perf pmus: Allow just core PMU scanning

Scanning all PMUs is expensive as all PMUs sysfs entries are loaded,
benchmarking shows more than 4x the cost:

```
$ perf bench internals pmu-scan -i 1000
Computing performance of sysfs PMU event scan for 1000 times
  Average core PMU scanning took: 989.231 usec (+- 1.535 usec)
  Average PMU scanning took: 4309.425 usec (+- 74.322 usec)
```

Add new perf_pmus__scan_core routine that scans just core
PMUs. Replace perf_pmus__scan calls with perf_pmus__scan_core when
non-core PMUs are being ignored.

Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ali Saidi <alisaidi@amazon.com>
Cc: Athira Rajeev <atrajeev@linux.vnet.ibm.com>
Cc: Dmitrii Dolgov <9erthalion6@gmail.com>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jing Zhang <renyu.zj@linux.alibaba.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.g.garry@oracle.com>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Kang Minchul <tegongkang@gmail.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Madhavan Srinivasan <maddy@linux.ibm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Ming Wang <wangming01@loongson.cn>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Sandipan Das <sandipan.das@amd.com>
Cc: Sean Christopherson <seanjc@google.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: https://lore.kernel.org/r/20230527072210.2900565-30-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Author: captain5050

tools/perf/arch/arm64/util/pmu.c

@@ -11,10 +11,7 @@ static struct perf_pmu *pmu__find_core_pmu(void)
 {
 	struct perf_pmu *pmu = NULL;
 
-	while ((pmu = perf_pmus__scan(pmu))) {
-		if (!is_pmu_core(pmu->name))
-			continue;
-
+	while ((pmu = perf_pmus__scan_core(pmu))) {
 		/*
 		 * The cpumap should cover all CPUs. Otherwise, some CPUs may
 		 * not support some events or have different event IDs.

tools/perf/arch/x86/util/evlist.c

@@ -33,13 +33,10 @@ static int ___evlist__add_default_attrs(struct evlist *evlist,
 			continue;
 		}
 
-		while ((pmu = perf_pmus__scan(pmu)) != NULL) {
+		while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
 			struct perf_cpu_map *cpus;
 			struct evsel *evsel;
 
-			if (!pmu->is_core)
-				continue;
-
 			evsel = evsel__new(attrs + i);
 			if (evsel == NULL)
 				goto out_delete_partial_list;

tools/perf/arch/x86/util/perf_regs.c

@@ -300,11 +300,9 @@ uint64_t arch__intr_reg_mask(void)
 		 * The same register set is supported among different hybrid PMUs.
 		 * Only check the first available one.
 		 */
-		while ((pmu = perf_pmus__scan(pmu)) != NULL) {
-			if (pmu->is_core) {
-				type = pmu->type;
-				break;
-			}
+		while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
+			type = pmu->type;
+			break;
 		}
 		attr.config |= type << PERF_PMU_TYPE_SHIFT;
 	}

tools/perf/bench/pmu-scan.c

@@ -22,6 +22,7 @@ struct pmu_scan_result {
 	int nr_aliases;
 	int nr_formats;
 	int nr_caps;
+	bool is_core;
 };
 
 static const struct option options[] = {
@@ -53,6 +54,7 @@ static int save_result(void)
 		r = results + nr_pmus;
 
 		r->name = strdup(pmu->name);
+		r->is_core = pmu->is_core;
 		r->nr_caps = pmu->nr_caps;
 
 		r->nr_aliases = 0;
@@ -72,7 +74,7 @@ static int save_result(void)
 	return 0;
 }
 
-static int check_result(void)
+static int check_result(bool core_only)
 {
 	struct pmu_scan_result *r;
 	struct perf_pmu *pmu;
@@ -81,6 +83,9 @@ static int check_result(void)
 
 	for (int i = 0; i < nr_pmus; i++) {
 		r = &results[i];
+		if (core_only && !r->is_core)
+			continue;
+
 		pmu = perf_pmus__find(r->name);
 		if (pmu == NULL) {
 			pr_err("Cannot find PMU %s\n", r->name);
@@ -130,7 +135,6 @@ static int run_pmu_scan(void)
 	struct timeval start, end, diff;
 	double time_average, time_stddev;
 	u64 runtime_us;
-	unsigned int i;
 	int ret;
 
 	init_stats(&stats);
@@ -142,26 +146,30 @@ static int run_pmu_scan(void)
 		return -1;
 	}
 
-	for (i = 0; i < iterations; i++) {
-		gettimeofday(&start, NULL);
-		perf_pmus__scan(NULL);
-		gettimeofday(&end, NULL);
-
-		timersub(&end, &start, &diff);
-		runtime_us = diff.tv_sec * USEC_PER_SEC + diff.tv_usec;
-		update_stats(&stats, runtime_us);
-
-		ret = check_result();
-		perf_pmus__destroy();
-		if (ret < 0)
-			break;
+	for (int j = 0; j < 2; j++) {
+		bool core_only = (j == 0);
+
+		for (unsigned int i = 0; i < iterations; i++) {
+			gettimeofday(&start, NULL);
+			if (core_only)
+				perf_pmus__scan_core(NULL);
+			else
+				perf_pmus__scan(NULL);
+			gettimeofday(&end, NULL);
+			timersub(&end, &start, &diff);
+			runtime_us = diff.tv_sec * USEC_PER_SEC + diff.tv_usec;
+			update_stats(&stats, runtime_us);
+
+			ret = check_result(core_only);
+			perf_pmus__destroy();
+			if (ret < 0)
+				break;
+		}
+		time_average = avg_stats(&stats);
+		time_stddev = stddev_stats(&stats);
+		pr_info("  Average%s PMU scanning took: %.3f usec (+- %.3f usec)\n",
+			core_only ? " core" : "", time_average, time_stddev);
 	}
-
-	time_average = avg_stats(&stats);
-	time_stddev = stddev_stats(&stats);
-	pr_info("  Average PMU scanning took: %.3f usec (+- %.3f usec)\n",
-		time_average, time_stddev);
-
 	delete_result();
 	return 0;
 }

tools/perf/tests/pmu-events.c

@@ -709,12 +709,9 @@ static int test__aliases(struct test_suite *test __maybe_unused,
 	struct perf_pmu *pmu = NULL;
 	unsigned long i;
 
-	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
+	while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
 		int count = 0;
 
-		if (!is_pmu_core(pmu->name))
-			continue;
-
 		if (list_empty(&pmu->format)) {
 			pr_debug2("skipping testing core PMU %s\n", pmu->name);
 			continue;

tools/perf/util/cputopo.c

@@ -477,10 +477,9 @@ struct hybrid_topology *hybrid_topology__new(void)
 	if (!perf_pmus__has_hybrid())
 		return NULL;
 
-	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
-		if (pmu->is_core)
-			nr++;
-	}
+	while ((pmu = perf_pmus__scan_core(pmu)) != NULL)
+		nr++;
+
 	if (nr == 0)
 		return NULL;
 
@@ -489,10 +488,7 @@ struct hybrid_topology *hybrid_topology__new(void)
 		return NULL;
 
 	tp->nr = nr;
-	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
-		if (!pmu->is_core)
-			continue;
-
+	while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
 		if (load_hybrid_node(&tp->nodes[i], pmu)) {
 			hybrid_topology__delete(tp);
 			return NULL;

tools/perf/util/header.c

@@ -1607,10 +1607,7 @@ static int write_pmu_caps(struct feat_fd *ff,
 	 */
 	if (perf_pmus__has_hybrid()) {
 		pmu = NULL;
-		while ((pmu = perf_pmus__scan(pmu))) {
-			if (!pmu->is_core)
-				continue;
-
+		while ((pmu = perf_pmus__scan_core(pmu))) {
 			ret = __write_pmu_caps(ff, pmu, true);
 			if (ret < 0)
 				return ret;

tools/perf/util/mem-events.c

@@ -136,10 +136,7 @@ int perf_mem_events__init(void)
 		} else {
 			struct perf_pmu *pmu = NULL;
 
-			while ((pmu = perf_pmus__scan(pmu)) != NULL) {
-				if (!pmu->is_core)
-					continue;
-
+			while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
 				scnprintf(sysfs_name, sizeof(sysfs_name),
 					  e->sysfs_name, pmu->name);
 				e->supported |= perf_mem_event__supported(mnt, sysfs_name);
@@ -176,10 +173,7 @@ static void perf_mem_events__print_unsupport_hybrid(struct perf_mem_event *e,
 	char sysfs_name[100];
 	struct perf_pmu *pmu = NULL;
 
-	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
-		if (!pmu->is_core)
-			continue;
-
+	while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
 		scnprintf(sysfs_name, sizeof(sysfs_name), e->sysfs_name,
 			  pmu->name);
 		if (!perf_mem_event__supported(mnt, sysfs_name)) {
@@ -217,9 +211,7 @@ int perf_mem_events__record_args(const char **rec_argv, int *argv_nr,
 				return -1;
 			}
 
-			while ((pmu = perf_pmus__scan(pmu)) != NULL) {
-				if (!pmu->is_core)
-					continue;
+			while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
 				rec_argv[i++] = "-e";
 				s = perf_mem_events__name(j, pmu->name);
 				if (s) {

tools/perf/util/parse-events.c

@@ -453,15 +453,12 @@ int parse_events_add_cache(struct list_head *list, int *idx, const char *name,
 	const char *config_name = get_config_name(head_config);
 	const char *metric_id = get_config_metric_id(head_config);
 
-	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
+	/* Legacy cache events are only supported by core PMUs. */
+	while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
 		LIST_HEAD(config_terms);
 		struct perf_event_attr attr;
 		int ret;
 
-		/* Skip unsupported PMUs. */
-		if (!perf_pmu__supports_legacy_cache(pmu))
-			continue;
-
 		if (parse_events__filter_pmu(parse_state, pmu))
 			continue;
 
@@ -1481,12 +1478,10 @@ int parse_events_add_numeric(struct parse_events_state *parse_state,
 		return __parse_events_add_numeric(parse_state, list, /*pmu=*/NULL,
 						  type, config, head_config);
 
-	while ((pmu = perf_pmus__scan(pmu)) != NULL) {
+	/* Wildcards on numeric values are only supported by core PMUs. */
+	while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
 		int ret;
 
-		if (!perf_pmu__supports_wildcard_numeric(pmu))
-			continue;
-
 		if (parse_events__filter_pmu(parse_state, pmu))
 			continue;
 

tools/perf/util/pmu.c

@@ -1427,21 +1427,11 @@ bool perf_pmu__supports_legacy_cache(const struct perf_pmu *pmu)
 	return pmu->is_core;
 }
 
-bool perf_pmu__supports_wildcard_numeric(const struct perf_pmu *pmu)
-{
-	return pmu->is_core;
-}
-
 bool perf_pmu__auto_merge_stats(const struct perf_pmu *pmu)
 {
 	return !is_pmu_hybrid(pmu->name);
 }
 
-bool perf_pmu__is_mem_pmu(const struct perf_pmu *pmu)
-{
-	return pmu->is_core;
-}
-
 bool perf_pmu__have_event(const struct perf_pmu *pmu, const char *name)
 {
 	struct perf_pmu_alias *alias;