Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1590049 > unrolled thread

perf: Improve support for uncore JSON event lists

Back to article view | Back to linux.kernel

Contents

  perf: Improve support for uncore JSON event lists Andi Kleen <andi@firstfloor.org> - 2017-03-01 08:00 +0100
    [PATCH 05/10] perf, tools: Special case uncore_ prefix Andi Kleen <andi@firstfloor.org> - 2017-03-01 08:00 +0100
    [PATCH 06/10] perf, tools: Add a simple expression parser for JSON Andi Kleen <andi@firstfloor.org> - 2017-03-01 09:30 +0100
    [PATCH 02/10] perf, tools, stat: Collapse identically named events Andi Kleen <andi@firstfloor.org> - 2017-03-01 09:50 +0100
      Re: [PATCH 02/10] perf, tools, stat: Collapse identically named  events Jiri Olsa <jolsa@redhat.com> - 2017-03-05 19:00 +0100
      Re: [PATCH 02/10] perf, tools, stat: Collapse identically named  events Jiri Olsa <jolsa@redhat.com> - 2017-03-07 11:10 +0100
    [PATCH 08/10] perf, tools, stat: Output JSON MetricExpr metric Andi Kleen <andi@firstfloor.org> - 2017-03-01 12:30 +0100
      Re: [PATCH 08/10] perf, tools, stat: Output JSON MetricExpr metric Jiri Olsa <jolsa@redhat.com> - 2017-03-07 11:10 +0100
    Re: perf: Improve support for uncore JSON event lists Jiri Olsa <jolsa@redhat.com> - 2017-03-07 11:10 +0100

#1590049 — perf: Improve support for uncore JSON event lists

This patch kit further improves support for Intel uncore events in
the Linux perf user tool. The basic support has been already
merged earlier, but this makes it nicer to use.

- Collapse counts from duplicated boxes to make the output
easier to read.
- Support specifying events for multiple duplicated boxes
in an abbreviated format to shorten event specifiers
- Add support for computing Metrics defined in the event lists,
so that the event lists can extend the metrics in perf stat.
This allows to represent many events in an easier to understand
format.

Available from

git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/builtin-json-28

v1: Initial post after being split off to own patchkit
Adding MetricName support and support for more than two events
in expressions.

v2: Address review comments. Move new hunk from refactor
patchkit to patch adding new features. Improve changelogs
slightly.

[toc] | [next] | [standalone]

#1590050 — [PATCH 05/10] perf, tools: Special case uncore_ prefix

From: Andi Kleen <ak@linux.intel.com>

Special case uncore_ prefix in PMU match, to allow for shorter event
uncore specifications.

Before

perf stat -a -e uncore_cbox/event=0x35,umask=0x1,filter_opc=0x19C/ sleep 1

After

perf stat -a -e cbox/event=0x35,umask=0x1,filter_opc=0x19C/ sleep 1

Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
 tools/perf/util/parse-events.y | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/perf/util/parse-events.y b/tools/perf/util/parse-events.y
index 3209f18cd5d8..eed371b971a8 100644
--- a/tools/perf/util/parse-events.y
+++ b/tools/perf/util/parse-events.y
@@ -239,6 +239,9 @@ PE_NAME opt_event_config
 		while ((pmu = perf_pmu__scan(pmu)) != NULL) {
 			char *name = pmu->name;
 
+			if (!strncmp(name, "uncore_", 7) &&
+			    strncmp($1, "uncore_", 7))
+				name += 7;
 			if (!strncmp($1, name, strlen($1))) {
 				if (parse_events_copy_term_list(orig_terms, &terms))
 					YYABORT;
-- 
2.9.3

[toc] | [prev] | [next] | [standalone]

#1590092 — [PATCH 06/10] perf, tools: Add a simple expression parser for JSON

From: Andi Kleen <ak@linux.intel.com>

Add a simple expression parser good enough to parse JSON relation
expressions. The parser is implemented using bison.

This is just intended as an simple parser for internal usage
in the event lists, not the beginning of a "perf scripting language"

v2: Use expr__ prefix instead of expr_
Support multiple free variables for parser
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
 tools/perf/tests/Build          |   1 +
 tools/perf/tests/builtin-test.c |   4 +
 tools/perf/tests/expr.c         |  55 +++++++++++++
 tools/perf/tests/tests.h        |   1 +
 tools/perf/util/Build           |   5 ++
 tools/perf/util/expr.h          |  25 ++++++
 tools/perf/util/expr.y          | 173 ++++++++++++++++++++++++++++++++++++++++
 7 files changed, 264 insertions(+)
 create mode 100644 tools/perf/tests/expr.c
 create mode 100644 tools/perf/util/expr.h
 create mode 100644 tools/perf/util/expr.y

diff --git a/tools/perf/tests/Build b/tools/perf/tests/Build
index 1cb3d9b540e9..af58ebc243ef 100644
--- a/tools/perf/tests/Build
+++ b/tools/perf/tests/Build
@@ -38,6 +38,7 @@ perf-y += cpumap.o
 perf-y += stat.o
 perf-y += event_update.o
 perf-y += event-times.o
+perf-y += expr.o
 perf-y += backward-ring-buffer.o
 perf-y += sdt.o
 perf-y += is_printable_array.o
diff --git a/tools/perf/tests/builtin-test.c b/tools/perf/tests/builtin-test.c
index 37e326bfd2dc..db03e853ba7f 100644
--- a/tools/perf/tests/builtin-test.c
+++ b/tools/perf/tests/builtin-test.c
@@ -44,6 +44,10 @@ static struct test generic_tests[] = {
 		.func = test__parse_events,
 	},
 	{
+		.desc = "Simple expression parser",
+		.func = test__expr,
+	},
+	{
 		.desc = "PERF_RECORD_* events & perf_sample fields",
 		.func = test__PERF_RECORD,
 	},
diff --git a/tools/perf/tests/expr.c b/tools/perf/tests/expr.c
new file mode 100644
index 000000000000..554695c06c5b
--- /dev/null
+++ b/tools/perf/tests/expr.c
@@ -0,0 +1,55 @@
+#include "util/debug.h"
+#include "util/expr.h"
+#include "tests.h"
+
+static int test(struct parse_ctx *ctx, const char *e, double val2)
+{
+	double val;
+
+	if (expr__parse(&val, ctx, &e))
+		TEST_ASSERT_VAL("parse test failed", 0);
+	TEST_ASSERT_VAL("unexpected value", val == val2);
+	return 0;
+}
+
+int test__expr(int subtest __maybe_unused)
+{
+	const char *p;
+	const char **other;
+	double val;
+	int ret;
+	struct parse_ctx ctx;
+	int num_other;
+
+	expr__ctx_init(&ctx);
+	expr__add_id(&ctx, "FOO", 1);
+	expr__add_id(&ctx, "BAR", 2);
+
+	ret = test(&ctx, "1+1", 2);
+	ret |= test(&ctx, "FOO+BAR", 3);
+	ret |= test(&ctx, "(BAR/2)%2", 1);
+	ret |= test(&ctx, "1 - -4",  5);
+	ret |= test(&ctx, "(FOO-1)*2 + (BAR/2)%2 - -4",  5);
+
+	if (ret)
+		return ret;
+
+	p = "FOO/0";
+	ret = expr__parse(&val, &ctx, &p);
+	TEST_ASSERT_VAL("division by zero", ret == 1);
+
+	p = "BAR/";
+	ret = expr__parse(&val, &ctx, &p);
+	TEST_ASSERT_VAL("missing operand", ret == 1);
+
+	TEST_ASSERT_VAL("find other",
+			expr__find_other("FOO + BAR + BAZ + BOZO", "FOO", &other, &num_other) == 0);
+	TEST_ASSERT_VAL("find other", num_other == 3);
+	TEST_ASSERT_VAL("find other", !strcmp(other[0], "BAR"));
+	TEST_ASSERT_VAL("find other", !strcmp(other[1], "BAZ"));
+	TEST_ASSERT_VAL("find other", !strcmp(other[2], "BOZO"));
+	TEST_ASSERT_VAL("find other", other[3] == NULL);
+	free((void *)other);
+
+	return 0;
+}
diff --git a/tools/perf/tests/tests.h b/tools/perf/tests/tests.h
index 1fa9b9d83aa5..631859629403 100644
--- a/tools/perf/tests/tests.h
+++ b/tools/perf/tests/tests.h
@@ -62,6 +62,7 @@ int test__sample_parsing(int subtest);
 int test__keep_tracking(int subtest);
 int test__parse_no_sample_id_all(int subtest);
 int test__dwarf_unwind(int subtest);
+int test__expr(int subtest);
 int test__hists_filter(int subtest);
 int test__mmap_thread_lookup(int subtest);
 int test__thread_mg_share(int subtest);
diff --git a/tools/perf/util/Build b/tools/perf/util/Build
index 5da376bc1afc..3dff5f5960d9 100644
--- a/tools/perf/util/Build
+++ b/tools/perf/util/Build
@@ -88,6 +88,7 @@ libperf-y += mem-events.o
 libperf-y += vsprintf.o
 libperf-y += drv_configs.o
 libperf-y += time-utils.o
+libperf-y += expr-bison.o
 
 libperf-$(CONFIG_LIBBPF) += bpf-loader.o
 libperf-$(CONFIG_BPF_PROLOGUE) += bpf-prologue.o
@@ -140,6 +141,10 @@ $(OUTPUT)util/parse-events-bison.c: util/parse-events.y
 	$(call rule_mkdir)
 	$(Q)$(call echo-cmd,bison)$(BISON) -v util/parse-events.y -d $(PARSER_DEBUG_BISON) -o $@ -p parse_events_
 
+$(OUTPUT)util/expr-bison.c: util/expr.y
+	$(call rule_mkdir)
+	$(Q)$(call echo-cmd,bison)$(BISON) -v util/expr.y -d $(PARSER_DEBUG_BISON) -o $@ -p expr__
+
 $(OUTPUT)util/pmu-flex.c: util/pmu.l $(OUTPUT)util/pmu-bison.c
 	$(call rule_mkdir)
 	$(Q)$(call echo-cmd,flex)$(FLEX) -o $@ --header-file=$(OUTPUT)util/pmu-flex.h util/pmu.l
diff --git a/tools/perf/util/expr.h b/tools/perf/util/expr.h
new file mode 100644
index 000000000000..9c2760a1a96e
--- /dev/null
+++ b/tools/perf/util/expr.h
@@ -0,0 +1,25 @@
+#ifndef PARSE_CTX_H
+#define PARSE_CTX_H 1
+
+#define EXPR_MAX_OTHER 8
+#define MAX_PARSE_ID EXPR_MAX_OTHER
+
+struct parse_id {
+	const char *name;
+	double val;
+};
+
+struct parse_ctx {
+	int num_ids;
+	struct parse_id ids[MAX_PARSE_ID];
+};
+
+void expr__ctx_init(struct parse_ctx *ctx);
+void expr__add_id(struct parse_ctx *ctx, const char *id, double val);
+#ifndef IN_EXPR_Y
+int expr__parse(double *final_val, struct parse_ctx *ctx, const char **pp);
+#endif
+int expr__find_other(const char *p, const char *one, const char ***other,
+		int *num_other);
+
+#endif
diff --git a/tools/perf/util/expr.y b/tools/perf/util/expr.y
new file mode 100644
index 000000000000..a8daa7a93605
--- /dev/null
+++ b/tools/perf/util/expr.y
@@ -0,0 +1,173 @@
+/* Simple expression parser */
+%{
+#include "util.h"
+#include "util/debug.h"
+#define IN_EXPR_Y 1
+#include "expr.h"
+#include <string.h>
+
+#define MAXIDLEN 256
+%}
+
+%define api.pure full
+%parse-param { double *final_val }
+%parse-param { struct parse_ctx *ctx }
+%parse-param { const char **pp }
+%lex-param { const char **pp }
+
+%union {
+	double num;
+	char id[MAXIDLEN+1];
+}
+
+%token <num> NUMBER
+%token <id> ID
+%left '|'
+%left '^'
+%left '&'
+%left '-' '+'
+%left '*' '/' '%'
+%left NEG NOT
+%type <num> expr
+
+%{
+static int expr__lex(YYSTYPE *res, const char **pp);
+
+static void expr__error(double *final_val __maybe_unused,
+		       struct parse_ctx *ctx __maybe_unused,
+		       const char **pp __maybe_unused,
+		       const char *s)
+{
+	pr_debug("%s\n", s);
+}
+
+static int lookup_id(struct parse_ctx *ctx, char *id, double *val)
+{
+	int i;
+
+	for (i = 0; i < ctx->num_ids; i++) {
+		if (!strcasecmp(ctx->ids[i].name, id)) {
+			*val = ctx->ids[i].val;
+			return 0;
+		}
+	}
+	return -1;
+}
+
+%}
+%%
+
+all_expr: expr			{ *final_val = $1; }
+	;
+
+expr:	  NUMBER
+	| ID			{ if (lookup_id(ctx, $1, &$$) < 0) {
+					pr_debug("%s not found", $1);
+					YYABORT;
+				  }
+				}
+	| expr '+' expr		{ $$ = $1 + $3; }
+	| expr '-' expr		{ $$ = $1 - $3; }
+	| expr '*' expr		{ $$ = $1 * $3; }
+	| expr '/' expr		{ if ($3 == 0) YYABORT; $$ = $1 / $3; }
+	| expr '%' expr		{ if ((long)$3 == 0) YYABORT; $$ = (long)$1 % (long)$3; }
+	| '-' expr %prec NEG	{ $$ = -$2; }
+	| '(' expr ')'		{ $$ = $2; }
+	;
+
+%%
+
+static int expr__symbol(YYSTYPE *res, const char *p, const char **pp)
+{
+	char *dst = res->id;
+	const char *s = p;
+
+	while (isalnum(*p) || *p == '_' || *p == '.') {
+		if (p - s >= MAXIDLEN)
+			return -1;
+		*dst++ = *p++;
+	}
+	*dst = 0;
+	*pp = p;
+	return ID;
+}
+
+static int expr__lex(YYSTYPE *res, const char **pp)
+{
+	int tok;
+	const char *s;
+	const char *p = *pp;
+
+	while (isspace(*p))
+		p++;
+	s = p;
+	switch (*p++) {
+	case 'a' ... 'z':
+	case 'A' ... 'Z':
+		return expr__symbol(res, p - 1, pp);
+	case '0' ... '9': case '.':
+		res->num = strtod(s, (char **)&p);
+		tok = NUMBER;
+		break;
+	default:
+		tok = *s;
+		break;
+	}
+	*pp = p;
+	return tok;
+}
+
+/* Caller must make sure id is allocated */
+void expr__add_id(struct parse_ctx *ctx, const char *name, double val)
+{
+	int idx;
+	assert(ctx->num_ids < MAX_PARSE_ID);
+	idx = ctx->num_ids++;
+	ctx->ids[idx].name = name;
+	ctx->ids[idx].val = val;
+}
+
+void expr__ctx_init(struct parse_ctx *ctx)
+{
+	ctx->num_ids = 0;
+}
+
+int expr__find_other(const char *p, const char *one, const char ***other,
+		     int *num_otherp)
+{
+	const char *orig = p;
+	int err = -1;
+	int num_other;
+
+	*other = malloc((EXPR_MAX_OTHER + 1) * sizeof(char *));
+	if (!*other)
+		return -1;
+
+	num_other = 0;
+	for (;;) {
+		YYSTYPE val;
+		int tok = expr__lex(&val, &p);
+		if (tok == 0) {
+			err = 0;
+			break;
+		}
+		if (tok == ID && strcasecmp(one, val.id)) {
+			if (num_other >= EXPR_MAX_OTHER - 1) {
+				pr_debug("Too many extra events in %s\n", orig);
+				break;
+			}
+			(*other)[num_other] = strdup(val.id);
+			if (!(*other)[num_other])
+				return -1;
+			num_other++;
+		}
+	}
+	(*other)[num_other] = NULL;
+	*num_otherp = num_other;
+	if (err) {
+		*num_otherp = 0;
+		free(*other);
+		*other = NULL;
+	}
+	return err;
+}
-- 
2.9.3

[toc] | [prev] | [next] | [standalone]

#1590095 — [PATCH 02/10] perf, tools, stat: Collapse identically named events

From: Andi Kleen <ak@linux.intel.com>

The uncore PMU has a lot of duplicated PMUs for different subsystems.
When expanding an uncore alias we usually end up with a large
number of identically named aliases, which makes perf stat
output difficult to read.

Automatically sum them up in perf stat, unless --no-merge is specified.

This can be default because only the uncores generally have duplicated
aliases. Other PMUs have unique names.

Before:

% perf stat --no-merge -a  -e unc_c_llc_lookup.any sleep 1

 Performance counter stats for 'system wide':

           694,976 Bytes unc_c_llc_lookup.any
           706,304 Bytes unc_c_llc_lookup.any
           956,608 Bytes unc_c_llc_lookup.any
           782,720 Bytes unc_c_llc_lookup.any
           605,696 Bytes unc_c_llc_lookup.any
           442,816 Bytes unc_c_llc_lookup.any
           659,328 Bytes unc_c_llc_lookup.any
           509,312 Bytes unc_c_llc_lookup.any
           263,936 Bytes unc_c_llc_lookup.any
           592,448 Bytes unc_c_llc_lookup.any
           672,448 Bytes unc_c_llc_lookup.any
           608,640 Bytes unc_c_llc_lookup.any
           641,024 Bytes unc_c_llc_lookup.any
           856,896 Bytes unc_c_llc_lookup.any
           808,832 Bytes unc_c_llc_lookup.any
           684,864 Bytes unc_c_llc_lookup.any
           710,464 Bytes unc_c_llc_lookup.any
           538,304 Bytes unc_c_llc_lookup.any

       1.002577660 seconds time elapsed

After:

% perf stat  -a  -e unc_c_llc_lookup.any sleep 1

 Performance counter stats for 'system wide':

         2,685,120 Bytes unc_c_llc_lookup.any

       1.002648032 seconds time elapsed

v2: Split collect_aliases. Rename alias flag.
v3: Make sure unsupported/not counted is always printed.
v4: Factor out callback change into separate patch.
v5: Move check for bad results here
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
 tools/perf/Documentation/perf-stat.txt |  3 +++
 tools/perf/builtin-stat.c              | 41 ++++++++++++++++++++++++++++++++++
 tools/perf/util/evsel.h                |  1 +
 3 files changed, 45 insertions(+)

diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index d96ccd4844df..320d8020bc5b 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -237,6 +237,9 @@ To interpret the results it is usually needed to know on which
 CPUs the workload runs on. If needed the CPUs can be forced using
 taskset.
 
+--no-merge::
+Do not merge results from same PMUs.
+
 EXAMPLES
 --------
 
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index c71916eaa255..389c8e457bf0 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -140,6 +140,7 @@ static unsigned int		unit_width			= 4; /* strlen("unit") */
 static bool			forever				= false;
 static bool			metric_only			= false;
 static bool			force_metric_only		= false;
+static bool			no_merge			= false;
 static struct timespec		ref_time;
 static struct cpu_map		*aggr_map;
 static aggr_get_id_t		aggr_get_id;
@@ -1178,12 +1179,34 @@ static void aggr_update_shadow(void)
 	}
 }
 
+static void collect_all_aliases(struct perf_evsel *counter,
+			    void (*cb)(struct perf_evsel *counter, void *data,
+				       bool first),
+			    void *data)
+{
+	struct perf_evsel *alias;
+
+	alias = list_prepare_entry(counter, &(evsel_list->entries), node);
+	list_for_each_entry_continue (alias, &evsel_list->entries, node) {
+		if (strcmp(perf_evsel__name(alias), perf_evsel__name(counter)) ||
+		    alias->scale != counter->scale ||
+		    alias->cgrp != counter->cgrp ||
+		    strcmp(alias->unit, counter->unit) ||
+		    nsec_counter(alias) != nsec_counter(counter))
+			break;
+		alias->merged_stat = true;
+		cb(alias, data, false);
+	}
+}
+
 static void collect_data(struct perf_evsel *counter,
 			    void (*cb)(struct perf_evsel *counter, void *data,
 				       bool first),
 			    void *data)
 {
 	cb(counter, data, true);
+	if (!no_merge)
+		collect_all_aliases(counter, cb, data);
 }
 
 struct aggr_data {
@@ -1208,6 +1231,16 @@ static void aggr_cb(struct perf_evsel *counter, void *data, bool first)
 		if (first)
 			ad->nr++;
 		counts = perf_counts(counter->counts, cpu, 0);
+		/*
+		 * When any result is bad, make them all to give
+		 * consistent output in interval mode.
+		 */
+		if (counts->ena == 0 || counts->run == 0 ||
+		    counter->counts->scaled == -1) {
+			ad->ena = 0;
+			ad->run = 0;
+			break;
+		}
 		ad->val += counts->val;
 		ad->ena += counts->ena;
 		ad->run += counts->run;
@@ -1240,6 +1273,8 @@ static void print_aggr(char *prefix)
 		ad.id = id = aggr_map->map[s];
 		first = true;
 		evlist__for_each_entry(evsel_list, counter) {
+			if (counter->merged_stat)
+				continue;
 			ad.val = ad.ena = ad.run = 0;
 			ad.nr = 0;
 			collect_data(counter, aggr_cb, &ad);
@@ -1315,6 +1350,8 @@ static void print_counter_aggr(struct perf_evsel *counter, char *prefix)
 	double uval;
 	struct caggr_data cd = { .avg = 0.0 };
 
+	if (counter->merged_stat)
+		return;
 	collect_data(counter, counter_aggr_cb, &cd);
 
 	if (prefix && !metric_only)
@@ -1347,6 +1384,9 @@ static void print_counter(struct perf_evsel *counter, char *prefix)
 	double uval;
 	int cpu;
 
+	if (counter->merged_stat)
+		return;
+
 	for (cpu = 0; cpu < perf_evsel__nr_cpus(counter); cpu++) {
 		struct aggr_data ad = { .cpu = cpu };
 
@@ -1691,6 +1731,7 @@ static const struct option stat_options[] = {
 		    "list of cpus to monitor in system-wide"),
 	OPT_SET_UINT('A', "no-aggr", &stat_config.aggr_mode,
 		    "disable CPU count aggregation", AGGR_NONE),
+	OPT_BOOLEAN(0, "no-merge", &no_merge, "Do not merge identical named events"),
 	OPT_STRING('x', "field-separator", &csv_sep, "separator",
 		   "print counts with custom separator"),
 	OPT_CALLBACK('G', "cgroup", &evsel_list, "name",
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 06ef6f29efa1..bd2e9b112d49 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -131,6 +131,7 @@ struct perf_evsel {
 	bool			cmdline_group_boundary;
 	struct list_head	config_terms;
 	int			bpf_fd;
+	bool			merged_stat;
 };
 
 union u64_swap {
-- 
2.9.3

[toc] | [prev] | [next] | [standalone]

#1592816 — Re: [PATCH 02/10] perf, tools, stat: Collapse identically named events

On Tue, Feb 28, 2017 at 10:49:16PM -0800, Andi Kleen wrote:

SNIP

>  
>  struct aggr_data {
> @@ -1208,6 +1231,16 @@ static void aggr_cb(struct perf_evsel *counter, void *data, bool first)
>  		if (first)
>  			ad->nr++;
>  		counts = perf_counts(counter->counts, cpu, 0);
> +		/*
> +		 * When any result is bad, make them all to give
> +		 * consistent output in interval mode.
> +		 */
> +		if (counts->ena == 0 || counts->run == 0 ||
> +		    counter->counts->scaled == -1) {
> +			ad->ena = 0;
> +			ad->run = 0;
> +			break;
> +		}

I think this needs to be in separate patch, showing
the interval output before and after the change..

thanks,
jirka

[toc] | [prev] | [next] | [standalone]

#1594069 — Re: [PATCH 02/10] perf, tools, stat: Collapse identically named events

On Tue, Feb 28, 2017 at 10:49:16PM -0800, Andi Kleen wrote:

SNIP

>  	}
>  }
>  
> +static void collect_all_aliases(struct perf_evsel *counter,
> +			    void (*cb)(struct perf_evsel *counter, void *data,
> +				       bool first),
> +			    void *data)
> +{
> +	struct perf_evsel *alias;
> +
> +	alias = list_prepare_entry(counter, &(evsel_list->entries), node);
> +	list_for_each_entry_continue (alias, &evsel_list->entries, node) {
> +		if (strcmp(perf_evsel__name(alias), perf_evsel__name(counter)) ||
> +		    alias->scale != counter->scale ||
> +		    alias->cgrp != counter->cgrp ||
> +		    strcmp(alias->unit, counter->unit) ||
> +		    nsec_counter(alias) != nsec_counter(counter))
> +			break;
> +		alias->merged_stat = true;
> +		cb(alias, data, false);
> +	}
> +}
> +
>  static void collect_data(struct perf_evsel *counter,
>  			    void (*cb)(struct perf_evsel *counter, void *data,
>  				       bool first),
>  			    void *data)
>  {

could we place the merged_stat condition in here and have
collect_data return value to process or not the current
counter.. looks like we could get rid of 2 merged_stat
conditions in print_counter_aggr and print_aggr

jirka

[toc] | [prev] | [next] | [standalone]

#1590262 — [PATCH 08/10] perf, tools, stat: Output JSON MetricExpr metric

From: Andi Kleen <ak@linux.intel.com>

Add generic infrastructure to perf stat to output ratios for "MetricExpr"
entries in the event lists. Many events are more useful as ratios
than in raw form, typically some count in relation to total ticks.

Transfer the MetricExpr information from the alias to the evsel.

We mark the events that need to be collected for MetricExpr, and also
link the events using them with a pointer. The code is careful
to always prefer the right event in the same group to minimize
multiplexing errors. At the moment only a single relation is supported.

Then add a rblist to the stat shadow code that remembers stats based
on the cpu and context.

Then finally update and retrieve and print these values similarly to the
existing hardcoded perf metrics. We use the simple expression parser
added earlier to evaluate the expression.

Normally we just output the result without further commentary,
but for --metric-only this would lead to empty columns. So for this
case use the original event as description.

So far there is no attempt to automatically add the MetricExpr event,
if it is missing, however we suggest it to the user.

$ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}'
     1.000228813        800,139,950      unc_p_clockticks
     1.000228813        789,833,783      unc_p_freq_max_os_cycles  #     98.7
     2.000654229        800,308,990      unc_p_clockticks
     2.000654229        396,214,238      unc_p_freq_max_os_cycles  #     49.5

$ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' --metric-only
     1.000206740     48.0
     2.000451543     48.1

v2: Change from DivideBy to MetricExpr
v3: Use expr__ prefix.  Support more than one other event.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
 tools/perf/builtin-stat.c      |   3 +
 tools/perf/util/evsel.c        |   3 +
 tools/perf/util/evsel.h        |   3 +
 tools/perf/util/parse-events.c |   1 +
 tools/perf/util/pmu.c          |   2 +
 tools/perf/util/pmu.h          |   1 +
 tools/perf/util/stat-shadow.c  | 188 +++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/stat.h         |   2 +
 8 files changed, 203 insertions(+)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 389c8e457bf0..a6cc5ad479b9 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1141,6 +1141,7 @@ static void printout(int id, int nr, struct perf_evsel *counter, double uval,
 	out.print_metric = pm;
 	out.new_line = nl;
 	out.ctx = &os;
+	out.force_header = false;
 
 	if (csv_output && !metric_only) {
 		print_noise(counter, noise);
@@ -1478,6 +1479,7 @@ static void print_metric_headers(const char *prefix, bool no_indent)
 		out.ctx = &os;
 		out.print_metric = print_metric_header;
 		out.new_line = new_line_metric;
+		out.force_header = true;
 		os.evsel = counter;
 		perf_stat__print_shadow_stats(counter, 0,
 					      0,
@@ -2460,6 +2462,7 @@ int cmd_stat(int argc, const char **argv, const char *prefix __maybe_unused)
 	argc = parse_options_subcommand(argc, argv, stat_options, stat_subcommands,
 					(const char **) stat_usage,
 					PARSE_OPT_STOP_AT_NON_OPTION);
+	perf_stat__collect_metric_expr(evsel_list);
 	perf_stat__init_shadow_stats();
 
 	if (csv_sep) {
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index ac59710b79e0..a25dc688ab74 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -236,6 +236,9 @@ void perf_evsel__init(struct perf_evsel *evsel,
 	evsel->sample_size = __perf_evsel__sample_size(attr->sample_type);
 	perf_evsel__calc_id_pos(evsel);
 	evsel->cmdline_group_boundary = false;
+	evsel->metric_expr   = NULL;
+	evsel->metric_events = NULL;
+	evsel->collect_stat  = false;
 }
 
 struct perf_evsel *perf_evsel__new_idx(struct perf_event_attr *attr, int idx)
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index bd2e9b112d49..8f1f61826fdf 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -132,6 +132,9 @@ struct perf_evsel {
 	struct list_head	config_terms;
 	int			bpf_fd;
 	bool			merged_stat;
+	const char *		metric_expr;
+	struct perf_evsel	**metric_events;
+	bool			collect_stat;
 };
 
 union u64_swap {
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index d693e2acbda7..fcb0b6854642 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -1254,6 +1254,7 @@ int parse_events_add_pmu(struct parse_events_evlist *data,
 		evsel->scale = info.scale;
 		evsel->per_pkg = info.per_pkg;
 		evsel->snapshot = info.snapshot;
+		evsel->metric_expr = info.metric_expr;
 	}
 
 	return evsel ? 0 : -ENOMEM;
diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
index 0583e1bdc762..4d9632390ab1 100644
--- a/tools/perf/util/pmu.c
+++ b/tools/perf/util/pmu.c
@@ -985,6 +985,7 @@ int perf_pmu__check_alias(struct perf_pmu *pmu, struct list_head *head_terms,
 	info->unit     = NULL;
 	info->scale    = 0.0;
 	info->snapshot = false;
+	info->metric_expr = NULL;
 
 	list_for_each_entry_safe(term, h, head_terms, list) {
 		alias = pmu_find_alias(pmu, term);
@@ -1000,6 +1001,7 @@ int perf_pmu__check_alias(struct perf_pmu *pmu, struct list_head *head_terms,
 
 		if (alias->per_pkg)
 			info->per_pkg = true;
+		info->metric_expr = alias->metric_expr;
 
 		list_del(&term->list);
 		free(term);
diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
index 3dccb15f29e9..27f078ccc594 100644
--- a/tools/perf/util/pmu.h
+++ b/tools/perf/util/pmu.h
@@ -31,6 +31,7 @@ struct perf_pmu {
 
 struct perf_pmu_info {
 	const char *unit;
+	const char *metric_expr;
 	double scale;
 	bool per_pkg;
 	bool snapshot;
diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
index 8a2bbd2a4d82..e8e76744fa32 100644
--- a/tools/perf/util/stat-shadow.c
+++ b/tools/perf/util/stat-shadow.c
@@ -3,6 +3,9 @@
 #include "stat.h"
 #include "color.h"
 #include "pmu.h"
+#include "rblist.h"
+#include "evlist.h"
+#include "expr.h"
 
 enum {
 	CTX_BIT_USER	= 1 << 0,
@@ -41,13 +44,73 @@ static struct stats runtime_topdown_slots_issued[NUM_CTX][MAX_NR_CPUS];
 static struct stats runtime_topdown_slots_retired[NUM_CTX][MAX_NR_CPUS];
 static struct stats runtime_topdown_fetch_bubbles[NUM_CTX][MAX_NR_CPUS];
 static struct stats runtime_topdown_recovery_bubbles[NUM_CTX][MAX_NR_CPUS];
+static struct rblist runtime_saved_values;
 static bool have_frontend_stalled;
 
 struct stats walltime_nsecs_stats;
 
+struct saved_value {
+	struct rb_node rb_node;
+	struct perf_evsel *evsel;
+	int cpu;
+	int ctx;
+	struct stats stats;
+};
+
+static int saved_value_cmp(struct rb_node *rb_node, const void *entry)
+{
+	struct saved_value *a = container_of(rb_node,
+					     struct saved_value,
+					     rb_node);
+	const struct saved_value *b = entry;
+
+	if (a->ctx != b->ctx)
+		return a->ctx - b->ctx;
+	if (a->cpu != b->cpu)
+		return a->cpu - b->cpu;
+	return a->evsel - b->evsel;
+}
+
+static struct rb_node *saved_value_new(struct rblist *rblist __maybe_unused,
+				     const void *entry)
+{
+	struct saved_value *nd = malloc(sizeof(struct saved_value));
+
+	if (!nd)
+		return NULL;
+	memcpy(nd, entry, sizeof(struct saved_value));
+	return &nd->rb_node;
+}
+
+static struct saved_value *saved_value_lookup(struct perf_evsel *evsel,
+					      int cpu, int ctx,
+					      bool create)
+{
+	struct rb_node *nd;
+	struct saved_value dm = {
+		.cpu = cpu,
+		.ctx = ctx,
+		.evsel = evsel,
+	};
+	nd = rblist__find(&runtime_saved_values, &dm);
+	if (nd)
+		return container_of(nd, struct saved_value, rb_node);
+	if (create) {
+		rblist__add_node(&runtime_saved_values, &dm);
+		nd = rblist__find(&runtime_saved_values, &dm);
+		if (nd)
+			return container_of(nd, struct saved_value, rb_node);
+	}
+	return NULL;
+}
+
 void perf_stat__init_shadow_stats(void)
 {
 	have_frontend_stalled = pmu_have_event("cpu", "stalled-cycles-frontend");
+	rblist__init(&runtime_saved_values);
+	runtime_saved_values.node_cmp = saved_value_cmp;
+	runtime_saved_values.node_new = saved_value_new;
+	/* No delete for now */
 }
 
 static int evsel_context(struct perf_evsel *evsel)
@@ -70,6 +133,8 @@ static int evsel_context(struct perf_evsel *evsel)
 
 void perf_stat__reset_shadow_stats(void)
 {
+	struct rb_node *pos, *next;
+
 	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));
@@ -92,6 +157,15 @@ void perf_stat__reset_shadow_stats(void)
 	memset(runtime_topdown_slots_issued, 0, sizeof(runtime_topdown_slots_issued));
 	memset(runtime_topdown_fetch_bubbles, 0, sizeof(runtime_topdown_fetch_bubbles));
 	memset(runtime_topdown_recovery_bubbles, 0, sizeof(runtime_topdown_recovery_bubbles));
+
+	next = rb_first(&runtime_saved_values.entries);
+	while (next) {
+		pos = next;
+		next = rb_next(pos);
+		memset(&container_of(pos, struct saved_value, rb_node)->stats,
+		       0,
+		       sizeof(struct stats));
+	}
 }
 
 /*
@@ -143,6 +217,12 @@ void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
 		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]);
+
+	if (counter->collect_stat) {
+		struct saved_value *v = saved_value_lookup(counter, cpu, ctx,
+							   true);
+		update_stats(&v->stats, count[0]);
+	}
 }
 
 /* used for get_ratio_color() */
@@ -172,6 +252,88 @@ static const char *get_ratio_color(enum grc_type type, double ratio)
 	return color;
 }
 
+static struct perf_evsel *perf_stat__find_event(struct perf_evlist *evsel_list,
+						const char *name)
+{
+	struct perf_evsel *c2;
+
+	evlist__for_each_entry (evsel_list, c2) {
+		if (!strcasecmp(c2->name, name))
+			return c2;
+	}
+	return NULL;
+}
+
+/* Mark MetricExpr target events and link events using them to them. */
+void perf_stat__collect_metric_expr(struct perf_evlist *evsel_list)
+{
+	struct perf_evsel *counter, *leader, **metric_events, *oc;
+	bool found;
+	const char **metric_names;
+	int i;
+	int num_metric_names;
+
+	evlist__for_each_entry(evsel_list, counter) {
+		bool invalid = false;
+
+		leader = counter->leader;
+		if (!counter->metric_expr)
+			continue;
+		metric_events = counter->metric_events;
+		if (!metric_events) {
+			if (expr__find_other(counter->metric_expr, counter->name,
+						&metric_names, &num_metric_names) < 0)
+				continue;
+
+			metric_events = calloc(sizeof(struct perf_evsel *),
+					       num_metric_names + 1);
+			if (!metric_events)
+				return;
+			counter->metric_events = metric_events;
+		}
+
+		for (i = 0; i < num_metric_names; i++) {
+			found = false;
+			if (leader) {
+				/* Search in group */
+				for_each_group_member (oc, leader) {
+					if (!strcasecmp(oc->name, metric_names[i])) {
+						found = true;
+						break;
+					}
+				}
+			}
+			if (!found) {
+				/* Search ignoring groups */
+				oc = perf_stat__find_event(evsel_list, metric_names[i]);
+			}
+			if (!oc) {
+				/*
+				 * Adding events automatically would be difficult, because
+				 * it would risk creating groups that are not schedulable.
+				 * perf stat doesn't understand all the scheduling constraints
+				 * of events. So we ask the user instead to add the missing
+				 * events.
+				 */
+				fprintf(stderr, "Add %s event to groups to get metric expression for %s\n",
+						metric_names[i],
+						counter->name);
+				invalid = true;
+				continue;
+			}
+			metric_events[i] = oc;
+			oc->collect_stat = true;
+		}
+		metric_events[i] = NULL;
+		free(metric_names);
+		if (invalid) {
+			free(metric_events);
+			counter->metric_events = NULL;
+			counter->metric_expr = NULL;
+		}
+	}
+}
+
 static void print_stalled_cycles_frontend(int cpu,
 					  struct perf_evsel *evsel, double avg,
 					  struct perf_stat_output_ctx *out)
@@ -614,6 +776,32 @@ void perf_stat__print_shadow_stats(struct perf_evsel *evsel,
 					be_bound * 100.);
 		else
 			print_metric(ctxp, NULL, NULL, name, 0);
+	} else if (evsel->metric_expr) {
+		struct parse_ctx pctx;
+		int i;
+
+		expr__ctx_init(&pctx);
+		expr__add_id(&pctx, evsel->name, avg);
+		for (i = 0; evsel->metric_events[i]; i++) {
+			struct saved_value *v;
+
+			v = saved_value_lookup(evsel->metric_events[i], cpu, ctx, false);
+			if (!v)
+				break;
+			expr__add_id(&pctx, evsel->metric_events[i]->name,
+					     avg_stats(&v->stats));
+		}
+		if (!evsel->metric_events[i]) {
+			const char *p = evsel->metric_expr;
+
+			if (expr__parse(&ratio, &pctx, &p) == 0)
+				print_metric(ctxp, NULL, "%8.1f",
+					out->force_header ? evsel->name : "",
+					ratio);
+			else
+				print_metric(ctxp, NULL, NULL, "", 0);
+		} else
+			print_metric(ctxp, NULL, NULL, "", 0);
 	} else if (runtime_nsecs_stats[cpu].n != 0) {
 		char unit = 'M';
 		char unit_buf[10];
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index c29bb94c48a4..0a65ae23f495 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -85,11 +85,13 @@ struct perf_stat_output_ctx {
 	void *ctx;
 	print_metric_t print_metric;
 	new_line_t new_line;
+	bool force_header;
 };
 
 void perf_stat__print_shadow_stats(struct perf_evsel *evsel,
 				   double avg, int cpu,
 				   struct perf_stat_output_ctx *out);
+void perf_stat__collect_metric_expr(struct perf_evlist *);
 
 int perf_evlist__alloc_stats(struct perf_evlist *evlist, bool alloc_raw);
 void perf_evlist__free_stats(struct perf_evlist *evlist);
-- 
2.9.3

[toc] | [prev] | [next] | [standalone]

#1594064 — Re: [PATCH 08/10] perf, tools, stat: Output JSON MetricExpr metric

On Tue, Feb 28, 2017 at 10:49:22PM -0800, Andi Kleen wrote:
> From: Andi Kleen <ak@linux.intel.com>
> 
> Add generic infrastructure to perf stat to output ratios for "MetricExpr"
> entries in the event lists. Many events are more useful as ratios
> than in raw form, typically some count in relation to total ticks.
> 
> Transfer the MetricExpr information from the alias to the evsel.
> 
> We mark the events that need to be collected for MetricExpr, and also
> link the events using them with a pointer. The code is careful
> to always prefer the right event in the same group to minimize
> multiplexing errors. At the moment only a single relation is supported.
> 
> Then add a rblist to the stat shadow code that remembers stats based
> on the cpu and context.
> 
> Then finally update and retrieve and print these values similarly to the
> existing hardcoded perf metrics. We use the simple expression parser
> added earlier to evaluate the expression.
> 
> Normally we just output the result without further commentary,
> but for --metric-only this would lead to empty columns. So for this
> case use the original event as description.
> 
> So far there is no attempt to automatically add the MetricExpr event,
> if it is missing, however we suggest it to the user.
> 
> $ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}'
>      1.000228813        800,139,950      unc_p_clockticks
>      1.000228813        789,833,783      unc_p_freq_max_os_cycles  #     98.7
>      2.000654229        800,308,990      unc_p_clockticks
>      2.000654229        396,214,238      unc_p_freq_max_os_cycles  #     49.5
> 
> $ perf stat -a -I 1000 -e '{unc_p_clockticks,unc_p_freq_max_os_cycles}' --metric-only
>      1.000206740     48.0
>      2.000451543     48.1

The comment above suggest I'll see 'original event as description' in this example??

However, could you just make those metric names working in here?
Displaying original event for metric value is misleading.

jirka

[toc] | [prev] | [next] | [standalone]

#1594062

On Tue, Feb 28, 2017 at 10:49:14PM -0800, Andi Kleen wrote:
> This patch kit further improves support for Intel uncore events in
> the Linux perf user tool. The basic support has been already
> merged earlier, but this makes it nicer to use.
> 
> - Collapse counts from duplicated boxes to make the output
> easier to read.
> - Support specifying events for multiple duplicated boxes
> in an abbreviated format to shorten event specifiers
> - Add support for computing Metrics defined in the event lists,
> so that the event lists can extend the metrics in perf stat.
> This allows to represent many events in an easier to understand
> format.
> 
> Available from
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/builtin-json-28
> 
> v1: Initial post after being split off to own patchkit
> Adding MetricName support and support for more than two events
> in expressions.
> 
> v2: Address review comments. Move new hunk from refactor
> patchkit to patch adding new features. Improve changelogs
> slightly.
> 

with my comments addressed, I'll be ok to ack it in next version

thanks,
jirka

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web