Commit 9d6a1df9 authored by Ian Rogers's avatar Ian Rogers Committed by Arnaldo Carvalho de Melo
Browse files

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: default avatarKan Liang <kan.liang@linux.intel.com>
Signed-off-by: default avatarIan 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: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
parent 15c57a80

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

+1 −4
Original line number Diff line number Diff line
@@ -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

+1 −4
Original line number Diff line number Diff line
@@ -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

+3 −5
Original line number Diff line number Diff line
@@ -300,12 +300,10 @@ 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) {

		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

+29 −21
Original line number Diff line number Diff line
@@ -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++) {

	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();

			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 PMU scanning took: %.3f usec (+- %.3f usec)\n",

		time_average, time_stddev);



		pr_info("  Average%s PMU scanning took: %.3f usec (+- %.3f usec)\n",

			core_only ? " core" : "", time_average, time_stddev);

	}

	delete_result();

	return 0;

}

tools/perf/tests/pmu-events.c

+1 −4
Original line number Diff line number Diff line
@@ -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;