(+), 5 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index fd577f725d23..a6ae1007f1f9 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -147,10 +147,6 @@ static int			(*aggr_get_id)(struct cpu_map *m, int cpu);
 
 static volatile int done = 0;
 
-struct perf_stat {
-	struct stats	  res_stats[3];
-};
-
 static inline void diff_timespec(struct timespec *r, struct timespec *a,
 				 struct timespec *b)
 {
@@ -180,6 +176,8 @@ static void perf_evsel__reset_stat_priv(struct perf_evsel *evsel)
 
 	for (i = 0; i < 3; i++)
 		init_stats(&ps->res_stats[i]);
+
+	perf_stat_evsel_id_init(evsel);
 }
 
 static int perf_evsel__alloc_stat_priv(struct perf_evsel *evsel)
diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
index 6506b3dfb605..8e9f6bb7581b 100644
--- a/tools/perf/util/stat.c
+++ b/tools/perf/util/stat.c
@@ -1,6 +1,6 @@
 #include <math.h>
-
 #include "stat.h"
+#include "evsel.h"
 
 void update_stats(struct stats *stats, u64 val)
 {
@@ -61,3 +61,32 @@ double rel_stddev_stats(double stddev, double avg)
 
 	return pct;
 }
+
+bool __perf_evsel_stat__is(struct perf_evsel *evsel,
+			   enum perf_stat_evsel_id id)
+{
+	struct perf_stat *ps = evsel->priv;
+
+	return ps->id == id;
+}
+
+#define ID(id, name) [PERF_STAT_EVSEL_ID__##id] = #name
+static const char *id_str[PERF_STAT_EVSEL_ID__MAX] = {
+	ID(NONE, x),
+};
+#undef ID
+
+void perf_stat_evsel_id_init(struct perf_evsel *evsel)
+{
+	struct perf_stat *ps = evsel->priv;
+	int i;
+
+	/* ps->id is 0 hence PERF_STAT_EVSEL_ID__NONE by default */
+
+	for (i = 0; i < PERF_STAT_EVSEL_ID__MAX; i++) {
+		if (!strcmp(perf_evsel__name(evsel), id_str[i])) {
+			ps->id = i;
+			break;
+		}
+	}
+}
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index 5667fc3e39cf..f4136cfd3cc9 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -9,6 +9,16 @@ struct stats
 	u64 max, min;
 };
 
+enum perf_stat_evsel_id {
+	PERF_STAT_EVSEL_ID__NONE = 0,
+	PERF_STAT_EVSEL_ID__MAX,
+};
+
+struct perf_stat {
+	struct stats		res_stats[3];
+	enum perf_stat_evsel_id	id;
+};
+
 void update_stats(struct stats *stats, u64 val);
 double avg_stats(struct stats *stats);
 double stddev_stats(struct stats *stats);
@@ -22,4 +32,14 @@ static inline void init_stats(struct stats *stats)
 	stats->min  = (u64) -1;
 	stats->max  = 0;
 }
+
+struct perf_evsel;
+bool __perf_evsel_stat__is(struct perf_evsel *evsel,
+			   enum perf_stat_evsel_id id);
+
+#define perf_stat_evsel__is(evsel, id) \
+	__perf_evsel_stat__is(evsel, PERF_STAT_EVSEL_ID__ ## id)
+
+void perf_stat_evsel_id_init(struct perf_evsel *evsel);
+
 #endif
-- 
1.9.3

Re: [PATCHv3] perf stat: Add id into perf_stat struct

[Arnaldo Carvalho de Melo]

Em Thu, Jun 04, 2015 at 03:50:55PM +0200, Jiri Olsa escreveu:
> On Wed, Jun 03, 2015 at 04:25:51PM +0200, Jiri Olsa wrote:
> > We need fast way to identify evsel as transaction event
> > for shadow counters computation. Currently we are using
> > possition (in evlist) based way.
> > 
> > Adding 'id' into 'struct perf_stat' so it can carry transaction
> > event ID and we can use it for shadow counters computations.
> > 
> > Link: http://lkml.kernel.org/n/tip-9lnvjksibrs0flhkpix2qkm9@git.kernel.org
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > ---
> 
> using perf_evsel__name instead of evsel->name pointer

It now passes 'perf test' on rhel7/fedora21 systems, but it fails on a
rhel6 system, I applied it, will test it further on the rhel6 system to
see what went wrong there, unless you do it first ;-)

[root@sandy ~]# perf test 15
15: struct perf_event_attr setup                             :sh: line
1:  5381 Aborted                 (core dumped)
PERF_TEST_ATTR=/tmp/tmp8ZBLwc /home/acme/bin/perf stat -o
/tmp/tmp8ZBLwc/perf.data --group -e cycles,instructions kill > /dev/null
2>&1
sh: line 1:  5438 Aborted                 (core dumped)
PERF_TEST_ATTR=/tmp/tmpnRDADB /home/acme/bin/perf stat -o
/tmp/tmpnRDADB/perf.data -e '{cycles,instructions}' kill > /dev/null
2>&1
 Ok
[root@sandy ~]# uname -r
2.6.32-504.8.1.el6.x86_64
[root@sandy ~]# cat /etc/redhat-release 
Red Hat Enterprise Linux Server release 6.6 (Santiago)
[root@sandy ~]# 

- Arnaldo

[tip:perf/core] perf stat: Add id into perf_stat struct

[tip-bot for Jiri Olsa]

Commit-ID:  e2f56da1d6670070f6f55d43007cb7b03ee04c2f
Gitweb:     http://git.kernel.org/tip/e2f56da1d6670070f6f55d43007cb7b03ee04c2f
Author:     Jiri Olsa <jolsa@redhat.com>
AuthorDate: Thu, 4 Jun 2015 15:50:55 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:30 -0300

perf stat: Add id into perf_stat struct

We need fast way to identify evsel as transaction event for shadow
counters computation. Currently we are using possition (in evlist) based
way.

Adding 'id' into 'struct perf_stat' so it can carry transaction event ID
and we can use it for shadow counters computations.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/20150604135055.GB23625@krava.redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c |  6 ++----
 tools/perf/util/stat.c    | 31 ++++++++++++++++++++++++++++++-
 tools/perf/util/stat.h    | 20 ++++++++++++++++++++
 3 files changed, 52 insertions(+), 5 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index fd577f7..a6ae100 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -147,10 +147,6 @@ static int			(*aggr_get_id)(struct cpu_map *m, int cpu);
 
 static volatile int done = 0;
 
-struct perf_stat {
-	struct stats	  res_stats[3];
-};
-
 static inline void diff_timespec(struct timespec *r, struct timespec *a,
 				 struct timespec *b)
 {
@@ -180,6 +176,8 @@ static void perf_evsel__reset_stat_priv(struct perf_evsel *evsel)
 
 	for (i = 0; i < 3; i++)
 		init_stats(&ps->res_stats[i]);
+
+	perf_stat_evsel_id_init(evsel);
 }
 
 static int perf_evsel__alloc_stat_priv(struct perf_evsel *evsel)
diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
index 6506b3d..8e9f6bb 100644
--- a/tools/perf/util/stat.c
+++ b/tools/perf/util/stat.c
@@ -1,6 +1,6 @@
 #include <math.h>
-
 #include "stat.h"
+#include "evsel.h"
 
 void update_stats(struct stats *stats, u64 val)
 {
@@ -61,3 +61,32 @@ double rel_stddev_stats(double stddev, double avg)
 
 	return pct;
 }
+
+bool __perf_evsel_stat__is(struct perf_evsel *evsel,
+			   enum perf_stat_evsel_id id)
+{
+	struct perf_stat *ps = evsel->priv;
+
+	return ps->id == id;
+}
+
+#define ID(id, name) [PERF_STAT_EVSEL_ID__##id] = #name
+static const char *id_str[PERF_STAT_EVSEL_ID__MAX] = {
+	ID(NONE, x),
+};
+#undef ID
+
+void perf_stat_evsel_id_init(struct perf_evsel *evsel)
+{
+	struct perf_stat *ps = evsel->priv;
+	int i;
+
+	/* ps->id is 0 hence PERF_STAT_EVSEL_ID__NONE by default */
+
+	for (i = 0; i < PERF_STAT_EVSEL_ID__MAX; i++) {
+		if (!strcmp(perf_evsel__name(evsel), id_str[i])) {
+			ps->id = i;
+			break;
+		}
+	}
+}
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index 5667fc3..f4136cf 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -9,6 +9,16 @@ struct stats
 	u64 max, min;
 };
 
+enum perf_stat_evsel_id {
+	PERF_STAT_EVSEL_ID__NONE = 0,
+	PERF_STAT_EVSEL_ID__MAX,
+};
+
+struct perf_stat {
+	struct stats		res_stats[3];
+	enum perf_stat_evsel_id	id;
+};
+
 void update_stats(struct stats *stats, u64 val);
 double avg_stats(struct stats *stats);
 double stddev_stats(struct stats *stats);
@@ -22,4 +32,14 @@ static inline void init_stats(struct stats *stats)
 	stats->min  = (u64) -1;
 	stats->max  = 0;
 }
+
+struct perf_evsel;
+bool __perf_evsel_stat__is(struct perf_evsel *evsel,
+			   enum perf_stat_evsel_id id);
+
+#define perf_stat_evsel__is(evsel, id) \
+	__perf_evsel_stat__is(evsel, PERF_STAT_EVSEL_ID__ ## id)
+
+void perf_stat_evsel_id_init(struct perf_evsel *evsel);
+
 #endif

[tip:perf/core] perf stat: Replace transaction event possition check with id check

[tip-bot for Jiri Olsa]

Commit-ID:  4c358d5cf36192f22b8d331779cb92e3ede9cddf
Gitweb:     http://git.kernel.org/tip/4c358d5cf36192f22b8d331779cb92e3ede9cddf
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Wed, 3 Jun 2015 16:25:52 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:30 -0300

perf stat: Replace transaction event possition check with id check

Using perf_stat::id to check for transaction events, instead of current
position based way.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1433341559-31848-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c | 55 ++++++-----------------------------------------
 tools/perf/util/stat.c    |  6 +++++-
 tools/perf/util/stat.h    |  4 ++++
 3 files changed, 16 insertions(+), 49 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index a6ae100..514493d 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -96,17 +96,6 @@ static const char * const transaction_limited_attrs[] = {
 	"}"
 };
 
-/* must match transaction_attrs and the beginning limited_attrs */
-enum {
-	T_TASK_CLOCK,
-	T_INSTRUCTIONS,
-	T_CYCLES,
-	T_CYCLES_IN_TX,
-	T_TRANSACTION_START,
-	T_ELISION_START,
-	T_CYCLES_IN_TX_CP,
-};
-
 static struct perf_evlist	*evsel_list;
 
 static struct target target = {
@@ -352,29 +341,6 @@ static inline int nsec_counter(struct perf_evsel *evsel)
 	return 0;
 }
 
-static struct perf_evsel *nth_evsel(int n)
-{
-	static struct perf_evsel **array;
-	static int array_len;
-	struct perf_evsel *ev;
-	int j;
-
-	/* Assumes this only called when evsel_list does not change anymore. */
-	if (!array) {
-		evlist__for_each(evsel_list, ev)
-			array_len++;
-		array = malloc(array_len * sizeof(void *));
-		if (!array)
-			exit(ENOMEM);
-		j = 0;
-		evlist__for_each(evsel_list, ev)
-			array[j++] = ev;
-	}
-	if (n < array_len)
-		return array[n];
-	return NULL;
-}
-
 /*
  * Update various tracking values we maintain to print
  * more semantic information such as miss/hit ratios,
@@ -389,14 +355,11 @@ static void update_shadow_stats(struct perf_evsel *counter, u64 *count,
 		update_stats(&runtime_nsecs_stats[cpu], count[0]);
 	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
 		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
-	else if (transaction_run &&
-		 perf_evsel__cmp(counter, nth_evsel(T_CYCLES_IN_TX)))
+	else if (transaction_run && perf_stat_evsel__is(counter, CYCLES_IN_TX))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (transaction_run &&
-		 perf_evsel__cmp(counter, nth_evsel(T_TRANSACTION_START)))
+	else if (transaction_run && perf_stat_evsel__is(counter, TRANSACTION_START))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (transaction_run &&
-		 perf_evsel__cmp(counter, nth_evsel(T_ELISION_START)))
+	else if (transaction_run && perf_stat_evsel__is(counter, ELISION_START))
 		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
 	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
 		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
@@ -1207,15 +1170,13 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 		} else {
 			fprintf(output, "                                   ");
 		}
-	} else if (transaction_run &&
-		   perf_evsel__cmp(evsel, nth_evsel(T_CYCLES_IN_TX))) {
+	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total)
 			fprintf(output,
 				" #   %5.2f%% transactional cycles   ",
 				100.0 * (avg / total));
-	} else if (transaction_run &&
-		   perf_evsel__cmp(evsel, nth_evsel(T_CYCLES_IN_TX_CP))) {
+	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
 		if (total2 < avg)
@@ -1224,8 +1185,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 			fprintf(output,
 				" #   %5.2f%% aborted cycles         ",
 				100.0 * ((total2-avg) / total));
-	} else if (transaction_run &&
-		   perf_evsel__cmp(evsel, nth_evsel(T_TRANSACTION_START)) &&
+	} else if (transaction_run && perf_stat_evsel__is(evsel, TRANSACTION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
@@ -1234,8 +1194,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 			ratio = total / avg;
 
 		fprintf(output, " # %8.0f cycles / transaction   ", ratio);
-	} else if (transaction_run &&
-		   perf_evsel__cmp(evsel, nth_evsel(T_ELISION_START)) &&
+	} else if (transaction_run && perf_stat_evsel__is(evsel, ELISION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
index 8e9f6bb..60b9282 100644
--- a/tools/perf/util/stat.c
+++ b/tools/perf/util/stat.c
@@ -72,7 +72,11 @@ bool __perf_evsel_stat__is(struct perf_evsel *evsel,
 
 #define ID(id, name) [PERF_STAT_EVSEL_ID__##id] = #name
 static const char *id_str[PERF_STAT_EVSEL_ID__MAX] = {
-	ID(NONE, x),
+	ID(NONE,		x),
+	ID(CYCLES_IN_TX,	cpu/cycles-t/),
+	ID(TRANSACTION_START,	cpu/tx-start/),
+	ID(ELISION_START,	cpu/el-start/),
+	ID(CYCLES_IN_TX_CP,	cpu/cycles-ct/),
 };
 #undef ID
 
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index f4136cf..3df529b 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -11,6 +11,10 @@ struct stats
 
 enum perf_stat_evsel_id {
 	PERF_STAT_EVSEL_ID__NONE = 0,
+	PERF_STAT_EVSEL_ID__CYCLES_IN_TX,
+	PERF_STAT_EVSEL_ID__TRANSACTION_START,
+	PERF_STAT_EVSEL_ID__ELISION_START,
+	PERF_STAT_EVSEL_ID__CYCLES_IN_TX_CP,
 	PERF_STAT_EVSEL_ID__MAX,
 };
 

[tip:perf/core] perf stat: Remove setup_events function

[tip-bot for Jiri Olsa]

Commit-ID:  a454742c1252d6242e00b5f4f6f9e5fbce3859d7
Gitweb:     http://git.kernel.org/tip/a454742c1252d6242e00b5f4f6f9e5fbce3859d7
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Wed, 3 Jun 2015 16:25:53 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:30 -0300

perf stat: Remove setup_events function

We can use already existing parse_events interface.

Both transaction_attrs and transaction_limited_attrs are changed to be
single strings.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1433341559-31848-4-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c | 27 +++++++--------------------
 1 file changed, 7 insertions(+), 20 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 514493d..0c0071c 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -73,8 +73,8 @@ static void print_counter(struct perf_evsel *counter, char *prefix);
 static void print_aggr(char *prefix);
 
 /* Default events used for perf stat -T */
-static const char * const transaction_attrs[] = {
-	"task-clock",
+static const char *transaction_attrs = {
+	"task-clock,"
 	"{"
 	"instructions,"
 	"cycles,"
@@ -86,8 +86,8 @@ static const char * const transaction_attrs[] = {
 };
 
 /* More limited version when the CPU does not have all events. */
-static const char * const transaction_limited_attrs[] = {
-	"task-clock",
+static const char * transaction_limited_attrs = {
+	"task-clock,"
 	"{"
 	"instructions,"
 	"cycles,"
@@ -1533,17 +1533,6 @@ static int perf_stat_init_aggr_mode(void)
 	return 0;
 }
 
-static int setup_events(const char * const *attrs, unsigned len)
-{
-	unsigned i;
-
-	for (i = 0; i < len; i++) {
-		if (parse_events(evsel_list, attrs[i], NULL))
-			return -1;
-	}
-	return 0;
-}
-
 /*
  * Add default attributes, if there were no attributes specified or
  * if -d/--detailed, -d -d or -d -d -d is used:
@@ -1665,12 +1654,10 @@ static int add_default_attributes(void)
 		int err;
 		if (pmu_have_event("cpu", "cycles-ct") &&
 		    pmu_have_event("cpu", "el-start"))
-			err = setup_events(transaction_attrs,
-					ARRAY_SIZE(transaction_attrs));
+			err = parse_events(evsel_list, transaction_attrs, NULL);
 		else
-			err = setup_events(transaction_limited_attrs,
-				 ARRAY_SIZE(transaction_limited_attrs));
-		if (err < 0) {
+			err = parse_events(evsel_list, transaction_limited_attrs, NULL);
+		if (err) {
 			fprintf(stderr, "Cannot set up transaction events\n");
 			return -1;
 		}

[tip:perf/core] perf stat: Remove transaction_run from shadow update/print code

[tip-bot for Jiri Olsa]

Commit-ID:  3e99e2f5e78e601591dbcf777c67a84aa9ea2ae5
Gitweb:     http://git.kernel.org/tip/3e99e2f5e78e601591dbcf777c67a84aa9ea2ae5
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Wed, 3 Jun 2015 16:25:54 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:30 -0300

perf stat: Remove transaction_run from shadow update/print code

It's no longer needed, because we use nameid to recognize transaction
events.

Keeping it only in stat code to initialize transaction events.

I.e. struct perf_stat::id, accessible via evsel->priv, will be only set
for transaction related events.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1433341559-31848-5-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 0c0071c..b3e08ce 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -355,11 +355,11 @@ static void update_shadow_stats(struct perf_evsel *counter, u64 *count,
 		update_stats(&runtime_nsecs_stats[cpu], count[0]);
 	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
 		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
-	else if (transaction_run && perf_stat_evsel__is(counter, CYCLES_IN_TX))
+	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (transaction_run && perf_stat_evsel__is(counter, TRANSACTION_START))
+	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (transaction_run && perf_stat_evsel__is(counter, ELISION_START))
+	else if (perf_stat_evsel__is(counter, ELISION_START))
 		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
 	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
 		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
@@ -1170,13 +1170,13 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 		} else {
 			fprintf(output, "                                   ");
 		}
-	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
+	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total)
 			fprintf(output,
 				" #   %5.2f%% transactional cycles   ",
 				100.0 * (avg / total));
-	} else if (transaction_run && perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
+	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
 		if (total2 < avg)
@@ -1185,7 +1185,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 			fprintf(output,
 				" #   %5.2f%% aborted cycles         ",
 				100.0 * ((total2-avg) / total));
-	} else if (transaction_run && perf_stat_evsel__is(evsel, TRANSACTION_START) &&
+	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
@@ -1194,7 +1194,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 			ratio = total / avg;
 
 		fprintf(output, " # %8.0f cycles / transaction   ", ratio);
-	} else if (transaction_run && perf_stat_evsel__is(evsel, ELISION_START) &&
+	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);

[tip:perf/core] perf stat: Introduce reset_shadow_stats function

[tip-bot for Jiri Olsa]

Commit-ID:  1eda3b2144391e1ec9e1870bb32d5216ac7b384c
Gitweb:     http://git.kernel.org/tip/1eda3b2144391e1ec9e1870bb32d5216ac7b384c
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Wed, 3 Jun 2015 16:25:55 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:30 -0300

perf stat: Introduce reset_shadow_stats function

Move shadow counters reset code into separate function
as preparation for moving it into its own object.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1433341559-31848-6-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c | 21 +++++++++++++--------
 1 file changed, 13 insertions(+), 8 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index b3e08ce..fc85e6b 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -279,15 +279,8 @@ static int evsel_context(struct perf_evsel *evsel)
 	return ctx;
 }
 
-static void perf_stat__reset_stats(struct perf_evlist *evlist)
+static void reset_shadow_stats(void)
 {
-	struct perf_evsel *evsel;
-
-	evlist__for_each(evlist, evsel) {
-		perf_evsel__reset_stat_priv(evsel);
-		perf_evsel__reset_counts(evsel, perf_evsel__nr_cpus(evsel));
-	}
-
 	memset(runtime_nsecs_stats, 0, sizeof(runtime_nsecs_stats));
 	memset(runtime_cycles_stats, 0, sizeof(runtime_cycles_stats));
 	memset(runtime_stalled_cycles_front_stats, 0, sizeof(runtime_stalled_cycles_front_stats));
@@ -307,6 +300,18 @@ static void perf_stat__reset_stats(struct perf_evlist *evlist)
 	memset(&walltime_nsecs_stats, 0, sizeof(walltime_nsecs_stats));
 }
 
+static void perf_stat__reset_stats(struct perf_evlist *evlist)
+{
+	struct perf_evsel *evsel;
+
+	evlist__for_each(evlist, evsel) {
+		perf_evsel__reset_stat_priv(evsel);
+		perf_evsel__reset_counts(evsel, perf_evsel__nr_cpus(evsel));
+	}
+
+	reset_shadow_stats();
+}
+
 static int create_perf_stat_counter(struct perf_evsel *evsel)
 {
 	struct perf_event_attr *attr = &evsel->attr;

[tip:perf/core] perf stat: Introduce print_shadow_stats function

[tip-bot for Jiri Olsa]

Commit-ID:  556b1fb7f9c1a9fd43ea4dacd5d14ec39ac6296a
Gitweb:     http://git.kernel.org/tip/556b1fb7f9c1a9fd43ea4dacd5d14ec39ac6296a
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Wed, 3 Jun 2015 16:25:56 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:31 -0300

perf stat: Introduce print_shadow_stats function

Move shadow counters display code into separate function as preparation
for moving it into its own object.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1433341559-31848-7-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c | 72 +++++++++++++++++++++++++----------------------
 1 file changed, 39 insertions(+), 33 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index fc85e6b..2ff2e22 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1059,43 +1059,11 @@ static void print_ll_cache_misses(int cpu,
 	fprintf(output, " of all LL-cache hits   ");
 }
 
-static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
+static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 {
 	double total, ratio = 0.0, total2;
-	double sc =  evsel->scale;
-	const char *fmt;
-	int cpu = cpu_map__id_to_cpu(id);
 	int ctx = evsel_context(evsel);
 
-	if (csv_output) {
-		fmt = sc != 1.0 ?  "%.2f%s" : "%.0f%s";
-	} else {
-		if (big_num)
-			fmt = sc != 1.0 ? "%'18.2f%s" : "%'18.0f%s";
-		else
-			fmt = sc != 1.0 ? "%18.2f%s" : "%18.0f%s";
-	}
-
-	aggr_printout(evsel, id, nr);
-
-	if (aggr_mode == AGGR_GLOBAL)
-		cpu = 0;
-
-	fprintf(output, fmt, avg, csv_sep);
-
-	if (evsel->unit)
-		fprintf(output, "%-*s%s",
-			csv_output ? 0 : unit_width,
-			evsel->unit, csv_sep);
-
-	fprintf(output, "%-*s", csv_output ? 0 : 25, perf_evsel__name(evsel));
-
-	if (evsel->cgrp)
-		fprintf(output, "%s%s", csv_sep, evsel->cgrp->name);
-
-	if (csv_output || interval)
-		return;
-
 	if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total) {
@@ -1226,6 +1194,44 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 	}
 }
 
+static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
+{
+	double sc =  evsel->scale;
+	const char *fmt;
+	int cpu = cpu_map__id_to_cpu(id);
+
+	if (csv_output) {
+		fmt = sc != 1.0 ?  "%.2f%s" : "%.0f%s";
+	} else {
+		if (big_num)
+			fmt = sc != 1.0 ? "%'18.2f%s" : "%'18.0f%s";
+		else
+			fmt = sc != 1.0 ? "%18.2f%s" : "%18.0f%s";
+	}
+
+	aggr_printout(evsel, id, nr);
+
+	if (aggr_mode == AGGR_GLOBAL)
+		cpu = 0;
+
+	fprintf(output, fmt, avg, csv_sep);
+
+	if (evsel->unit)
+		fprintf(output, "%-*s%s",
+			csv_output ? 0 : unit_width,
+			evsel->unit, csv_sep);
+
+	fprintf(output, "%-*s", csv_output ? 0 : 25, perf_evsel__name(evsel));
+
+	if (evsel->cgrp)
+		fprintf(output, "%s%s", csv_sep, evsel->cgrp->name);
+
+	if (csv_output || interval)
+		return;
+
+	print_shadow_stats(evsel, avg, cpu);
+}
+
 static void print_aggr(char *prefix)
 {
 	struct perf_evsel *counter;

[tip:perf/core] perf stat: Add output file argument to print_shadow_stats function

[tip-bot for Jiri Olsa]

Commit-ID:  4d982740cd598bdd876d9a396cc919724af32bc9
Gitweb:     http://git.kernel.org/tip/4d982740cd598bdd876d9a396cc919724af32bc9
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Wed, 3 Jun 2015 16:25:57 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:31 -0300

perf stat: Add output file argument to print_shadow_stats function

As preparation for moving shadow counters code into its own object.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1433341559-31848-8-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c | 112 +++++++++++++++++++++++-----------------------
 1 file changed, 56 insertions(+), 56 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 2ff2e22..14a75dd 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -899,7 +899,7 @@ static const char *get_ratio_color(enum grc_type type, double ratio)
 	return color;
 }
 
-static void print_stalled_cycles_frontend(int cpu,
+static void print_stalled_cycles_frontend(FILE *out, int cpu,
 					  struct perf_evsel *evsel
 					  __maybe_unused, double avg)
 {
@@ -914,12 +914,12 @@ static void print_stalled_cycles_frontend(int cpu,
 
 	color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " frontend cycles idle   ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " frontend cycles idle   ");
 }
 
-static void print_stalled_cycles_backend(int cpu,
+static void print_stalled_cycles_backend(FILE *out, int cpu,
 					 struct perf_evsel *evsel
 					 __maybe_unused, double avg)
 {
@@ -934,12 +934,12 @@ static void print_stalled_cycles_backend(int cpu,
 
 	color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " backend  cycles idle   ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " backend  cycles idle   ");
 }
 
-static void print_branch_misses(int cpu,
+static void print_branch_misses(FILE *out, int cpu,
 				struct perf_evsel *evsel __maybe_unused,
 				double avg)
 {
@@ -954,12 +954,12 @@ static void print_branch_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all branches        ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all branches        ");
 }
 
-static void print_l1_dcache_misses(int cpu,
+static void print_l1_dcache_misses(FILE *out, int cpu,
 				   struct perf_evsel *evsel __maybe_unused,
 				   double avg)
 {
@@ -974,12 +974,12 @@ static void print_l1_dcache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all L1-dcache hits  ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all L1-dcache hits  ");
 }
 
-static void print_l1_icache_misses(int cpu,
+static void print_l1_icache_misses(FILE *out, int cpu,
 				   struct perf_evsel *evsel __maybe_unused,
 				   double avg)
 {
@@ -994,12 +994,12 @@ static void print_l1_icache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all L1-icache hits  ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all L1-icache hits  ");
 }
 
-static void print_dtlb_cache_misses(int cpu,
+static void print_dtlb_cache_misses(FILE *out, int cpu,
 				    struct perf_evsel *evsel __maybe_unused,
 				    double avg)
 {
@@ -1014,12 +1014,12 @@ static void print_dtlb_cache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all dTLB cache hits ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all dTLB cache hits ");
 }
 
-static void print_itlb_cache_misses(int cpu,
+static void print_itlb_cache_misses(FILE *out, int cpu,
 				    struct perf_evsel *evsel __maybe_unused,
 				    double avg)
 {
@@ -1034,12 +1034,12 @@ static void print_itlb_cache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all iTLB cache hits ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all iTLB cache hits ");
 }
 
-static void print_ll_cache_misses(int cpu,
+static void print_ll_cache_misses(FILE *out, int cpu,
 				  struct perf_evsel *evsel __maybe_unused,
 				  double avg)
 {
@@ -1054,12 +1054,12 @@ static void print_ll_cache_misses(int cpu,
 
 	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
 
-	fprintf(output, " #  ");
-	color_fprintf(output, color, "%6.2f%%", ratio);
-	fprintf(output, " of all LL-cache hits   ");
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all LL-cache hits   ");
 }
 
-static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
+static void print_shadow_stats(FILE *out, struct perf_evsel *evsel, double avg, int cpu)
 {
 	double total, ratio = 0.0, total2;
 	int ctx = evsel_context(evsel);
@@ -1068,59 +1068,59 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total) {
 			ratio = avg / total;
-			fprintf(output, " #   %5.2f  insns per cycle        ", ratio);
+			fprintf(out, " #   %5.2f  insns per cycle        ", ratio);
 		} else {
-			fprintf(output, "                                   ");
+			fprintf(out, "                                   ");
 		}
 		total = avg_stats(&runtime_stalled_cycles_front_stats[ctx][cpu]);
 		total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[ctx][cpu]));
 
 		if (total && avg) {
 			ratio = total / avg;
-			fprintf(output, "\n");
+			fprintf(out, "\n");
 			if (aggr_mode == AGGR_NONE)
-				fprintf(output, "        ");
-			fprintf(output, "                                                  #   %5.2f  stalled cycles per insn", ratio);
+				fprintf(out, "        ");
+			fprintf(out, "                                                  #   %5.2f  stalled cycles per insn", ratio);
 		}
 
 	} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
 			runtime_branches_stats[ctx][cpu].n != 0) {
-		print_branch_misses(cpu, evsel, avg);
+		print_branch_misses(out, 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[ctx][cpu].n != 0) {
-		print_l1_dcache_misses(cpu, evsel, avg);
+		print_l1_dcache_misses(out, 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[ctx][cpu].n != 0) {
-		print_l1_icache_misses(cpu, evsel, avg);
+		print_l1_icache_misses(out, 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[ctx][cpu].n != 0) {
-		print_dtlb_cache_misses(cpu, evsel, avg);
+		print_dtlb_cache_misses(out, 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[ctx][cpu].n != 0) {
-		print_itlb_cache_misses(cpu, evsel, avg);
+		print_itlb_cache_misses(out, 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[ctx][cpu].n != 0) {
-		print_ll_cache_misses(cpu, evsel, avg);
+		print_ll_cache_misses(out, cpu, evsel, avg);
 	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
 			runtime_cacherefs_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cacherefs_stats[ctx][cpu]);
@@ -1128,25 +1128,25 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		if (total)
 			ratio = avg * 100 / total;
 
-		fprintf(output, " # %8.3f %% of all cache refs    ", ratio);
+		fprintf(out, " # %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);
+		print_stalled_cycles_frontend(out, cpu, evsel, avg);
 	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
-		print_stalled_cycles_backend(cpu, evsel, avg);
+		print_stalled_cycles_backend(out, cpu, evsel, avg);
 	} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
 		total = avg_stats(&runtime_nsecs_stats[cpu]);
 
 		if (total) {
 			ratio = avg / total;
-			fprintf(output, " # %8.3f GHz                    ", ratio);
+			fprintf(out, " # %8.3f GHz                    ", ratio);
 		} else {
-			fprintf(output, "                                   ");
+			fprintf(out, "                                   ");
 		}
 	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
 		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
 		if (total)
-			fprintf(output,
+			fprintf(out,
 				" #   %5.2f%% transactional cycles   ",
 				100.0 * (avg / total));
 	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
@@ -1155,7 +1155,7 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		if (total2 < avg)
 			total2 = avg;
 		if (total)
-			fprintf(output,
+			fprintf(out,
 				" #   %5.2f%% aborted cycles         ",
 				100.0 * ((total2-avg) / total));
 	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
@@ -1166,7 +1166,7 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		if (total)
 			ratio = total / avg;
 
-		fprintf(output, " # %8.0f cycles / transaction   ", ratio);
+		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
 	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
 		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
@@ -1175,7 +1175,7 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 		if (total)
 			ratio = total / avg;
 
-		fprintf(output, " # %8.0f cycles / elision       ", ratio);
+		fprintf(out, " # %8.0f cycles / elision       ", ratio);
 	} else if (runtime_nsecs_stats[cpu].n != 0) {
 		char unit = 'M';
 
@@ -1188,9 +1188,9 @@ static void print_shadow_stats(struct perf_evsel *evsel, double avg, int cpu)
 			unit = 'K';
 		}
 
-		fprintf(output, " # %8.3f %c/sec                  ", ratio, unit);
+		fprintf(out, " # %8.3f %c/sec                  ", ratio, unit);
 	} else {
-		fprintf(output, "                                   ");
+		fprintf(out, "                                   ");
 	}
 }
 
@@ -1229,7 +1229,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 	if (csv_output || interval)
 		return;
 
-	print_shadow_stats(evsel, avg, cpu);
+	print_shadow_stats(output, evsel, avg, cpu);
 }
 
 static void print_aggr(char *prefix)

[tip:perf/core] perf stat: Add aggr_mode argument to print_shadow_stats function

[tip-bot for Jiri Olsa]

Commit-ID:  7a23f57c89cec0e6d3189d420d992902d4465ff4
Gitweb:     http://git.kernel.org/tip/7a23f57c89cec0e6d3189d420d992902d4465ff4
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Wed, 3 Jun 2015 16:25:58 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:31 -0300

perf stat: Add aggr_mode argument to print_shadow_stats function

As preparation for moving shadow counters code into its own object.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1433341559-31848-9-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 14a75dd..50918dc 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1059,7 +1059,8 @@ static void print_ll_cache_misses(FILE *out, int cpu,
 	fprintf(out, " of all LL-cache hits   ");
 }
 
-static void print_shadow_stats(FILE *out, struct perf_evsel *evsel, double avg, int cpu)
+static void print_shadow_stats(FILE *out, struct perf_evsel *evsel,
+			       double avg, int cpu, enum aggr_mode aggr)
 {
 	double total, ratio = 0.0, total2;
 	int ctx = evsel_context(evsel);
@@ -1078,7 +1079,7 @@ static void print_shadow_stats(FILE *out, struct perf_evsel *evsel, double avg,
 		if (total && avg) {
 			ratio = total / avg;
 			fprintf(out, "\n");
-			if (aggr_mode == AGGR_NONE)
+			if (aggr == AGGR_NONE)
 				fprintf(out, "        ");
 			fprintf(out, "                                                  #   %5.2f  stalled cycles per insn", ratio);
 		}
@@ -1229,7 +1230,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 	if (csv_output || interval)
 		return;
 
-	print_shadow_stats(output, evsel, avg, cpu);
+	print_shadow_stats(output, evsel, avg, cpu, aggr_mode);
 }
 
 static void print_aggr(char *prefix)

[tip:perf/core] perf stat: Move shadow stat counters into separate object

[tip-bot for Jiri Olsa]

Commit-ID:  f87027b9689d591ec22720944563a2d43ec835c4
Gitweb:     http://git.kernel.org/tip/f87027b9689d591ec22720944563a2d43ec835c4
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Wed, 3 Jun 2015 16:25:59 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Mon, 8 Jun 2015 10:30:31 -0300

perf stat: Move shadow stat counters into separate object

Separating shadow counters code into separate object as a cleanup, but
mainly for upcomming changes, so could use it from script command
context.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1433341559-31848-10-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c     | 444 +-----------------------------------------
 tools/perf/util/Build         |   1 +
 tools/perf/util/stat-shadow.c | 434 +++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/stat.h        |  16 ++
 4 files changed, 455 insertions(+), 440 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 50918dc..ff3d258 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -102,13 +102,6 @@ static struct target target = {
 	.uid	= UINT_MAX,
 };
 
-enum aggr_mode {
-	AGGR_NONE,
-	AGGR_GLOBAL,
-	AGGR_SOCKET,
-	AGGR_CORE,
-};
-
 static int			run_count			=  1;
 static bool			no_inherit			= false;
 static bool			scale				=  true;
@@ -234,72 +227,6 @@ out_free:
 	return -1;
 }
 
-enum {
-	CTX_BIT_USER	= 1 << 0,
-	CTX_BIT_KERNEL	= 1 << 1,
-	CTX_BIT_HV	= 1 << 2,
-	CTX_BIT_HOST	= 1 << 3,
-	CTX_BIT_IDLE	= 1 << 4,
-	CTX_BIT_MAX	= 1 << 5,
-};
-
-#define NUM_CTX CTX_BIT_MAX
-
-static struct stats runtime_nsecs_stats[MAX_NR_CPUS];
-static struct stats runtime_cycles_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_stalled_cycles_front_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_stalled_cycles_back_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_branches_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_cacherefs_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_l1_dcache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_l1_icache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_ll_cache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_itlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_dtlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_cycles_in_tx_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats walltime_nsecs_stats;
-static struct stats runtime_transaction_stats[NUM_CTX][MAX_NR_CPUS];
-static struct stats runtime_elision_stats[NUM_CTX][MAX_NR_CPUS];
-
-static int evsel_context(struct perf_evsel *evsel)
-{
-	int ctx = 0;
-
-	if (evsel->attr.exclude_kernel)
-		ctx |= CTX_BIT_KERNEL;
-	if (evsel->attr.exclude_user)
-		ctx |= CTX_BIT_USER;
-	if (evsel->attr.exclude_hv)
-		ctx |= CTX_BIT_HV;
-	if (evsel->attr.exclude_host)
-		ctx |= CTX_BIT_HOST;
-	if (evsel->attr.exclude_idle)
-		ctx |= CTX_BIT_IDLE;
-
-	return ctx;
-}
-
-static void reset_shadow_stats(void)
-{
-	memset(runtime_nsecs_stats, 0, sizeof(runtime_nsecs_stats));
-	memset(runtime_cycles_stats, 0, sizeof(runtime_cycles_stats));
-	memset(runtime_stalled_cycles_front_stats, 0, sizeof(runtime_stalled_cycles_front_stats));
-	memset(runtime_stalled_cycles_back_stats, 0, sizeof(runtime_stalled_cycles_back_stats));
-	memset(runtime_branches_stats, 0, sizeof(runtime_branches_stats));
-	memset(runtime_cacherefs_stats, 0, sizeof(runtime_cacherefs_stats));
-	memset(runtime_l1_dcache_stats, 0, sizeof(runtime_l1_dcache_stats));
-	memset(runtime_l1_icache_stats, 0, sizeof(runtime_l1_icache_stats));
-	memset(runtime_ll_cache_stats, 0, sizeof(runtime_ll_cache_stats));
-	memset(runtime_itlb_cache_stats, 0, sizeof(runtime_itlb_cache_stats));
-	memset(runtime_dtlb_cache_stats, 0, sizeof(runtime_dtlb_cache_stats));
-	memset(runtime_cycles_in_tx_stats, 0,
-			sizeof(runtime_cycles_in_tx_stats));
-	memset(runtime_transaction_stats, 0,
-		sizeof(runtime_transaction_stats));
-	memset(runtime_elision_stats, 0, sizeof(runtime_elision_stats));
-	memset(&walltime_nsecs_stats, 0, sizeof(walltime_nsecs_stats));
-}
-
 static void perf_stat__reset_stats(struct perf_evlist *evlist)
 {
 	struct perf_evsel *evsel;
@@ -309,7 +236,7 @@ static void perf_stat__reset_stats(struct perf_evlist *evlist)
 		perf_evsel__reset_counts(evsel, perf_evsel__nr_cpus(evsel));
 	}
 
-	reset_shadow_stats();
+	perf_stat__reset_shadow_stats();
 }
 
 static int create_perf_stat_counter(struct perf_evsel *evsel)
@@ -346,46 +273,6 @@ static inline int nsec_counter(struct perf_evsel *evsel)
 	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,
-				int cpu)
-{
-	int ctx = evsel_context(counter);
-
-	if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
-		update_stats(&runtime_nsecs_stats[cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
-		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
-	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
-		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
-		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
-	else if (perf_stat_evsel__is(counter, ELISION_START))
-		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
-		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
-		update_stats(&runtime_stalled_cycles_back_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
-		update_stats(&runtime_branches_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
-		update_stats(&runtime_cacherefs_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
-		update_stats(&runtime_l1_dcache_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1I))
-		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_LL))
-		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_DTLB))
-		update_stats(&runtime_dtlb_cache_stats[ctx][cpu], count[0]);
-	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_ITLB))
-		update_stats(&runtime_itlb_cache_stats[ctx][cpu], count[0]);
-}
-
 static void zero_per_pkg(struct perf_evsel *counter)
 {
 	if (counter->per_pkg_mask)
@@ -446,7 +333,7 @@ static int read_cb(struct perf_evsel *evsel, int cpu, int thread __maybe_unused,
 		perf_counts_values__scale(count, scale, NULL);
 		evsel->counts->cpu[cpu] = *count;
 		if (aggr_mode == AGGR_NONE)
-			update_shadow_stats(evsel, count->values, cpu);
+			perf_stat__update_shadow_stats(evsel, count->values, cpu);
 		break;
 	case AGGR_GLOBAL:
 		aggr->val += count->val;
@@ -494,7 +381,7 @@ static int read_counter_aggr(struct perf_evsel *counter)
 	/*
 	 * Save the full runtime - to allow normalization during printout:
 	 */
-	update_shadow_stats(counter, count, 0);
+	perf_stat__update_shadow_stats(counter, count, 0);
 
 	return 0;
 }
@@ -872,329 +759,6 @@ static void nsec_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 		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(FILE *out, int cpu,
-					  struct perf_evsel *evsel
-					  __maybe_unused, double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " frontend cycles idle   ");
-}
-
-static void print_stalled_cycles_backend(FILE *out, int cpu,
-					 struct perf_evsel *evsel
-					 __maybe_unused, double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " backend  cycles idle   ");
-}
-
-static void print_branch_misses(FILE *out, int cpu,
-				struct perf_evsel *evsel __maybe_unused,
-				double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_branches_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all branches        ");
-}
-
-static void print_l1_dcache_misses(FILE *out, int cpu,
-				   struct perf_evsel *evsel __maybe_unused,
-				   double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_l1_dcache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all L1-dcache hits  ");
-}
-
-static void print_l1_icache_misses(FILE *out, int cpu,
-				   struct perf_evsel *evsel __maybe_unused,
-				   double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_l1_icache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all L1-icache hits  ");
-}
-
-static void print_dtlb_cache_misses(FILE *out, int cpu,
-				    struct perf_evsel *evsel __maybe_unused,
-				    double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_dtlb_cache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all dTLB cache hits ");
-}
-
-static void print_itlb_cache_misses(FILE *out, int cpu,
-				    struct perf_evsel *evsel __maybe_unused,
-				    double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_itlb_cache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all iTLB cache hits ");
-}
-
-static void print_ll_cache_misses(FILE *out, int cpu,
-				  struct perf_evsel *evsel __maybe_unused,
-				  double avg)
-{
-	double total, ratio = 0.0;
-	const char *color;
-	int ctx = evsel_context(evsel);
-
-	total = avg_stats(&runtime_ll_cache_stats[ctx][cpu]);
-
-	if (total)
-		ratio = avg / total * 100.0;
-
-	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
-
-	fprintf(out, " #  ");
-	color_fprintf(out, color, "%6.2f%%", ratio);
-	fprintf(out, " of all LL-cache hits   ");
-}
-
-static void print_shadow_stats(FILE *out, struct perf_evsel *evsel,
-			       double avg, int cpu, enum aggr_mode aggr)
-{
-	double total, ratio = 0.0, total2;
-	int ctx = evsel_context(evsel);
-
-	if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
-		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-		if (total) {
-			ratio = avg / total;
-			fprintf(out, " #   %5.2f  insns per cycle        ", ratio);
-		} else {
-			fprintf(out, "                                   ");
-		}
-		total = avg_stats(&runtime_stalled_cycles_front_stats[ctx][cpu]);
-		total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[ctx][cpu]));
-
-		if (total && avg) {
-			ratio = total / avg;
-			fprintf(out, "\n");
-			if (aggr == AGGR_NONE)
-				fprintf(out, "        ");
-			fprintf(out, "                                                  #   %5.2f  stalled cycles per insn", ratio);
-		}
-
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
-			runtime_branches_stats[ctx][cpu].n != 0) {
-		print_branch_misses(out, 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[ctx][cpu].n != 0) {
-		print_l1_dcache_misses(out, 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[ctx][cpu].n != 0) {
-		print_l1_icache_misses(out, 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[ctx][cpu].n != 0) {
-		print_dtlb_cache_misses(out, 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[ctx][cpu].n != 0) {
-		print_itlb_cache_misses(out, 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[ctx][cpu].n != 0) {
-		print_ll_cache_misses(out, cpu, evsel, avg);
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
-			runtime_cacherefs_stats[ctx][cpu].n != 0) {
-		total = avg_stats(&runtime_cacherefs_stats[ctx][cpu]);
-
-		if (total)
-			ratio = avg * 100 / total;
-
-		fprintf(out, " # %8.3f %% of all cache refs    ", ratio);
-
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
-		print_stalled_cycles_frontend(out, cpu, evsel, avg);
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
-		print_stalled_cycles_backend(out, cpu, evsel, avg);
-	} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
-		total = avg_stats(&runtime_nsecs_stats[cpu]);
-
-		if (total) {
-			ratio = avg / total;
-			fprintf(out, " # %8.3f GHz                    ", ratio);
-		} else {
-			fprintf(out, "                                   ");
-		}
-	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
-		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-		if (total)
-			fprintf(out,
-				" #   %5.2f%% transactional cycles   ",
-				100.0 * (avg / total));
-	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
-		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
-		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
-		if (total2 < avg)
-			total2 = avg;
-		if (total)
-			fprintf(out,
-				" #   %5.2f%% aborted cycles         ",
-				100.0 * ((total2-avg) / total));
-	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
-		   avg > 0 &&
-		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
-		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
-
-		if (total)
-			ratio = total / avg;
-
-		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
-	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
-		   avg > 0 &&
-		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
-		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
-
-		if (total)
-			ratio = total / avg;
-
-		fprintf(out, " # %8.0f cycles / elision       ", 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(out, " # %8.3f %c/sec                  ", ratio, unit);
-	} else {
-		fprintf(out, "                                   ");
-	}
-}
-
 static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 {
 	double sc =  evsel->scale;
@@ -1230,7 +794,7 @@ static void abs_printout(int id, int nr, struct perf_evsel *evsel, double avg)
 	if (csv_output || interval)
 		return;
 
-	print_shadow_stats(output, evsel, avg, cpu, aggr_mode);
+	perf_stat__print_shadow_stats(output, evsel, avg, cpu, aggr_mode);
 }
 
 static void print_aggr(char *prefix)
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index e4b676d..586a59d 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -68,6 +68,7 @@ libperf-y += rblist.o
 libperf-y += intlist.o
 libperf-y += vdso.o
 libperf-y += stat.o
+libperf-y += stat-shadow.o
 libperf-y += record.o
 libperf-y += srcline.o
 libperf-y += data.o
diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
new file mode 100644
index 0000000..53e8bb7
--- /dev/null
+++ b/tools/perf/util/stat-shadow.c
@@ -0,0 +1,434 @@
+#include <stdio.h>
+#include "evsel.h"
+#include "stat.h"
+#include "color.h"
+
+enum {
+	CTX_BIT_USER	= 1 << 0,
+	CTX_BIT_KERNEL	= 1 << 1,
+	CTX_BIT_HV	= 1 << 2,
+	CTX_BIT_HOST	= 1 << 3,
+	CTX_BIT_IDLE	= 1 << 4,
+	CTX_BIT_MAX	= 1 << 5,
+};
+
+#define NUM_CTX CTX_BIT_MAX
+
+static struct stats runtime_nsecs_stats[MAX_NR_CPUS];
+static struct stats runtime_cycles_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_stalled_cycles_front_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_stalled_cycles_back_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_branches_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_cacherefs_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_l1_dcache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_l1_icache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_ll_cache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_itlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_dtlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_cycles_in_tx_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_transaction_stats[NUM_CTX][MAX_NR_CPUS];
+static struct stats runtime_elision_stats[NUM_CTX][MAX_NR_CPUS];
+
+struct stats walltime_nsecs_stats;
+
+static int evsel_context(struct perf_evsel *evsel)
+{
+	int ctx = 0;
+
+	if (evsel->attr.exclude_kernel)
+		ctx |= CTX_BIT_KERNEL;
+	if (evsel->attr.exclude_user)
+		ctx |= CTX_BIT_USER;
+	if (evsel->attr.exclude_hv)
+		ctx |= CTX_BIT_HV;
+	if (evsel->attr.exclude_host)
+		ctx |= CTX_BIT_HOST;
+	if (evsel->attr.exclude_idle)
+		ctx |= CTX_BIT_IDLE;
+
+	return ctx;
+}
+
+void perf_stat__reset_shadow_stats(void)
+{
+	memset(runtime_nsecs_stats, 0, sizeof(runtime_nsecs_stats));
+	memset(runtime_cycles_stats, 0, sizeof(runtime_cycles_stats));
+	memset(runtime_stalled_cycles_front_stats, 0, sizeof(runtime_stalled_cycles_front_stats));
+	memset(runtime_stalled_cycles_back_stats, 0, sizeof(runtime_stalled_cycles_back_stats));
+	memset(runtime_branches_stats, 0, sizeof(runtime_branches_stats));
+	memset(runtime_cacherefs_stats, 0, sizeof(runtime_cacherefs_stats));
+	memset(runtime_l1_dcache_stats, 0, sizeof(runtime_l1_dcache_stats));
+	memset(runtime_l1_icache_stats, 0, sizeof(runtime_l1_icache_stats));
+	memset(runtime_ll_cache_stats, 0, sizeof(runtime_ll_cache_stats));
+	memset(runtime_itlb_cache_stats, 0, sizeof(runtime_itlb_cache_stats));
+	memset(runtime_dtlb_cache_stats, 0, sizeof(runtime_dtlb_cache_stats));
+	memset(runtime_cycles_in_tx_stats, 0,
+			sizeof(runtime_cycles_in_tx_stats));
+	memset(runtime_transaction_stats, 0,
+		sizeof(runtime_transaction_stats));
+	memset(runtime_elision_stats, 0, sizeof(runtime_elision_stats));
+	memset(&walltime_nsecs_stats, 0, sizeof(walltime_nsecs_stats));
+}
+
+/*
+ * Update various tracking values we maintain to print
+ * more semantic information such as miss/hit ratios,
+ * instruction rates, etc:
+ */
+void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
+				    int cpu)
+{
+	int ctx = evsel_context(counter);
+
+	if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
+		update_stats(&runtime_nsecs_stats[cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
+		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
+	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
+		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
+	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
+		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
+	else if (perf_stat_evsel__is(counter, ELISION_START))
+		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
+		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
+		update_stats(&runtime_stalled_cycles_back_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
+		update_stats(&runtime_branches_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
+		update_stats(&runtime_cacherefs_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
+		update_stats(&runtime_l1_dcache_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1I))
+		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_LL))
+		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_DTLB))
+		update_stats(&runtime_dtlb_cache_stats[ctx][cpu], count[0]);
+	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_ITLB))
+		update_stats(&runtime_itlb_cache_stats[ctx][cpu], count[0]);
+}
+
+/* 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(FILE *out, int cpu,
+					  struct perf_evsel *evsel
+					  __maybe_unused, double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " frontend cycles idle   ");
+}
+
+static void print_stalled_cycles_backend(FILE *out, int cpu,
+					 struct perf_evsel *evsel
+					 __maybe_unused, double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " backend  cycles idle   ");
+}
+
+static void print_branch_misses(FILE *out, int cpu,
+				struct perf_evsel *evsel __maybe_unused,
+				double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_branches_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all branches        ");
+}
+
+static void print_l1_dcache_misses(FILE *out, int cpu,
+				   struct perf_evsel *evsel __maybe_unused,
+				   double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_l1_dcache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all L1-dcache hits  ");
+}
+
+static void print_l1_icache_misses(FILE *out, int cpu,
+				   struct perf_evsel *evsel __maybe_unused,
+				   double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_l1_icache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all L1-icache hits  ");
+}
+
+static void print_dtlb_cache_misses(FILE *out, int cpu,
+				    struct perf_evsel *evsel __maybe_unused,
+				    double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_dtlb_cache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all dTLB cache hits ");
+}
+
+static void print_itlb_cache_misses(FILE *out, int cpu,
+				    struct perf_evsel *evsel __maybe_unused,
+				    double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_itlb_cache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all iTLB cache hits ");
+}
+
+static void print_ll_cache_misses(FILE *out, int cpu,
+				  struct perf_evsel *evsel __maybe_unused,
+				  double avg)
+{
+	double total, ratio = 0.0;
+	const char *color;
+	int ctx = evsel_context(evsel);
+
+	total = avg_stats(&runtime_ll_cache_stats[ctx][cpu]);
+
+	if (total)
+		ratio = avg / total * 100.0;
+
+	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
+
+	fprintf(out, " #  ");
+	color_fprintf(out, color, "%6.2f%%", ratio);
+	fprintf(out, " of all LL-cache hits   ");
+}
+
+void perf_stat__print_shadow_stats(FILE *out, struct perf_evsel *evsel,
+				   double avg, int cpu, enum aggr_mode aggr)
+{
+	double total, ratio = 0.0, total2;
+	int ctx = evsel_context(evsel);
+
+	if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
+		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+		if (total) {
+			ratio = avg / total;
+			fprintf(out, " #   %5.2f  insns per cycle        ", ratio);
+		} else {
+			fprintf(out, "                                   ");
+		}
+		total = avg_stats(&runtime_stalled_cycles_front_stats[ctx][cpu]);
+		total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[ctx][cpu]));
+
+		if (total && avg) {
+			ratio = total / avg;
+			fprintf(out, "\n");
+			if (aggr == AGGR_NONE)
+				fprintf(out, "        ");
+			fprintf(out, "                                                  #   %5.2f  stalled cycles per insn", ratio);
+		}
+
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
+			runtime_branches_stats[ctx][cpu].n != 0) {
+		print_branch_misses(out, 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[ctx][cpu].n != 0) {
+		print_l1_dcache_misses(out, 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[ctx][cpu].n != 0) {
+		print_l1_icache_misses(out, 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[ctx][cpu].n != 0) {
+		print_dtlb_cache_misses(out, 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[ctx][cpu].n != 0) {
+		print_itlb_cache_misses(out, 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[ctx][cpu].n != 0) {
+		print_ll_cache_misses(out, cpu, evsel, avg);
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
+			runtime_cacherefs_stats[ctx][cpu].n != 0) {
+		total = avg_stats(&runtime_cacherefs_stats[ctx][cpu]);
+
+		if (total)
+			ratio = avg * 100 / total;
+
+		fprintf(out, " # %8.3f %% of all cache refs    ", ratio);
+
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
+		print_stalled_cycles_frontend(out, cpu, evsel, avg);
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
+		print_stalled_cycles_backend(out, cpu, evsel, avg);
+	} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
+		total = avg_stats(&runtime_nsecs_stats[cpu]);
+
+		if (total) {
+			ratio = avg / total;
+			fprintf(out, " # %8.3f GHz                    ", ratio);
+		} else {
+			fprintf(out, "                                   ");
+		}
+	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
+		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+		if (total)
+			fprintf(out,
+				" #   %5.2f%% transactional cycles   ",
+				100.0 * (avg / total));
+	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
+		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
+		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
+		if (total2 < avg)
+			total2 = avg;
+		if (total)
+			fprintf(out,
+				" #   %5.2f%% aborted cycles         ",
+				100.0 * ((total2-avg) / total));
+	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
+		   avg > 0 &&
+		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
+		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
+
+		if (total)
+			ratio = total / avg;
+
+		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
+	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
+		   avg > 0 &&
+		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
+		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
+
+		if (total)
+			ratio = total / avg;
+
+		fprintf(out, " # %8.0f cycles / elision       ", 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(out, " # %8.3f %c/sec                  ", ratio, unit);
+	} else {
+		fprintf(out, "                                   ");
+	}
+}
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index 3df529b..615c779 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -2,6 +2,7 @@
 #define __PERF_STATS_H
 
 #include <linux/types.h>
+#include <stdio.h>
 
 struct stats
 {
@@ -23,6 +24,13 @@ struct perf_stat {
 	enum perf_stat_evsel_id	id;
 };
 
+enum aggr_mode {
+	AGGR_NONE,
+	AGGR_GLOBAL,
+	AGGR_SOCKET,
+	AGGR_CORE,
+};
+
 void update_stats(struct stats *stats, u64 val);
 double avg_stats(struct stats *stats);
 double stddev_stats(struct stats *stats);
@@ -46,4 +54,12 @@ bool __perf_evsel_stat__is(struct perf_evsel *evsel,
 
 void perf_stat_evsel_id_init(struct perf_evsel *evsel);
 
+extern struct stats walltime_nsecs_stats;
+
+void perf_stat__reset_shadow_stats(void);
+void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
+				    int cpu);
+void perf_stat__print_shadow_stats(FILE *out, struct perf_evsel *evsel,
+				   double avg, int cpu, enum aggr_mode aggr);
+
 #endif

Re: [PATCHv3] perf stat: Add id into perf_stat struct

[Jiri Olsa]

On Mon, Jun 08, 2015 at 11:03:40AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Thu, Jun 04, 2015 at 03:50:55PM +0200, Jiri Olsa escreveu:
> > On Wed, Jun 03, 2015 at 04:25:51PM +0200, Jiri Olsa wrote:
> > > We need fast way to identify evsel as transaction event
> > > for shadow counters computation. Currently we are using
> > > possition (in evlist) based way.
> > > 
> > > Adding 'id' into 'struct perf_stat' so it can carry transaction
> > > event ID and we can use it for shadow counters computations.
> > > 
> > > Link: http://lkml.kernel.org/n/tip-9lnvjksibrs0flhkpix2qkm9@git.kernel.org
> > > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > > ---
> > 
> > using perf_evsel__name instead of evsel->name pointer
> 
> It now passes 'perf test' on rhel7/fedora21 systems, but it fails on a
> rhel6 system, I applied it, will test it further on the rhel6 system to
> see what went wrong there, unless you do it first ;-)
> 
> [root@sandy ~]# perf test 15
> 15: struct perf_event_attr setup                             :sh: line
> 1:  5381 Aborted                 (core dumped)
> PERF_TEST_ATTR=/tmp/tmp8ZBLwc /home/acme/bin/perf stat -o
> /tmp/tmp8ZBLwc/perf.data --group -e cycles,instructions kill > /dev/null
> 2>&1
> sh: line 1:  5438 Aborted                 (core dumped)
> PERF_TEST_ATTR=/tmp/tmpnRDADB /home/acme/bin/perf stat -o
> /tmp/tmpnRDADB/perf.data -e '{cycles,instructions}' kill > /dev/null
> 2>&1
>  Ok
> [root@sandy ~]# uname -r
> 2.6.32-504.8.1.el6.x86_64
> [root@sandy ~]# cat /etc/redhat-release 
> Red Hat Enterprise Linux Server release 6.6 (Santiago)
> [root@sandy ~]# 

hum, cant reproduce

[root@ibm-x3650m4-02 perf]# ./perf test 15
15: struct perf_event_attr setup                             : Ok


the server is latest rhel6 kernel (scratch):

[root@ibm-x3650m4-02 perf]# uname -a
Linux ibm-x3650m4-02.lab.eng.brq.redhat.com 2.6.32-556.el6perf_ivb_haswell_1.x86_64 #1 SMP Wed Apr 29 06:00:24 EDT 2015 x86_64 x86_64 x86_64 GNU/Linux
[root@ibm-x3650m4-02 perf]# cat /etc/redhat-release 
Red Hat Enterprise Linux Server release 6.6 (Santiago)


jirka