PMU refactoring and improvements
Ian Rogers
reviewerv4 cover PMU refactoring and improvements
Separate the code in pmu.[ch] into the set/list of PMUs and the code for a particular PMU. Move the set/list of PMUs code into pmus.[ch]. Clean up hybrid code and remove hybrid PMU list, it is sufficient to scan PMUs looking for core ones. Add core PMU list and perf_pmus__scan_core that just reads core PMUs. Switch code that skips non-core PMUs during a perf_pmus__scan, to use the perf_pmus__scan_core variant. Don't scan sysfs for PMUs if all such PMUs have been previously scanned/loaded. Scanning just core PMUs, for the cases it is applicable, can improve the sysfs reading time by more than 4 fold on my laptop, as servers generally have many more uncore PMUs the improvement there should be larger:
``` $ 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) ```
The patch "perf pmu: Separate pmu and pmus" moves and renames a lot of functions, and is consequently large. The changes are trivial, but kept together to keep the overall number of patches more reasonable.
v4. On patch 16 (perf pmu: Remove perf_pmu__hybrid_mounted) remove the
handling of no cpus for a hybrid core PMU following discussion with Kan:
https://lore.kernel.org/lkml/[email protected]/
On patch 9 (perf evlist: Propagate user CPU maps intersecting core PMU maps) fix the comment on struct perf_evsel's system_wide variable from conversation with Namhyung:
https://lore.kernel.org/lkml/[email protected]/
v3. Address fixing hybrid user specified CPU maps by doing it in
propagate maps. Remove nearly all references to cpu_core/cpu_atom in particular by removing is_pmu_hybrid - hybrid is now >1 core PMU. Addresses comments by Kan and Namhyung.
v2. Address Kan's review comments wrt "cycles" -> "cycles:P" and
"uncore_pmus" -> "other_pmus".
Ian Rogers (35):
perf cpumap: Add intersect function
perf tests: Organize cpu_map tests into a single suite perf cpumap: Add equal function
libperf cpumap: Add "any CPU"/dummy test function perf pmu: Detect ARM and hybrid PMUs with sysfs
perf pmu: Add is_core to pmu
perf evsel: Add is_pmu_core inorder to interpret own_cpus perf pmu: Add CPU map for "cpu" PMUs
perf evlist: Propagate user CPU maps intersecting core PMU maps perf evlist: Allow has_user_cpus to be set on hybrid perf target: Remove unused hybrid value
perf tools: Warn if no user requested CPUs match PMU's CPUs perf evlist: Remove evlist__warn_hybrid_group
perf evlist: Remove __evlist__add_default
perf evlist: Reduce scope of evlist__has_hybrid
perf pmu: Remove perf_pmu__hybrid_mounted
perf pmu: Rewrite perf_pmu__has_hybrid to avoid list perf x86: Iterate hybrid PMUs as core PMUs
perf topology: Avoid hybrid list for hybrid topology perf evsel: Compute is_hybrid from PMU being core perf header: Avoid hybrid PMU list in write_pmu_caps perf metrics: Remove perf_pmu__is_hybrid use
perf stat: Avoid hybrid PMU list
perf mem: Avoid hybrid PMU list
perf pmu: Remove perf_pmu__hybrid_pmus list
perf pmus: Prefer perf_pmu__scan over perf_pmus__for_each_pmu perf x86 mem: minor refactor to is_mem_loads_aux_event perf pmu: Separate pmu and pmus
perf pmus: Split pmus list into core and other
perf pmus: Allow just core PMU scanning
perf pmus: Avoid repeated sysfs scanning
perf pmus: Ensure all PMUs are read for find_by_type perf pmus: Add function to return count of core PMUs perf pmus: Remove perf_pmus__has_hybrid
perf pmu: Remove is_pmu_hybrid
tools/lib/perf/cpumap.c | 61 +++ tools/lib/perf/evlist.c | 25 +- tools/lib/perf/include/internal/evsel.h | 15 +- tools/lib/perf/include/perf/cpumap.h | 14 + tools/perf/arch/arm/util/auxtrace.c | 7 +- tools/perf/arch/arm/util/cs-etm.c | 4 +- tools/perf/arch/arm64/util/pmu.c | 6 +- tools/perf/arch/x86/tests/hybrid.c | 7 +- tools/perf/arch/x86/util/auxtrace.c | 5 +- tools/perf/arch/x86/util/evlist.c | 25 +- tools/perf/arch/x86/util/evsel.c | 27 +- tools/perf/arch/x86/util/intel-bts.c | 4 +- tools/perf/arch/x86/util/intel-pt.c | 4 +- tools/perf/arch/x86/util/mem-events.c | 17 +- tools/perf/arch/x86/util/perf_regs.c | 15 +- tools/perf/arch/x86/util/topdown.c | 5 +- tools/perf/bench/pmu-scan.c | 60 +-- tools/perf/builtin-c2c.c | 9 +- tools/perf/builtin-list.c | 4 +- tools/perf/builtin-mem.c | 9 +- tools/perf/builtin-record.c | 29 +- tools/perf/builtin-stat.c | 14 +- tools/perf/builtin-top.c | 10 +- tools/perf/tests/attr.c | 11 +- tools/perf/tests/builtin-test.c | 4 +- tools/perf/tests/cpumap.c | 92 ++++- tools/perf/tests/event_groups.c | 7 +- tools/perf/tests/parse-events.c | 15 +- tools/perf/tests/parse-metric.c | 9 +- tools/perf/tests/pmu-events.c | 6 +- tools/perf/tests/switch-tracking.c | 14 +- tools/perf/tests/tests.h | 4 +- tools/perf/tests/topology.c | 16 +- tools/perf/util/Build | 2 - tools/perf/util/cpumap.c | 4 +- tools/perf/util/cpumap.h | 4 +- tools/perf/util/cputopo.c | 12 +- tools/perf/util/env.c | 5 +- tools/perf/util/evlist-hybrid.c | 162 -------- tools/perf/util/evlist-hybrid.h | 15 - tools/perf/util/evlist.c | 64 +++- tools/perf/util/evlist.h | 9 +- tools/perf/util/evsel.c | 60 +-- tools/perf/util/evsel.h | 3 - tools/perf/util/header.c | 27 +- tools/perf/util/mem-events.c | 25 +- tools/perf/util/metricgroup.c | 9 +- tools/perf/util/parse-events.c | 25 +- tools/perf/util/parse-events.y | 3 +- tools/perf/util/pfm.c | 6 +- tools/perf/util/pmu-hybrid.c | 52 --- tools/perf/util/pmu-hybrid.h | 32 -- tools/perf/util/pmu.c | 470 +----------------------- tools/perf/util/pmu.h | 25 +- tools/perf/util/pmus.c | 465 ++++++++++++++++++++++- tools/perf/util/pmus.h | 15 +- tools/perf/util/print-events.c | 15 +- tools/perf/util/python-ext-sources | 1 - tools/perf/util/stat-display.c | 19 +- tools/perf/util/target.h | 1 - 60 files changed, 992 insertions(+), 1092 deletions(-) delete mode 100644 tools/perf/util/evlist-hybrid.c delete mode 100644 tools/perf/util/evlist-hybrid.h delete mode 100644 tools/perf/util/pmu-hybrid.c delete mode 100644 tools/perf/util/pmu-hybrid.h
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 1/35 perf cpumap: Add intersect function
The merge function gives the union of two cpu maps. Add an intersect function which is necessary, for example, when intersecting a PMUs supported CPUs with user requested.
--- tools/lib/perf/cpumap.c | 35 ++++++++++++++++++++++++ tools/lib/perf/include/perf/cpumap.h | 2 ++ tools/perf/tests/builtin-test.c | 1 + tools/perf/tests/cpumap.c | 41 ++++++++++++++++++++++++++++ tools/perf/tests/tests.h | 1 + 5 files changed, 80 insertions(+)
tools/lib/perf/cpumap.c +35 -0
@@ -402,3 +402,38 @@ struct perf_cpu_map *perf_cpu_map__merge(struct perf_cpu_map *orig,perf_cpu_map__put (orig);return merged; } + +struct perf_cpu_map *perf_cpu_map__intersect (struct perf_cpu_map *orig, +struct perf_cpu_map *other) +{ +struct perf_cpu *tmp_cpus; +int tmp_len; +int i, j, k; +struct perf_cpu_map *merged =NULL ; + +if (perf_cpu_map__is_subset (other, orig)) +return perf_cpu_map__get (orig); +if (perf_cpu_map__is_subset (orig, other)) +return perf_cpu_map__get (other); + + tmp_len =max (orig->nr, other->nr); + tmp_cpus =malloc (tmp_len *sizeof (struct perf_cpu )); +if (!tmp_cpus) +return NULL ; + + i = j = k =0 ; +while (i < orig->nr && j < other->nr) { +if (orig->map[i].cpu < other->map[j].cpu) + i++; +else if (orig->map[i].cpu > other->map[j].cpu) + j++; +else { + j++; + tmp_cpus[k++] = orig->map[i++]; + } + } +if (k) + merged =cpu_map__trim_new (k, tmp_cpus); +free (tmp_cpus); +return merged; +}
tools/lib/perf/include/perf/cpumap.h +2 -0
@@ -20,6 +20,8 @@ LIBPERF_API struct perf_cpu_map *perf_cpu_map__read(FILE *file); LIBPERF_API struct perf_cpu_map *perf_cpu_map__get(struct perf_cpu_map *map); LIBPERF_API struct perf_cpu_map *perf_cpu_map__merge(struct perf_cpu_map *orig,struct perf_cpu_map *other); +LIBPERF_API struct perf_cpu_map *perf_cpu_map__intersect(struct perf_cpu_map *orig, +struct perf_cpu_map *other); LIBPERF_API void perf_cpu_map__put(struct perf_cpu_map *map); LIBPERF_API struct perf_cpu perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx); LIBPERF_API int perf_cpu_map__nr(const struct perf_cpu_map *cpus);
tools/perf/tests/builtin-test.c +1 -0
@@ -97,6 +97,7 @@ static struct test_suite *generic_tests[] = { &suite__backward_ring_buffer, &suite__cpu_map_print, &suite__cpu_map_merge, + &suite__cpu_map_intersect, &suite__sdt_event, &suite__is_printable_array, &suite__bitmap_print,
tools/perf/tests/cpumap.c +41 -0
47 lines not on this page · raw message
tools/perf/tests/tests.h +1 -0
@@ -127,6 +127,7 @@ DECLARE_SUITE(event_times);DECLARE_SUITE (backward_ring_buffer);DECLARE_SUITE (cpu_map_print);DECLARE_SUITE (cpu_map_merge); +DECLARE_SUITE (cpu_map_intersect);DECLARE_SUITE (sdt_event);DECLARE_SUITE (is_printable_array);DECLARE_SUITE (bitmap_print);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 2/35 perf tests: Organize cpu_map tests into a single suite
In reply to Ian RogersGo from 4 suites to a single suite with 4 test cases.
--- tools/perf/tests/builtin-test.c | 5 +---- tools/perf/tests/cpumap.c | 16 ++++++++++++---- tools/perf/tests/tests.h | 5 +---- 3 files changed, 14 insertions(+), 12 deletions(-)
tools/perf/tests/builtin-test.c +1 -4
@@ -88,16 +88,13 @@ static struct test_suite *generic_tests[] = { &suite__bpf, &suite__thread_map_synthesize, &suite__thread_map_remove, - &suite__cpu_map_synthesize, + &suite__cpu_map, &suite__synthesize_stat_config, &suite__synthesize_stat, &suite__synthesize_stat_round, &suite__event_update, &suite__event_times, &suite__backward_ring_buffer, - &suite__cpu_map_print, - &suite__cpu_map_merge, - &suite__cpu_map_intersect, &suite__sdt_event, &suite__is_printable_array, &suite__bitmap_print,
tools/perf/tests/cpumap.c +12 -4
@@ -211,7 +211,15 @@ static int test__cpu_map_intersect(struct test_suite *test __maybe_unused,return ret; } -DEFINE_SUITE ("Synthesize cpu map" , cpu_map_synthesize); -DEFINE_SUITE ("Print cpu map" , cpu_map_print); -DEFINE_SUITE ("Merge cpu map" , cpu_map_merge); -DEFINE_SUITE ("Intersect cpu map" , cpu_map_intersect); +static struct test_case tests__cpu_map[] = { +TEST_CASE ("Synthesize cpu map" , cpu_map_synthesize), +TEST_CASE ("Print cpu map" , cpu_map_print), +TEST_CASE ("Merge cpu map" , cpu_map_merge), +TEST_CASE ("Intersect cpu map" , cpu_map_intersect), + { .name =NULL , } +}; + +struct test_suite suite__cpu_map = { + .desc ="CPU map" , + .test_cases = tests__cpu_map, +};
tools/perf/tests/tests.h +1 -4
@@ -118,16 +118,13 @@ DECLARE_SUITE(bpf);DECLARE_SUITE (session_topology);DECLARE_SUITE (thread_map_synthesize);DECLARE_SUITE (thread_map_remove); -DECLARE_SUITE (cpu_map_synthesize); +DECLARE_SUITE (cpu_map);DECLARE_SUITE (synthesize_stat_config);DECLARE_SUITE (synthesize_stat);DECLARE_SUITE (synthesize_stat_round);DECLARE_SUITE (event_update);DECLARE_SUITE (event_times);DECLARE_SUITE (backward_ring_buffer); -DECLARE_SUITE (cpu_map_print); -DECLARE_SUITE (cpu_map_merge); -DECLARE_SUITE (cpu_map_intersect);DECLARE_SUITE (sdt_event);DECLARE_SUITE (is_printable_array);DECLARE_SUITE (bitmap_print);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 3/35 perf cpumap: Add equal function
In reply to Ian RogersEquality is a useful property to compare after merging and intersecting maps.
--- tools/lib/perf/cpumap.c | 21 ++++++++++++++++ tools/lib/perf/include/perf/cpumap.h | 2 ++ tools/perf/tests/cpumap.c | 37 ++++++++++++++++++++++++++++ 3 files changed, 60 insertions(+)
tools/lib/perf/cpumap.c +21 -0
@@ -321,6 +321,27 @@ bool perf_cpu_map__has(const struct perf_cpu_map *cpus, struct perf_cpu cpu)return perf_cpu_map__idx (cpus, cpu) !=-1 ; } +bool perf_cpu_map__equal (const struct perf_cpu_map *lhs,const struct perf_cpu_map *rhs) +{ +int nr; + +if (lhs == rhs) +return true; + +if (!lhs || !rhs) +return false; + + nr =perf_cpu_map__nr (lhs); +if (nr !=perf_cpu_map__nr (rhs)) +return false; + +for (int idx =0 ; idx < nr; idx++) { +if (RC_CHK_ACCESS (lhs)->map[idx].cpu !=RC_CHK_ACCESS (rhs)->map[idx].cpu) +return false; + } +return true; +} +struct perf_cpu perf_cpu_map__max (const struct perf_cpu_map *map) {struct perf_cpu result = {
tools/lib/perf/include/perf/cpumap.h +2 -0
@@ -28,6 +28,8 @@ LIBPERF_API int perf_cpu_map__nr(const struct perf_cpu_map *cpus); LIBPERF_API bool perf_cpu_map__empty(const struct perf_cpu_map *map); LIBPERF_API struct perf_cpu perf_cpu_map__max(const struct perf_cpu_map *map); LIBPERF_API bool perf_cpu_map__has(const struct perf_cpu_map *map, struct perf_cpu cpu); +LIBPERF_API bool perf_cpu_map__equal(const struct perf_cpu_map *lhs, +const struct perf_cpu_map *rhs);#define perf_cpu_map__for_each_cpu (cpu, idx, cpus) \ for ((idx) = 0, (cpu) = perf_cpu_map__cpu(cpus, idx); \
tools/perf/tests/cpumap.c +37 -0
@@ -211,11 +211,48 @@ static int test__cpu_map_intersect(struct test_suite *test __maybe_unused,return ret; } +static int test__cpu_map_equal (struct test_suite *test__maybe_unused ,int subtest__maybe_unused ) +{ +struct perf_cpu_map *any =perf_cpu_map__dummy_new (); +struct perf_cpu_map *one =perf_cpu_map__new ("1" ); +struct perf_cpu_map *two =perf_cpu_map__new ("2" ); +struct perf_cpu_map *empty =perf_cpu_map__intersect (one, two); +struct perf_cpu_map *pair =perf_cpu_map__new ("1-2" ); +struct perf_cpu_map *tmp; +struct perf_cpu_map *maps[] = {empty, any, one, two, pair}; + +for (size_t i =0 ; i <ARRAY_SIZE (maps); i++) { +/* Maps equal themself. */ +TEST_ASSERT_VAL ("equal" ,perf_cpu_map__equal (maps[i], maps[i])); +for (size_t j =0 ; j <ARRAY_SIZE (maps); j++) { +/* Maps dont't equal each other. */ +if (i == j) +continue ; +TEST_ASSERT_VAL ("not equal" , !perf_cpu_map__equal (maps[i], maps[j])); + } + } + +/* Maps equal made maps. */ + tmp =perf_cpu_map__merge (perf_cpu_map__get (one), two); +TEST_ASSERT_VAL ("pair" ,perf_cpu_map__equal (pair, tmp)); +perf_cpu_map__put (tmp); + + tmp =perf_cpu_map__intersect (pair, one); +TEST_ASSERT_VAL ("one" ,perf_cpu_map__equal (one, tmp)); +perf_cpu_map__put (tmp); + +for (size_t i =0 ; i <ARRAY_SIZE (maps); i++) +perf_cpu_map__put (maps[i]); + +return TEST_OK ; +} +static struct test_case tests__cpu_map[] = {TEST_CASE ("Synthesize cpu map" , cpu_map_synthesize),TEST_CASE ("Print cpu map" , cpu_map_print),TEST_CASE ("Merge cpu map" , cpu_map_merge),TEST_CASE ("Intersect cpu map" , cpu_map_intersect), +TEST_CASE ("Equal cpu map" , cpu_map_equal), { .name =NULL , } };
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Arnaldo Carvalho de Melo
maintainerRe: [PATCH v4 03/35] perf cpumap: Add equal function
Em Fri, May 26, 2023 at 02:53:38PM -0700, Ian Rogers escreveu:
Quoted 26 lines
Equality is a useful property to compare after merging and intersecting maps.
Signed-off-by Ian Rogers <[email protected]>Reviewed-by Kan Liang <[email protected]>--- tools/lib/perf/cpumap.c | 21 ++++++++++++++++ tools/lib/perf/include/perf/cpumap.h | 2 ++ tools/perf/tests/cpumap.c | 37 ++++++++++++++++++++++++++++ 3 files changed, 60 insertions(+)tools/lib/perf/cpumap.c +7
-0@@ -321,6 +321,27 @@ bool perf_cpu_map__has(const struct perf_cpu_map *cpus, struct perf_cpu cpu)return perf_cpu_map__idx (cpus, cpu) !=-1 ; } +bool perf_cpu_map__equal (const struct perf_cpu_map *lhs,const struct perf_cpu_map *rhs) +{ +int nr; + +if (lhs == rhs) +return true; +
tools/lib/perf/cpumap.c bool perf_cpu_map__has(const struct perf_cpu_map *cpus, struct perf_cpu cpu) · line 340
Quoted 10 lines
tools/lib/perf/cpumap.c +10
-0+if (!lhs || !rhs) +return false; + + nr =perf_cpu_map__nr (lhs); +if (nr !=perf_cpu_map__nr (rhs)) +return false; + +for (int idx =0 ; idx < nr; idx++) { +if (RC_CHK_ACCESS (lhs)->map[idx].cpu !=RC_CHK_ACCESS (rhs)->map[idx].cpu) +return false;
Don't we have an accessor to avoid this RC_CHK_ACCESS()-> access?
Quoted 76 lines
tools/lib/perf/cpumap.c +4
-0+ } +return true; +} +struct perf_cpu perf_cpu_map__max (const struct perf_cpu_map *map) {struct perf_cpu result = {tools/lib/perf/include/perf/cpumap.h +2
-0@@ -28,6 +28,8 @@ LIBPERF_API int perf_cpu_map__nr(const struct perf_cpu_map *cpus); LIBPERF_API bool perf_cpu_map__empty(const struct perf_cpu_map *map); LIBPERF_API struct perf_cpu perf_cpu_map__max(const struct perf_cpu_map *map); LIBPERF_API bool perf_cpu_map__has(const struct perf_cpu_map *map, struct perf_cpu cpu); +LIBPERF_API bool perf_cpu_map__equal(const struct perf_cpu_map *lhs, +const struct perf_cpu_map *rhs);#define perf_cpu_map__for_each_cpu (cpu, idx, cpus) \ for ((idx) = 0, (cpu) = perf_cpu_map__cpu(cpus, idx); \tools/perf/tests/cpumap.c +37
-0@@ -211,11 +211,48 @@ static int test__cpu_map_intersect(struct test_suite *test __maybe_unused,return ret; } +static int test__cpu_map_equal (struct test_suite *test__maybe_unused ,int subtest__maybe_unused ) +{ +struct perf_cpu_map *any =perf_cpu_map__dummy_new (); +struct perf_cpu_map *one =perf_cpu_map__new ("1" ); +struct perf_cpu_map *two =perf_cpu_map__new ("2" ); +struct perf_cpu_map *empty =perf_cpu_map__intersect (one, two); +struct perf_cpu_map *pair =perf_cpu_map__new ("1-2" ); +struct perf_cpu_map *tmp; +struct perf_cpu_map *maps[] = {empty, any, one, two, pair}; + +for (size_t i =0 ; i <ARRAY_SIZE (maps); i++) { +/* Maps equal themself. */ +TEST_ASSERT_VAL ("equal" ,perf_cpu_map__equal (maps[i], maps[i])); +for (size_t j =0 ; j <ARRAY_SIZE (maps); j++) { +/* Maps dont't equal each other. */ +if (i == j) +continue ; +TEST_ASSERT_VAL ("not equal" , !perf_cpu_map__equal (maps[i], maps[j])); + } + } + +/* Maps equal made maps. */ + tmp =perf_cpu_map__merge (perf_cpu_map__get (one), two); +TEST_ASSERT_VAL ("pair" ,perf_cpu_map__equal (pair, tmp)); +perf_cpu_map__put (tmp); + + tmp =perf_cpu_map__intersect (pair, one); +TEST_ASSERT_VAL ("one" ,perf_cpu_map__equal (one, tmp)); +perf_cpu_map__put (tmp); + +for (size_t i =0 ; i <ARRAY_SIZE (maps); i++) +perf_cpu_map__put (maps[i]); + +return TEST_OK ; +} +static struct test_case tests__cpu_map[] = {TEST_CASE ("Synthesize cpu map" , cpu_map_synthesize),TEST_CASE ("Print cpu map" , cpu_map_print),TEST_CASE ("Merge cpu map" , cpu_map_merge),TEST_CASE ("Intersect cpu map" , cpu_map_intersect), +TEST_CASE ("Equal cpu map" , cpu_map_equal), { .name =NULL , } };Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Signature · 1 line
- Arnaldo
Arnaldo Carvalho de Melo
maintainerRe: [PATCH v4 03/35] perf cpumap: Add equal function
Em Fri, May 26, 2023 at 10:32:12PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 31 lines
Em Fri, May 26, 2023 at 02:53:38PM -0700, Ian Rogers escreveu:
Quoted 30 lines
Equality is a useful property to compare after merging and intersecting maps.
Signed-off-by Ian Rogers <[email protected]>Reviewed-by Kan Liang <[email protected]>--- tools/lib/perf/cpumap.c | 21 ++++++++++++++++ tools/lib/perf/include/perf/cpumap.h | 2 ++ tools/perf/tests/cpumap.c | 37 ++++++++++++++++++++++++++++ 3 files changed, 60 insertions(+)tools/lib/perf/cpumap.c +11
-0@@ -321,6 +321,27 @@ bool perf_cpu_map__has(const struct perf_cpu_map *cpus, struct perf_cpu cpu)return perf_cpu_map__idx (cpus, cpu) !=-1 ; } +bool perf_cpu_map__equal (const struct perf_cpu_map *lhs,const struct perf_cpu_map *rhs) +{ +int nr; + +if (lhs == rhs) +return true; + +if (!lhs || !rhs) +return false; + + nr =perf_cpu_map__nr (lhs);
Quoted 8 lines
tools/lib/perf/cpumap.c bool perf_cpu_map__has(const struct perf_cpu_map *cpus, struct perf_cpu cpu) · line 340
Quoted 6 lines
tools/lib/perf/cpumap.c +6
-0+if (nr !=perf_cpu_map__nr (rhs)) +return false; + +for (int idx =0 ; idx < nr; idx++) { +if (RC_CHK_ACCESS (lhs)->map[idx].cpu !=RC_CHK_ACCESS (rhs)->map[idx].cpu) +return false;Don't we have an accessor to avoid this RC_CHK_ACCESS()-> access?
In the following patch you use it:
+bool perf_cpu_map__has_any_cpu(const struct perf_cpu_map *map) +{
+ return map && perf_cpu_map__cpu(map, 0).cpu == -1;
+}
But it does extra checks you did already:
struct perf_cpu perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx)
{
struct perf_cpu result = {
.cpu = -1
};
if (cpus && idx < RC_CHK_ACCESS(cpus)->nr)
return RC_CHK_ACCESS(cpus)->map[idx];
return result;
}Usually we have:
struct perf_cpu __perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx)
{
return RC_CHK_ACCESS(cpus)->map[idx];
}struct perf_cpu perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx)
{
struct perf_cpu result = {
.cpu = -1
};
if (cpus && idx < __perf_cpu_map__nr(cpus))
return __perf_cpu_map__cpu(cpus, idx);
return result;
}Then you would have:
bool perf_cpu_map__equal(const struct perf_cpu_map *lhs, const struct perf_cpu_map *rhs)
{
int nr;
if (lhs == rhs)
return true;
if (!lhs || !rhs)
return false;
nr = __perf_cpu_map__nr(lhs); // no need to check lhs again for NULL
if (nr != __perf_cpu_map__nr(rhs)) // ditto for rhs
return false;
for (int idx = 0; idx < nr; idx++) {
if (__perf_cpu_map__cpu(lhs, idx)->cpu != __perf_cpu_map__cpu(rhs, idx)->cpu)
return false;Quoted 80 lines
Quoted 76 lines
tools/lib/perf/cpumap.c +4
-0+ } +return true; +} +struct perf_cpu perf_cpu_map__max (const struct perf_cpu_map *map) {struct perf_cpu result = {tools/lib/perf/include/perf/cpumap.h +2
-0@@ -28,6 +28,8 @@ LIBPERF_API int perf_cpu_map__nr(const struct perf_cpu_map *cpus); LIBPERF_API bool perf_cpu_map__empty(const struct perf_cpu_map *map); LIBPERF_API struct perf_cpu perf_cpu_map__max(const struct perf_cpu_map *map); LIBPERF_API bool perf_cpu_map__has(const struct perf_cpu_map *map, struct perf_cpu cpu); +LIBPERF_API bool perf_cpu_map__equal(const struct perf_cpu_map *lhs, +const struct perf_cpu_map *rhs);#define perf_cpu_map__for_each_cpu (cpu, idx, cpus) \ for ((idx) = 0, (cpu) = perf_cpu_map__cpu(cpus, idx); \tools/perf/tests/cpumap.c +37
-0@@ -211,11 +211,48 @@ static int test__cpu_map_intersect(struct test_suite *test __maybe_unused,return ret; } +static int test__cpu_map_equal (struct test_suite *test__maybe_unused ,int subtest__maybe_unused ) +{ +struct perf_cpu_map *any =perf_cpu_map__dummy_new (); +struct perf_cpu_map *one =perf_cpu_map__new ("1" ); +struct perf_cpu_map *two =perf_cpu_map__new ("2" ); +struct perf_cpu_map *empty =perf_cpu_map__intersect (one, two); +struct perf_cpu_map *pair =perf_cpu_map__new ("1-2" ); +struct perf_cpu_map *tmp; +struct perf_cpu_map *maps[] = {empty, any, one, two, pair}; + +for (size_t i =0 ; i <ARRAY_SIZE (maps); i++) { +/* Maps equal themself. */ +TEST_ASSERT_VAL ("equal" ,perf_cpu_map__equal (maps[i], maps[i])); +for (size_t j =0 ; j <ARRAY_SIZE (maps); j++) { +/* Maps dont't equal each other. */ +if (i == j) +continue ; +TEST_ASSERT_VAL ("not equal" , !perf_cpu_map__equal (maps[i], maps[j])); + } + } + +/* Maps equal made maps. */ + tmp =perf_cpu_map__merge (perf_cpu_map__get (one), two); +TEST_ASSERT_VAL ("pair" ,perf_cpu_map__equal (pair, tmp)); +perf_cpu_map__put (tmp); + + tmp =perf_cpu_map__intersect (pair, one); +TEST_ASSERT_VAL ("one" ,perf_cpu_map__equal (one, tmp)); +perf_cpu_map__put (tmp); + +for (size_t i =0 ; i <ARRAY_SIZE (maps); i++) +perf_cpu_map__put (maps[i]); + +return TEST_OK ; +} +static struct test_case tests__cpu_map[] = {TEST_CASE ("Synthesize cpu map" , cpu_map_synthesize),TEST_CASE ("Print cpu map" , cpu_map_print),TEST_CASE ("Merge cpu map" , cpu_map_merge),TEST_CASE ("Intersect cpu map" , cpu_map_intersect), +TEST_CASE ("Equal cpu map" , cpu_map_equal), { .name =NULL , } };Signature · 1 line
2.41.0.rc0.172.g3f132b7071-googSignature · 1 line
- Arnaldo
Signature · 1 line
- Arnaldo
Ian Rogers
reviewerRe: [PATCH v4 03/35] perf cpumap: Add equal function
Quoted 85 lines · Arnaldo Carvalho de Melo
Em Fri, May 26, 2023 at 10:32:12PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 31 lines
Em Fri, May 26, 2023 at 02:53:38PM -0700, Ian Rogers escreveu:
Quoted 30 lines
Equality is a useful property to compare after merging and intersecting maps.
Signed-off-by Ian Rogers <[email protected]>Reviewed-by Kan Liang <[email protected]>--- tools/lib/perf/cpumap.c | 21 ++++++++++++++++ tools/lib/perf/include/perf/cpumap.h | 2 ++ tools/perf/tests/cpumap.c | 37 ++++++++++++++++++++++++++++ 3 files changed, 60 insertions(+)tools/lib/perf/cpumap.c +11
-0@@ -321,6 +321,27 @@ bool perf_cpu_map__has(const struct perf_cpu_map *cpus, struct perf_cpu cpu)return perf_cpu_map__idx (cpus, cpu) !=-1 ; } +bool perf_cpu_map__equal (const struct perf_cpu_map *lhs,const struct perf_cpu_map *rhs) +{ +int nr; + +if (lhs == rhs) +return true; + +if (!lhs || !rhs) +return false; + + nr =perf_cpu_map__nr (lhs);Quoted 8 lines
tools/lib/perf/cpumap.c bool perf_cpu_map__has(const struct perf_cpu_map *cpus, struct perf_cpu cpu) · line 340
Quoted 6 lines
tools/lib/perf/cpumap.c +6
-0+if (nr !=perf_cpu_map__nr (rhs)) +return false; + +for (int idx =0 ; idx < nr; idx++) { +if (RC_CHK_ACCESS (lhs)->map[idx].cpu !=RC_CHK_ACCESS (rhs)->map[idx].cpu) +return false;Don't we have an accessor to avoid this RC_CHK_ACCESS()-> access?
In the following patch you use it:
tools/lib/perf/cpumap.c +4
-0+bool perf_cpu_map__has_any_cpu (const struct perf_cpu_map *map) +{ +return map &&perf_cpu_map__cpu (map,0 ).cpu ==-1 ; +}But it does extra checks you did already:
struct
perf_cpuperf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx){ struct perf_cpu result = { .cpu = -1 }; if (cpus && idx < RC_CHK_ACCESS(cpus)->nr) return RC_CHK_ACCESS(cpus)->map[idx]; return result; }Usually we have:
struct
perf_cpu__perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx){ return RC_CHK_ACCESS(cpus)->map[idx]; }struct
perf_cpuperf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx){ struct perf_cpu result = { .cpu = -1 }; if (cpus && idx < __perf_cpu_map__nr(cpus)) return __perf_cpu_map__cpu(cpus, idx); return result; }Then you would have:
bool
perf_cpu_map__equal(const struct perf_cpu_map *lhs, const struct perf_cpu_map *rhs)
Quoted 17 lines
{
int nr;
if (lhs == rhs)
return true;
if (!lhs || !rhs)
return false;
nr = __perf_cpu_map__nr(lhs); // no need to check lhs again for NULL
if (nr != __perf_cpu_map__nr(rhs)) // ditto for rhs
return false;
for (int idx = 0; idx < nr; idx++) {
if (__perf_cpu_map__cpu(lhs, idx)->cpu != __perf_cpu_map__cpu(rhs, idx)->cpu)
return false;Thanks, I'll update for v5. Fwiw, on intersect, I keep forgetting to mention that this was sent previously as:
https://lore.kernel.org/lkml/[email protected]/
Ian
Quoted 84 lines
Quoted 73 lines
Quoted 73 lines
tools/lib/perf/cpumap.c +4
-0+ } +return true; +} +struct perf_cpu perf_cpu_map__max (const struct perf_cpu_map *map) {struct perf_cpu result = {tools/lib/perf/include/perf/cpumap.h +2
-0@@ -28,6 +28,8 @@ LIBPERF_API int perf_cpu_map__nr(const struct perf_cpu_map *cpus); LIBPERF_API bool perf_cpu_map__empty(const struct perf_cpu_map *map); LIBPERF_API struct perf_cpu perf_cpu_map__max(const struct perf_cpu_map *map); LIBPERF_API bool perf_cpu_map__has(const struct perf_cpu_map *map, struct perf_cpu cpu); +LIBPERF_API bool perf_cpu_map__equal(const struct perf_cpu_map *lhs, +const struct perf_cpu_map *rhs);#define perf_cpu_map__for_each_cpu (cpu, idx, cpus) \ for ((idx) = 0, (cpu) = perf_cpu_map__cpu(cpus, idx); \tools/perf/tests/cpumap.c +37
-0@@ -211,11 +211,48 @@ static int test__cpu_map_intersect(struct test_suite *test __maybe_unused,return ret; } +static int test__cpu_map_equal (struct test_suite *test__maybe_unused ,int subtest__maybe_unused ) +{ +struct perf_cpu_map *any =perf_cpu_map__dummy_new (); +struct perf_cpu_map *one =perf_cpu_map__new ("1" ); +struct perf_cpu_map *two =perf_cpu_map__new ("2" ); +struct perf_cpu_map *empty =perf_cpu_map__intersect (one, two); +struct perf_cpu_map *pair =perf_cpu_map__new ("1-2" ); +struct perf_cpu_map *tmp; +struct perf_cpu_map *maps[] = {empty, any, one, two, pair}; + +for (size_t i =0 ; i <ARRAY_SIZE (maps); i++) { +/* Maps equal themself. */ +TEST_ASSERT_VAL ("equal" ,perf_cpu_map__equal (maps[i], maps[i])); +for (size_t j =0 ; j <ARRAY_SIZE (maps); j++) { +/* Maps dont't equal each other. */ +if (i == j) +continue ; +TEST_ASSERT_VAL ("not equal" , !perf_cpu_map__equal (maps[i], maps[j])); + } + } + +/* Maps equal made maps. */ + tmp =perf_cpu_map__merge (perf_cpu_map__get (one), two); +TEST_ASSERT_VAL ("pair" ,perf_cpu_map__equal (pair, tmp)); +perf_cpu_map__put (tmp); + + tmp =perf_cpu_map__intersect (pair, one); +TEST_ASSERT_VAL ("one" ,perf_cpu_map__equal (one, tmp)); +perf_cpu_map__put (tmp); + +for (size_t i =0 ; i <ARRAY_SIZE (maps); i++) +perf_cpu_map__put (maps[i]); + +return TEST_OK ; +} +static struct test_case tests__cpu_map[] = {TEST_CASE ("Synthesize cpu map" , cpu_map_synthesize),TEST_CASE ("Print cpu map" , cpu_map_print),TEST_CASE ("Merge cpu map" , cpu_map_merge),TEST_CASE ("Intersect cpu map" , cpu_map_intersect), +TEST_CASE ("Equal cpu map" , cpu_map_equal), { .name =NULL , } };Quoted 7 lines
Quoted 3 lines
--
2.41.0.rc0.172.g3f132b7071-goog--
- Arnaldo
--
- Arnaldo
Ian Rogers
reviewerPATCH v4 4/35 libperf cpumap: Add "any CPU"/dummy test function
In reply to Ian RogersIt is common in the code currently to test a map for "empty" when in fact the "any CPU"/dummy value of -1 is being sought. Add a new function to enable this and document the behavior of two other functions. The term "any CPU" comes from perf_event_open, where the value is consumed, but it is more typical in the code to see this value/map referred to as the dummy value. This could be misleading due to the dummy event and also dummy not being intention revealing, so it is hoped to migrate the code to referring to this as "any CPU".
--- tools/lib/perf/cpumap.c | 5 +++++ tools/lib/perf/include/perf/cpumap.h | 10 ++++++++++ 2 files changed, 15 insertions(+)
tools/lib/perf/cpumap.c +5 -0
@@ -342,6 +342,11 @@ bool perf_cpu_map__equal(const struct perf_cpu_map *lhs, const struct perf_cpu_mreturn true; } +bool perf_cpu_map__has_any_cpu (const struct perf_cpu_map *map) +{ +return map &&perf_cpu_map__cpu (map,0 ).cpu ==-1 ; +} +struct perf_cpu perf_cpu_map__max (const struct perf_cpu_map *map) {struct perf_cpu result = {
tools/lib/perf/include/perf/cpumap.h +10 -0
@@ -13,6 +13,9 @@ struct perf_cpu {struct perf_cpu_map ; +/** +* perf_cpu_map__dummy_new - a map with a singular "any CPU"/dummy -1 value. +*/ LIBPERF_API struct perf_cpu_map *perf_cpu_map__dummy_new(void); LIBPERF_API struct perf_cpu_map *perf_cpu_map__default_new(void); LIBPERF_API struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list);
@@ -25,11 +28,18 @@ LIBPERF_API struct perf_cpu_map *perf_cpu_map__intersect(struct perf_cpu_map *or LIBPERF_API void perf_cpu_map__put(struct perf_cpu_map *map); LIBPERF_API struct perf_cpu perf_cpu_map__cpu(const struct perf_cpu_map *cpus, int idx); LIBPERF_API int perf_cpu_map__nr(const struct perf_cpu_map *cpus); +/** +* perf_cpu_map__empty - is map either empty or the "any CPU"/dummy value. +*/ LIBPERF_API bool perf_cpu_map__empty(const struct perf_cpu_map *map); LIBPERF_API struct perf_cpu perf_cpu_map__max(const struct perf_cpu_map *map); LIBPERF_API bool perf_cpu_map__has(const struct perf_cpu_map *map, struct perf_cpu cpu); LIBPERF_API bool perf_cpu_map__equal(const struct perf_cpu_map *lhs,const struct perf_cpu_map *rhs); +/** +* perf_cpu_map__any_cpu - Does the map contain the "any CPU"/dummy -1 value? +*/ +LIBPERF_API bool perf_cpu_map__has_any_cpu(const struct perf_cpu_map *map);#define perf_cpu_map__for_each_cpu (cpu, idx, cpus) \ for ((idx) = 0, (cpu) = perf_cpu_map__cpu(cpus, idx); \
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 5/35 perf pmu: Detect ARM and hybrid PMUs with sysfs
In reply to Ian Rogersis_arm_pmu_core detects a core PMU via the presence of a "cpus" file rather than a "cpumask" file. This pattern holds for hybrid PMUs so rename the function and remove redundant perf_pmu__is_hybrid tests.
Add a new helper is_pmu_hybrid similar to is_pmu_core.
--- tools/perf/util/pmu.c | 29 ++++++++++++++++++----------- tools/perf/util/pmu.h | 1 + 2 files changed, 19 insertions(+), 11 deletions(-)
tools/perf/util/pmu.c +18 -11
@@ -643,12 +643,14 @@ static char *pmu_id(const char *name)return str; } -/* -* PMU CORE devices have different name other than cpu in sysfs on some -* platforms. -* Looking for possible sysfs files to identify the arm core device. +/** +* is_sysfs_pmu_core() - PMU CORE devices have different name other than cpu in +* sysfs on some platforms like ARM or Intel hybrid. Looking for +* possible the cpus file in sysfs files to identify whether this is a +* core device. +* @name: The PMU name such as "cpu_atom". */ -static int is_arm_pmu_core (const char *name) +static int is_sysfs_pmu_core (const char *name) {char path[PATH_MAX ];
@@ -814,7 +816,7 @@ void pmu_add_cpu_aliases_table(struct list_head *head, struct perf_pmu *pmu,struct pmu_add_cpu_aliases_map_data data = { .head = head, .name = pmu->name, - .cpu_name =is_arm_pmu_core (pmu->name) ? pmu->name :"cpu" , + .cpu_name =is_sysfs_pmu_core (pmu->name) ? pmu->name :"cpu" , .pmu = pmu, };
@@ -1647,22 +1649,27 @@ static int cmp_sevent(const void *a, const void *b)bool is_pmu_core (const char *name) { -return !strcmp (name,"cpu" ) ||is_arm_pmu_core (name); +return !strcmp (name,"cpu" ) ||is_sysfs_pmu_core (name); +} + +bool is_pmu_hybrid (const char *name) +{ +return !strcmp (name,"cpu_atom" ) || !strcmp (name,"cpu_core" ); }bool perf_pmu__supports_legacy_cache (const struct perf_pmu *pmu) { -return is_pmu_core (pmu->name) ||perf_pmu__is_hybrid (pmu->name); +return is_pmu_core (pmu->name); }bool perf_pmu__supports_wildcard_numeric (const struct perf_pmu *pmu) { -return is_pmu_core (pmu->name) ||perf_pmu__is_hybrid (pmu->name); +return is_pmu_core (pmu->name); }bool perf_pmu__auto_merge_stats (const struct perf_pmu *pmu) { -return !perf_pmu__is_hybrid (pmu->name); +return !is_pmu_hybrid (pmu->name); } static bool pmu_alias_is_duplicate(struct sevent *alias_a,
@@ -1716,7 +1723,7 @@ void print_pmu_events(const struct print_callbacks *print_cb, void *print_state) pmu =NULL ; j =0 ;while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { -bool is_cpu =is_pmu_core (pmu->name) ||perf_pmu__is_hybrid (pmu->name); +bool is_cpu =is_pmu_core (pmu->name);list_for_each_entry (event, &pmu->aliases, list) { aliases[j].event = event;
tools/perf/util/pmu.h +1 -0
@@ -220,6 +220,7 @@ void perf_pmu__del_formats(struct list_head *formats);struct perf_pmu *perf_pmu__scan (struct perf_pmu *pmu);bool is_pmu_core (const char *name); +bool is_pmu_hybrid (const char *name);bool perf_pmu__supports_legacy_cache (const struct perf_pmu *pmu);bool perf_pmu__supports_wildcard_numeric (const struct perf_pmu *pmu);bool perf_pmu__auto_merge_stats (const struct perf_pmu *pmu);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 6/35 perf pmu: Add is_core to pmu
In reply to Ian RogersCache is_pmu_core in the pmu to avoid recomputation.
--- tools/perf/util/pmu.c | 7 ++++--- tools/perf/util/pmu.h | 7 +++++++ 2 files changed, 11 insertions(+), 3 deletions(-)
tools/perf/util/pmu.c +4 -3
@@ -952,6 +952,7 @@ static struct perf_pmu *pmu_lookup(int dirfd, const char *lookup_name) } pmu->type = type; + pmu->is_core =is_pmu_core (name); pmu->is_uncore =pmu_is_uncore (dirfd, name);if (pmu->is_uncore) pmu->id =pmu_id (name);
@@ -1659,12 +1660,12 @@ bool is_pmu_hybrid(const char *name)bool perf_pmu__supports_legacy_cache (const struct perf_pmu *pmu) { -return is_pmu_core (pmu->name); +return pmu->is_core; }bool perf_pmu__supports_wildcard_numeric (const struct perf_pmu *pmu) { -return is_pmu_core (pmu->name); +return pmu->is_core; }bool perf_pmu__auto_merge_stats (const struct perf_pmu *pmu)
@@ -1723,7 +1724,7 @@ void print_pmu_events(const struct print_callbacks *print_cb, void *print_state) pmu =NULL ; j =0 ;while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { -bool is_cpu =is_pmu_core (pmu->name); +bool is_cpu = pmu->is_core;list_for_each_entry (event, &pmu->aliases, list) { aliases[j].event = event;
tools/perf/util/pmu.h +7 -0
@@ -59,6 +59,13 @@ struct perf_pmu {* @selectable: Can the PMU name be selected as if it were an event? */ bool selectable; +/** +* @is_core: Is the PMU the core CPU PMU? Determined by the name being +* "cpu" or by the presence of +* <sysfs>/bus/event_source/devices/<name>/cpus. There may be >1 core +* PMU on systems like Intel hybrid. +*/ +bool is_core;/** * @is_uncore: Is the PMU not within the CPU core? Determined by the * presence of <sysfs>/bus/event_source/devices/<name>/cpumask.
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 7/35 perf evsel: Add is_pmu_core inorder to interpret own_cpus
In reply to Ian RogersThe behaviour of handling cpu maps varies for core and other PMUs. For core PMUs the cpu map lists all valid CPUs, whereas for other PMUs the map is the default CPU. Add a flag in the evsel to indicate if a PMU is core to help with later interpreting of the cpu maps and populate it when the evsel is created during parsing. When propagating cpu maps, core PMUs should intersect the cpu map of the PMU with the user requested one.
--- tools/lib/perf/include/internal/evsel.h | 9 +++++++++ tools/perf/util/evsel.c | 1 + tools/perf/util/parse-events.c | 1 + 3 files changed, 11 insertions(+)
tools/lib/perf/include/internal/evsel.h +9 -0
@@ -41,7 +41,14 @@ struct perf_sample_id {struct perf_evsel {struct list_head node;struct perf_event_attr attr; +/** The commonly used cpu map of CPUs the event should be opened upon, etc. */ struct perf_cpu_map *cpus; +/** +* The cpu map read from the PMU. For core PMUs this is the list of all +* CPUs the event can be opened upon. For other PMUs this is the default +* cpu map for opening the event on, for example, the first CPU on a +* socket for an uncore event. +*/ struct perf_cpu_map *own_cpus;struct perf_thread_map *threads;struct xyarray *fd;
@@ -65,6 +72,8 @@ struct perf_evsel {* i.e. it cannot be the 'any CPU' value of -1. */ bool requires_cpu; +/** Is the PMU for the event a core one? Effects the handling of own_cpus. */ +bool is_pmu_core;int idx; };
tools/perf/util/evsel.c +1 -0
@@ -415,6 +415,7 @@ struct evsel *evsel__clone(struct evsel *orig) evsel->core.nr_members = orig->core.nr_members; evsel->core.system_wide = orig->core.system_wide; evsel->core.requires_cpu = orig->core.requires_cpu; + evsel->core.is_pmu_core = orig->core.is_pmu_core;if (orig->name) { evsel->name =strdup (orig->name);
tools/perf/util/parse-events.c +1 -0
@@ -260,6 +260,7 @@ __add_event(struct list_head *list, int *idx, evsel->core.cpus = cpus; evsel->core.own_cpus =perf_cpu_map__get (cpus); evsel->core.requires_cpu = pmu ? pmu->is_uncore : false; + evsel->core.is_pmu_core = pmu ? pmu->is_core : false; evsel->auto_merge_stats = auto_merge_stats; evsel->pmu = pmu; evsel->pmu_name = pmu && pmu->name ?strdup (pmu->name) :NULL ;
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 8/35 perf pmu: Add CPU map for "cpu" PMUs
In reply to Ian RogersA typical "cpu" PMU has no "cpus" or "cpumask" file meaning the CPU map is set to NULL, which also encodes an empty CPU map. Update pmu_cpumask so that if the "cpu" PMU fails to load a CPU map, use a default of all online PMUs.
Remove const from cpu_map__online for the sake of reference counting.
--- tools/perf/util/cpumap.c | 4 ++-- tools/perf/util/cpumap.h | 4 ++-- tools/perf/util/pmu.c | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-)
tools/perf/util/cpumap.c +2 -2
@@ -663,9 +663,9 @@ size_t cpu_map__snprint_mask(struct perf_cpu_map *map, char *buf, size_t size)return ptr - buf; } -const struct perf_cpu_map *cpu_map__online (void )/* thread unsafe */ +struct perf_cpu_map *cpu_map__online (void )/* thread unsafe */ { -static const struct perf_cpu_map *online =NULL ; +static struct perf_cpu_map *online;if (!online) online =perf_cpu_map__new (NULL );/* from /sys/devices/system/cpu/online */
tools/perf/util/cpumap.h +2 -2
@@ -48,7 +48,7 @@ struct perf_cpu_map *cpu_map__new_data(const struct perf_record_cpu_map_data *dasize_t cpu_map__snprint (struct perf_cpu_map *map,char *buf,size_t size);size_t cpu_map__snprint_mask (struct perf_cpu_map *map,char *buf,size_t size);size_t cpu_map__fprintf (struct perf_cpu_map *map,FILE *fp); -const struct perf_cpu_map *cpu_map__online (void );/* thread unsafe */ +struct perf_cpu_map *cpu_map__online (void );/* thread unsafe */ int cpu__setup_cpunode_map (void );
@@ -59,7 +59,7 @@ struct perf_cpu cpu__max_present_cpu(void);/** * cpu_map__is_dummy - Events associated with a pid, rather than a CPU, use a single dummy map with an entry of -1. */ -static inline bool cpu_map__is_dummy (struct perf_cpu_map *cpus) +static inline bool cpu_map__is_dummy (const struct perf_cpu_map *cpus) {return perf_cpu_map__nr (cpus) ==1 &&perf_cpu_map__cpu (cpus,0 ).cpu ==-1 ; }
tools/perf/util/pmu.c +1 -1
@@ -610,7 +610,7 @@ static struct perf_cpu_map *pmu_cpumask(int dirfd, const char *name)return cpus; } -return NULL ; +return !strcmp (name,"cpu" ) ?perf_cpu_map__get (cpu_map__online ()) :NULL ; }static bool pmu_is_uncore (int dirfd,const char *name)
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 9/35 perf evlist: Propagate user CPU maps intersecting core PMU maps
In reply to Ian RogersThe CPU map for a non-core PMU gives a default CPU value for perf_event_open. For core PMUs the CPU map lists all CPUs the evsel may be opened on. If there are >1 core PMU, the CPU maps will list the CPUs for that core PMU, but the user_requested_cpus may contain CPUs that are invalid for the PMU and cause perf_event_open to fail. To avoid this, when propagating the CPU map for core PMUs intersect it with the CPU map of the PMU (the evsel's "own_cpus").
Add comments to __perf_evlist__propagate_maps to explain its somewhat complex behavior. Fix the related comments for system_wide in struct perf_evsel.
--- tools/lib/perf/evlist.c | 25 ++++++++++++++++++++----- tools/lib/perf/include/internal/evsel.h | 6 +++--- 2 files changed, 23 insertions(+), 8 deletions(-)
tools/lib/perf/evlist.c +20 -5
@@ -36,18 +36,33 @@ void perf_evlist__init(struct perf_evlist *evlist)static void __perf_evlist__propagate_maps (struct perf_evlist *evlist,struct perf_evsel *evsel) { -/* -* We already have cpus for evsel (via PMU sysfs) so -* keep it, if there's no target cpu list defined. -*/ if (evsel->system_wide) { +/* System wide: set the cpu map of the evsel to all online CPUs. */ perf_cpu_map__put (evsel->cpus); evsel->cpus =perf_cpu_map__new (NULL ); + }else if (evlist->has_user_cpus && evsel->is_pmu_core) { +/* +* User requested CPUs on a core PMU, ensure the requested CPUs +* are valid by intersecting with those of the PMU. +*/ +perf_cpu_map__put (evsel->cpus); + evsel->cpus =perf_cpu_map__intersect (evlist->user_requested_cpus, evsel->own_cpus); }else if (!evsel->own_cpus || evlist->has_user_cpus || - (!evsel->requires_cpu &&perf_cpu_map__empty (evlist->user_requested_cpus))) { + (!evsel->requires_cpu &&perf_cpu_map__has_any_cpu (evlist->user_requested_cpus))) { +/* +* The PMU didn't specify a default cpu map, this isn't a core +* event and the user requested CPUs or the evlist user +* requested CPUs have the "any CPU" (aka dummy) CPU value. In +* which case use the user requested CPUs rather than the PMU +* ones. +*/ perf_cpu_map__put (evsel->cpus); evsel->cpus =perf_cpu_map__get (evlist->user_requested_cpus); }else if (evsel->cpus != evsel->own_cpus) { +/* +* No user requested cpu map but the PMU cpu map doesn't match +* the evsel's. Reset it back to the PMU cpu map. +*/ perf_cpu_map__put (evsel->cpus); evsel->cpus =perf_cpu_map__get (evsel->own_cpus); }
tools/lib/perf/include/internal/evsel.h +3 -3
@@ -62,9 +62,9 @@ struct perf_evsel {int nr_members;/* * system_wide is for events that need to be on every CPU, irrespective -* of user requested CPUs or threads. Map propagation will set cpus to -* this event's own_cpus, whereby they will contribute to evlist -* all_cpus. +* of user requested CPUs or threads. Tha main example of this is the +* dummy event. Map propagation will set cpus for this event to all CPUs +* as software PMU events like dummy, have a CPU map that is empty. */ bool system_wide;/*
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 10/35 perf evlist: Allow has_user_cpus to be set on hybrid
In reply to Ian RogersNow that CPU map propagation only sets valid CPUs for core PMUs, there is no reason to disable "has_user_cpus" for hybrid.
--- tools/perf/util/evlist.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/evlist.c +1 -1
@@ -1067,7 +1067,7 @@ int evlist__create_maps(struct evlist *evlist, struct target *target)if (!cpus) goto out_delete_threads; - evlist->core.has_user_cpus = !!target->cpu_list && !target->hybrid; + evlist->core.has_user_cpus = !!target->cpu_list;perf_evlist__set_maps (&evlist->core, cpus, threads);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 11/35 perf target: Remove unused hybrid value
In reply to Ian RogersPreviously this was used to modify CPU map propagation, but it is now unnecessary as map propagation ensure core PMUs only have valid PMUs in the CPU map from user requested CPUs.
--- tools/perf/builtin-record.c | 2 -- tools/perf/builtin-stat.c | 1 - tools/perf/util/target.h | 1 - 3 files changed, 4 deletions(-)
tools/perf/builtin-record.c +0 -2
@@ -4204,8 +4204,6 @@ int cmd_record(int argc, const char **argv) goto out; } - rec->opts.target.hybrid =perf_pmu__has_hybrid (); -if (callchain_param.enabled && callchain_param.record_mode ==CALLCHAIN_FP )arch__add_leaf_frame_record_opts (&rec->opts);
tools/perf/builtin-stat.c +0 -1
@@ -2467,7 +2467,6 @@ int cmd_stat(int argc, const char **argv) goto out; } - target.hybrid =perf_pmu__has_hybrid ();if (evlist__create_maps (evsel_list, &target) <0 ) {if (target__has_task (&target)) {pr_err ("Problems finding threads of monitor\n" );
tools/perf/util/target.h +0 -1
@@ -17,7 +17,6 @@ struct target {bool default_per_cpu;bool per_thread;bool use_bpf; -bool hybrid;int initial_delay;const char *attr_map; };
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 12/35 perf tools: Warn if no user requested CPUs match PMU's CPUs
In reply to Ian RogersIn commit 1d3351e631fc ("perf tools: Enable on a list of CPUs for hybrid") perf on hybrid will warn if a user requested CPU doesn't match the PMU of the given event but only for hybrid PMUs. Make the logic generic for all PMUs and remove the hybrid logic.
Warn if a CPU is requested that isn't present/offline for events not on the core. Warn if a CPU is requested for a core PMU, but the CPU isn't within the cpu map of that PMU.
For example on a 16 (0-15) CPU system:
```
$ perf stat -e imc_free_running/data_read/,cycles -C 16 true
WARNING: A requested CPU in '16' is not supported by PMU 'uncore_imc_free_running_1' (CPUs 0-15) for event 'imc_free_running/data_read/'
WARNING: A requested CPU in '16' is not supported by PMU 'uncore_imc_free_running_0' (CPUs 0-15) for event 'imc_free_running/data_read/'
WARNING: A requested CPU in '16' is not supported by PMU 'cpu' (CPUs 0-15) for event 'cycles'
Performance counter stats for 'CPU(s) 16':
<not supported> MiB imc_free_running/data_read/
<not supported> cycles
0.000575312 seconds time elapsed
```Remove evlist__fix_hybrid_cpus that previously produced the warnings and also perf_pmu__cpus_match that worked with evlist__fix_hybrid_cpus to change CPU maps for hybrid CPUs, something that is no longer necessary as CPU map propagation properly intersects user requested CPUs with the core PMU's CPU map.
--- tools/perf/builtin-record.c | 6 +-- tools/perf/builtin-stat.c | 5 +-- tools/perf/util/evlist-hybrid.c | 74 --------------------------------- tools/perf/util/evlist-hybrid.h | 1 - tools/perf/util/evlist.c | 39 +++++++++++++++++ tools/perf/util/evlist.h | 2 + tools/perf/util/pmu.c | 33 --------------- tools/perf/util/pmu.h | 4 -- 8 files changed, 43 insertions(+), 121 deletions(-)
tools/perf/builtin-record.c +1 -5
@@ -4198,11 +4198,7 @@ int cmd_record(int argc, const char **argv)/* Enable ignoring missing threads when -u/-p option is defined. */ rec->opts.ignore_missing_thread = rec->opts.target.uid !=UINT_MAX || rec->opts.target.pid; -if (evlist__fix_hybrid_cpus (rec->evlist, rec->opts.target.cpu_list)) { -pr_err ("failed to use cpu list %s\n" , - rec->opts.target.cpu_list); - goto out; - } +evlist__warn_user_requested_cpus (rec->evlist, rec->opts.target.cpu_list);if (callchain_param.enabled && callchain_param.record_mode ==CALLCHAIN_FP )arch__add_leaf_frame_record_opts (&rec->opts);
tools/perf/builtin-stat.c +1 -4
@@ -2462,10 +2462,7 @@ int cmd_stat(int argc, const char **argv) } } -if (evlist__fix_hybrid_cpus (evsel_list, target.cpu_list)) { -pr_err ("failed to use cpu list %s\n" , target.cpu_list); - goto out; - } +evlist__warn_user_requested_cpus (evsel_list, target.cpu_list);if (evlist__create_maps (evsel_list, &target) <0 ) {if (target__has_task (&target)) {
tools/perf/util/evlist-hybrid.c +0 -74
77 lines not on this page · raw message
tools/perf/util/evlist-hybrid.h +0 -1
@@ -10,6 +10,5 @@int evlist__add_default_hybrid (struct evlist *evlist,bool precise);void evlist__warn_hybrid_group (struct evlist *evlist);bool evlist__has_hybrid (struct evlist *evlist); -int evlist__fix_hybrid_cpus (struct evlist *evlist,const char *cpu_list);#endif /* __PERF_EVLIST_HYBRID_H */
tools/perf/util/evlist.c +39 -0
@@ -2465,3 +2465,42 @@ void evlist__check_mem_load_aux(struct evlist *evlist) } } } + +/** +* evlist__warn_user_requested_cpus() - Check each evsel against requested CPUs +* and warn if the user CPU list is inapplicable for the event's PMU's +* CPUs. Not core PMUs list a CPU in sysfs, but this may be overwritten by a +* user requested CPU and so any online CPU is applicable. Core PMUs handle +* events on the CPUs in their list and otherwise the event isn't supported. +* @evlist: The list of events being checked. +* @cpu_list: The user provided list of CPUs. +*/ +void evlist__warn_user_requested_cpus (struct evlist *evlist,const char *cpu_list) +{ +struct perf_cpu_map *user_requested_cpus; +struct evsel *pos; + +if (!cpu_list) +return ; + + user_requested_cpus =perf_cpu_map__new (cpu_list); +if (!user_requested_cpus) +return ; + +evlist__for_each_entry (evlist, pos) { +struct perf_cpu_map *intersect, *to_test; +const struct perf_pmu *pmu =evsel__find_pmu (pos); + + to_test = pmu && pmu->is_core ? pmu->cpus :cpu_map__online (); + intersect =perf_cpu_map__intersect (to_test, user_requested_cpus); +if (!perf_cpu_map__equal (intersect, user_requested_cpus)) { +char buf[128 ]; + +cpu_map__snprint (to_test, buf,sizeof (buf)); +pr_warning ("WARNING: A requested CPU in '%s' is not supported by PMU '%s' (CPUs %s) for event '%s'\n" , + cpu_list, pmu ? pmu->name :"cpu" , buf,evsel__name (pos)); + } +perf_cpu_map__put (intersect); + } +perf_cpu_map__put (user_requested_cpus); +}
tools/perf/util/evlist.h +2 -0
@@ -447,4 +447,6 @@ struct evsel *evlist__find_evsel(struct evlist *evlist, int idx);int evlist__scnprintf_evsels (struct evlist *evlist,size_t size,char *bf);void evlist__check_mem_load_aux (struct evlist *evlist); +void evlist__warn_user_requested_cpus (struct evlist *evlist,const char *cpu_list); +#endif /* __PERF_EVLIST_H */
tools/perf/util/pmu.c +0 -33
@@ -2046,39 +2046,6 @@ int perf_pmu__match(char *pattern, char *name, char *tok)return 0 ; } -int perf_pmu__cpus_match (struct perf_pmu *pmu,struct perf_cpu_map *cpus, -struct perf_cpu_map **mcpus_ptr, -struct perf_cpu_map **ucpus_ptr) -{ -struct perf_cpu_map *pmu_cpus = pmu->cpus; -struct perf_cpu_map *matched_cpus, *unmatched_cpus; -struct perf_cpu cpu; -int i, matched_nr =0 , unmatched_nr =0 ; - - matched_cpus =perf_cpu_map__default_new (); -if (!matched_cpus) -return -1 ; - - unmatched_cpus =perf_cpu_map__default_new (); -if (!unmatched_cpus) { -perf_cpu_map__put (matched_cpus); -return -1 ; - } - -perf_cpu_map__for_each_cpu (cpu, i, cpus) { -if (!perf_cpu_map__has (pmu_cpus, cpu)) -RC_CHK_ACCESS (unmatched_cpus)->map[unmatched_nr++] = cpu; -else -RC_CHK_ACCESS (matched_cpus)->map[matched_nr++] = cpu; - } - -perf_cpu_map__set_nr (unmatched_cpus, unmatched_nr); -perf_cpu_map__set_nr (matched_cpus, matched_nr); - *mcpus_ptr = matched_cpus; - *ucpus_ptr = unmatched_cpus; -return 0 ; -} -double __weak perf_pmu__cpu_slots_per_cycle (void ) {return NAN ;
tools/perf/util/pmu.h +0 -4
@@ -265,10 +265,6 @@ void perf_pmu__warn_invalid_formats(struct perf_pmu *pmu);bool perf_pmu__has_hybrid (void );int perf_pmu__match (char *pattern,char *name,char *tok); -int perf_pmu__cpus_match (struct perf_pmu *pmu,struct perf_cpu_map *cpus, -struct perf_cpu_map **mcpus_ptr, -struct perf_cpu_map **ucpus_ptr); -char *pmu_find_real_name (const char *name);char *pmu_find_alias_name (const char *name);double perf_pmu__cpu_slots_per_cycle (void );
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 13/35 perf evlist: Remove evlist__warn_hybrid_group
In reply to Ian RogersParse events now corrects PMU groups in parse_events__sort_events_and_fix_groups and so this warning is no longer possible.
--- tools/perf/builtin-stat.c | 3 --- tools/perf/util/evlist-hybrid.c | 32 -------------------------------- tools/perf/util/evlist-hybrid.h | 1 - 3 files changed, 36 deletions(-)
tools/perf/builtin-stat.c +0 -3
@@ -187,9 +187,6 @@ static void evlist__check_cpu_maps(struct evlist *evlist) {struct evsel *evsel, *warned_leader =NULL ; -if (evlist__has_hybrid (evlist)) -evlist__warn_hybrid_group (evlist); -evlist__for_each_entry (evlist, evsel) {struct evsel *leader =evsel__leader (evsel);
tools/perf/util/evlist-hybrid.c +0 -32
@@ -41,38 +41,6 @@ int evlist__add_default_hybrid(struct evlist *evlist, bool precise)return 0 ; } -static bool group_hybrid_conflict (struct evsel *leader) -{ -struct evsel *pos, *prev =NULL ; - -for_each_group_evsel (pos, leader) { -if (!evsel__is_hybrid (pos)) -continue ; - -if (prev &&strcmp (prev->pmu_name, pos->pmu_name)) -return true; - - prev = pos; - } - -return false; -} - -void evlist__warn_hybrid_group (struct evlist *evlist) -{ -struct evsel *evsel; - -evlist__for_each_entry (evlist, evsel) { -if (evsel__is_group_leader (evsel) && - evsel->core.nr_members >1 && -group_hybrid_conflict (evsel)) { -pr_warning ("WARNING: events in group from " -"different hybrid PMUs!\n" ); -return ; - } - } -} -bool evlist__has_hybrid (struct evlist *evlist) {struct evsel *evsel;
tools/perf/util/evlist-hybrid.h +0 -1
@@ -8,7 +8,6 @@#include <unistd.h> int evlist__add_default_hybrid (struct evlist *evlist,bool precise); -void evlist__warn_hybrid_group (struct evlist *evlist);bool evlist__has_hybrid (struct evlist *evlist);#endif /* __PERF_EVLIST_HYBRID_H */
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 14/35 perf evlist: Remove __evlist__add_default
In reply to Ian Rogers__evlist__add_default adds a cycles event to a typically empty evlist and was extended for hybrid with evlist__add_default_hybrid, as more than 1 PMU was necessary. Rather than have dedicated logic for the cycles event, this change switches to parsing 'cycles:P' which will handle wildcarding the PMUs appropriately for hybrid.
--- tools/perf/arch/x86/util/evsel.c | 20 -------------- tools/perf/builtin-record.c | 13 +++------ tools/perf/builtin-top.c | 10 ++++--- tools/perf/util/evlist-hybrid.c | 25 ----------------- tools/perf/util/evlist-hybrid.h | 1 - tools/perf/util/evlist.c | 22 ++++++--------- tools/perf/util/evlist.h | 7 ----- tools/perf/util/evsel.c | 46 -------------------------------- tools/perf/util/evsel.h | 3 --- 9 files changed, 17 insertions(+), 130 deletions(-)
tools/perf/arch/x86/util/evsel.c +0 -20
@@ -16,26 +16,6 @@ void arch_evsel__set_sample_weight(struct evsel *evsel)evsel__set_sample_bit (evsel,WEIGHT_STRUCT ); } -void arch_evsel__fixup_new_cycles (struct perf_event_attr *attr) -{ -struct perf_env env = { .total_mem =0 , } ; - -if (!perf_env__cpuid (&env)) -return ; - -/* -* On AMD, precise cycles event sampling internally uses IBS pmu. -* But IBS does not have filtering capabilities and perf by default -* sets exclude_guest = 1. This makes IBS pmu event init fail and -* thus perf ends up doing non-precise sampling. Avoid it by clearing -* exclude_guest. -*/ -if (env.cpuid &&strstarts (env.cpuid,"AuthenticAMD" )) - attr->exclude_guest =0 ; - -free (env.cpuid); -} -/* Check whether the evsel's PMU supports the perf metrics */ bool evsel__sys_has_perf_metrics (const struct evsel *evsel) {
tools/perf/builtin-record.c +3 -10
@@ -4161,18 +4161,11 @@ int cmd_record(int argc, const char **argv) record.opts.tail_synthesize = true;if (rec->evlist->core.nr_entries ==0 ) { -if (perf_pmu__has_hybrid ()) { - err =evlist__add_default_hybrid (rec->evlist, - !record.opts.no_samples); - }else { - err =__evlist__add_default (rec->evlist, - !record.opts.no_samples); - } +bool can_profile_kernel =perf_event_paranoid_check (1 ); -if (err <0 ) { -pr_err ("Not enough memory for event selector list\n" ); + err =parse_event (rec->evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); +if (err) goto out; - } }if (rec->opts.target.tid && !rec->opts.no_inherit_set)
tools/perf/builtin-top.c +6 -4
@@ -1653,10 +1653,12 @@ int cmd_top(int argc, const char **argv)if (annotate_check_args (&top.annotation_opts) <0 ) goto out_delete_evlist; -if (!top.evlist->core.nr_entries && -evlist__add_default (top.evlist) <0 ) { -pr_err ("Not enough memory for event selector list\n" ); - goto out_delete_evlist; +if (!top.evlist->core.nr_entries) { +bool can_profile_kernel =perf_event_paranoid_check (1 ); +int err =parse_event (top.evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); + +if (err) + goto out_delete_evlist; } status =evswitch__init (&top.evswitch, top.evlist, stderr);
tools/perf/util/evlist-hybrid.c +0 -25
@@ -16,31 +16,6 @@#include <perf/evsel.h> #include <perf/cpumap.h> -int evlist__add_default_hybrid (struct evlist *evlist,bool precise) -{ -struct evsel *evsel; -struct perf_pmu *pmu; -__u64 config; -struct perf_cpu_map *cpus; - -perf_pmu__for_each_hybrid_pmu (pmu) { - config =PERF_COUNT_HW_CPU_CYCLES | - ((__u64 )pmu->type <<PERF_PMU_TYPE_SHIFT ); - evsel =evsel__new_cycles (precise,PERF_TYPE_HARDWARE , - config); -if (!evsel) -return -ENOMEM ; - - cpus =perf_cpu_map__get (pmu->cpus); - evsel->core.cpus = cpus; - evsel->core.own_cpus =perf_cpu_map__get (cpus); - evsel->pmu_name =strdup (pmu->name); -evlist__add (evlist, evsel); - } - -return 0 ; -} -bool evlist__has_hybrid (struct evlist *evlist) {struct evsel *evsel;
tools/perf/util/evlist-hybrid.h +0 -1
@@ -7,7 +7,6 @@#include "evlist.h" #include <unistd.h> -int evlist__add_default_hybrid (struct evlist *evlist,bool precise);bool evlist__has_hybrid (struct evlist *evlist);#endif /* __PERF_EVLIST_HYBRID_H */
tools/perf/util/evlist.c +8 -14
@@ -93,8 +93,15 @@ struct evlist *evlist__new(void)struct evlist *evlist__new_default (void ) {struct evlist *evlist =evlist__new (); +bool can_profile_kernel; +int err; + +if (!evlist) +return NULL ; -if (evlist &&evlist__add_default (evlist)) { + can_profile_kernel =perf_event_paranoid_check (1 ); + err =parse_event (evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); +if (err) {evlist__delete (evlist); evlist =NULL ; }
@@ -237,19 +244,6 @@ static void evlist__set_leader(struct evlist *evlist)perf_evlist__set_leader (&evlist->core); } -int __evlist__add_default (struct evlist *evlist,bool precise) -{ -struct evsel *evsel; - - evsel =evsel__new_cycles (precise,PERF_TYPE_HARDWARE , -PERF_COUNT_HW_CPU_CYCLES ); -if (evsel ==NULL ) -return -ENOMEM ; - -evlist__add (evlist, evsel); -return 0 ; -} -static struct evsel *evlist__dummy_event (struct evlist *evlist) {struct perf_event_attr attr = {
tools/perf/util/evlist.h +0 -7
@@ -100,13 +100,6 @@ void evlist__delete(struct evlist *evlist);void evlist__add (struct evlist *evlist,struct evsel *entry);void evlist__remove (struct evlist *evlist,struct evsel *evsel); -int __evlist__add_default (struct evlist *evlist,bool precise); - -static inline int evlist__add_default (struct evlist *evlist) -{ -return __evlist__add_default (evlist, true); -} -int evlist__add_attrs (struct evlist *evlist,struct perf_event_attr *attrs,size_t nr_attrs); int __evlist__add_default_attrs(struct evlist *evlist,
tools/perf/util/evsel.c +0 -46
58 lines not on this page · raw message
tools/perf/util/evsel.h +0 -3
@@ -243,8 +243,6 @@ static inline struct evsel *evsel__newtp(const char *sys, const char *name) }#endif -struct evsel *evsel__new_cycles (bool precise,__u32 type,__u64 config); -#ifdef HAVE_LIBTRACEEVENT struct tep_event *event_format__new (const char *sys,const char *name);#endif
@@ -312,7 +310,6 @@ void __evsel__reset_sample_bit(struct evsel *evsel, enum perf_event_sample_formavoid evsel__set_sample_id (struct evsel *evsel,bool use_sample_identifier);void arch_evsel__set_sample_weight (struct evsel *evsel); -void arch_evsel__fixup_new_cycles (struct perf_event_attr *attr);void arch__post_evsel_config (struct evsel *evsel,struct perf_event_attr *attr);int evsel__set_filter (struct evsel *evsel,const char *filter);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Arnaldo Carvalho de Melo
maintainerRe: [PATCH v4 14/35] perf evlist: Remove __evlist__add_default
Em Fri, May 26, 2023 at 02:53:49PM -0700, Ian Rogers escreveu:
Quoted 5 lines
__evlist__add_defaultadds a cycles event to a typically empty evlist and was extended for hybrid withevlist__add_default_hybrid, as more than 1 PMU was necessary. Rather than have dedicated logic for the cycles event, this change switches to parsing 'cycles:P' which will handle wildcarding the PMUs appropriately for hybrid.
I think I reported this earlier, but at this point 'perf test python' breaks, I fixed it in the tmp.perf-tools-next:
19: 'import perf' in python : FAILED!
⬢[acme@toolbox perf-tools-next]$ git log --oneline -1fe4f622c4fc7a02a (HEAD) perf evlist: Remove __evlist__add_default
⬢[acme@toolbox perf-tools-next]$
⬢[acme@toolbox perf-tools-next]$ perf test -v python
Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc
19: 'import perf' in python :
--- start ---
test child forked, pid 2976621
python usage test: "echo "import sys ; sys.path.append('/tmp/build/perf-tools-next/python'); import perf" | '/usr/bin/python3' " Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ImportError: /tmp/build/perf-tools-next/python/perf.cpython-310-x86_64-linux-gnu.so: undefined symbol: parse_event test child finished with -1
---- end ----
'import perf' in python: FAILED!
⬢[acme@toolbox perf-tools-next]$
Probably there will be a few more cases in the next patches, please check.
- Arnaldo
Quoted 287 lines
Signed-off-by Ian Rogers <[email protected]>Reviewed-by Kan Liang <[email protected]>--- tools/perf/arch/x86/util/evsel.c | 20 -------------- tools/perf/builtin-record.c | 13 +++------ tools/perf/builtin-top.c | 10 ++++--- tools/perf/util/evlist-hybrid.c | 25 ----------------- tools/perf/util/evlist-hybrid.h | 1 - tools/perf/util/evlist.c | 22 ++++++--------- tools/perf/util/evlist.h | 7 ----- tools/perf/util/evsel.c | 46 -------------------------------- tools/perf/util/evsel.h | 3 --- 9 files changed, 17 insertions(+), 130 deletions(-)tools/perf/arch/x86/util/evsel.c +0
-20@@ -16,26 +16,6 @@ void arch_evsel__set_sample_weight(struct evsel *evsel)evsel__set_sample_bit (evsel,WEIGHT_STRUCT ); } -void arch_evsel__fixup_new_cycles (struct perf_event_attr *attr) -{ -struct perf_env env = { .total_mem =0 , } ; - -if (!perf_env__cpuid (&env)) -return ; - -/* -* On AMD, precise cycles event sampling internally uses IBS pmu. -* But IBS does not have filtering capabilities and perf by default -* sets exclude_guest = 1. This makes IBS pmu event init fail and -* thus perf ends up doing non-precise sampling. Avoid it by clearing -* exclude_guest. -*/ -if (env.cpuid &&strstarts (env.cpuid,"AuthenticAMD" )) - attr->exclude_guest =0 ; - -free (env.cpuid); -} -/* Check whether the evsel's PMU supports the perf metrics */ bool evsel__sys_has_perf_metrics (const struct evsel *evsel) {tools/perf/builtin-record.c +3
-10@@ -4161,18 +4161,11 @@ int cmd_record(int argc, const char **argv) record.opts.tail_synthesize = true;if (rec->evlist->core.nr_entries ==0 ) { -if (perf_pmu__has_hybrid ()) { - err =evlist__add_default_hybrid (rec->evlist, - !record.opts.no_samples); - }else { - err =__evlist__add_default (rec->evlist, - !record.opts.no_samples); - } +bool can_profile_kernel =perf_event_paranoid_check (1 ); -if (err <0 ) { -pr_err ("Not enough memory for event selector list\n" ); + err =parse_event (rec->evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); +if (err) goto out; - } }if (rec->opts.target.tid && !rec->opts.no_inherit_set)tools/perf/builtin-top.c +6
-4@@ -1653,10 +1653,12 @@ int cmd_top(int argc, const char **argv)if (annotate_check_args (&top.annotation_opts) <0 ) goto out_delete_evlist; -if (!top.evlist->core.nr_entries && -evlist__add_default (top.evlist) <0 ) { -pr_err ("Not enough memory for event selector list\n" ); - goto out_delete_evlist; +if (!top.evlist->core.nr_entries) { +bool can_profile_kernel =perf_event_paranoid_check (1 ); +int err =parse_event (top.evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); + +if (err) + goto out_delete_evlist; } status =evswitch__init (&top.evswitch, top.evlist, stderr);tools/perf/util/evlist-hybrid.c +0
-25@@ -16,31 +16,6 @@#include <perf/evsel.h> #include <perf/cpumap.h> -int evlist__add_default_hybrid (struct evlist *evlist,bool precise) -{ -struct evsel *evsel; -struct perf_pmu *pmu; -__u64 config; -struct perf_cpu_map *cpus; - -perf_pmu__for_each_hybrid_pmu (pmu) { - config =PERF_COUNT_HW_CPU_CYCLES | - ((__u64 )pmu->type <<PERF_PMU_TYPE_SHIFT ); - evsel =evsel__new_cycles (precise,PERF_TYPE_HARDWARE , - config); -if (!evsel) -return -ENOMEM ; - - cpus =perf_cpu_map__get (pmu->cpus); - evsel->core.cpus = cpus; - evsel->core.own_cpus =perf_cpu_map__get (cpus); - evsel->pmu_name =strdup (pmu->name); -evlist__add (evlist, evsel); - } - -return 0 ; -} -bool evlist__has_hybrid (struct evlist *evlist) {struct evsel *evsel;tools/perf/util/evlist-hybrid.h +0
-1@@ -7,7 +7,6 @@#include "evlist.h" #include <unistd.h> -int evlist__add_default_hybrid (struct evlist *evlist,bool precise);bool evlist__has_hybrid (struct evlist *evlist);#endif /* __PERF_EVLIST_HYBRID_H */ tools/perf/util/evlist.c +8
-14@@ -93,8 +93,15 @@ struct evlist *evlist__new(void)struct evlist *evlist__new_default (void ) {struct evlist *evlist =evlist__new (); +bool can_profile_kernel; +int err; + +if (!evlist) +return NULL ; -if (evlist &&evlist__add_default (evlist)) { + can_profile_kernel =perf_event_paranoid_check (1 ); + err =parse_event (evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); +if (err) {evlist__delete (evlist); evlist =NULL ; }@@ -237,19 +244,6 @@ static void evlist__set_leader(struct evlist *evlist)perf_evlist__set_leader (&evlist->core); } -int __evlist__add_default (struct evlist *evlist,bool precise) -{ -struct evsel *evsel; - - evsel =evsel__new_cycles (precise,PERF_TYPE_HARDWARE , -PERF_COUNT_HW_CPU_CYCLES ); -if (evsel ==NULL ) -return -ENOMEM ; - -evlist__add (evlist, evsel); -return 0 ; -} -static struct evsel *evlist__dummy_event (struct evlist *evlist) {struct perf_event_attr attr = {tools/perf/util/evlist.h +0
-7@@ -100,13 +100,6 @@ void evlist__delete(struct evlist *evlist);void evlist__add (struct evlist *evlist,struct evsel *entry);void evlist__remove (struct evlist *evlist,struct evsel *evsel); -int __evlist__add_default (struct evlist *evlist,bool precise); - -static inline int evlist__add_default (struct evlist *evlist) -{ -return __evlist__add_default (evlist, true); -} -int evlist__add_attrs (struct evlist *evlist,struct perf_event_attr *attrs,size_t nr_attrs); int __evlist__add_default_attrs(struct evlist *evlist,tools/perf/util/evsel.c +0
-46@@ -316,48 +316,6 @@ struct evsel *evsel__new_idx(struct perf_event_attr *attr, int idx)return evsel; } -static bool perf_event_can_profile_kernel (void ) -{ -return perf_event_paranoid_check (1 ); -} - -struct evsel *evsel__new_cycles (bool precise__maybe_unused ,__u32 type,__u64 config) -{ -struct perf_event_attr attr = { - .type = type, - .config = config, - .exclude_kernel = !perf_event_can_profile_kernel (), - }; -struct evsel *evsel; - -event_attr_init (&attr); - -/* -* Now let the usual logic to set up the perf_event_attr defaults -* to kick in when we return and before perf_evsel__open() is called. -*/ - evsel =evsel__new (&attr); -if (evsel ==NULL ) - goto out; - -arch_evsel__fixup_new_cycles (&evsel->core.attr); - - evsel->precise_max = true; - -/* use asprintf() because free(evsel) assumes name is allocated */ -if (asprintf (&evsel->name,"cycles%s%s%.*s" , - (attr.precise_ip || attr.exclude_kernel) ?":" :"" , - attr.exclude_kernel ?"u" :"" , - attr.precise_ip ? attr.precise_ip +1 :0 ,"ppp" ) <0 ) - goto error_free; -out: -return evsel; -error_free: -evsel__delete (evsel); - evsel =NULL ; - goto out; -} -int copy_config_terms (struct list_head *dst,struct list_head *src) {struct evsel_config_term *pos, *tmp;@@ -1131,10 +1089,6 @@ void __weak arch_evsel__set_sample_weight(struct evsel *evsel)evsel__set_sample_bit (evsel,WEIGHT ); } -void __weak arch_evsel__fixup_new_cycles (struct perf_event_attr *attr__maybe_unused ) -{ -} -void __weak arch__post_evsel_config (struct evsel *evsel__maybe_unused ,struct perf_event_attr *attr__maybe_unused ) {tools/perf/util/evsel.h +0
-3@@ -243,8 +243,6 @@ static inline struct evsel *evsel__newtp(const char *sys, const char *name) }#endif -struct evsel *evsel__new_cycles (bool precise,__u32 type,__u64 config); -#ifdef HAVE_LIBTRACEEVENT struct tep_event *event_format__new (const char *sys,const char *name);#endif @@ -312,7 +310,6 @@ void __evsel__reset_sample_bit(struct evsel *evsel, enum perf_event_sample_formavoid evsel__set_sample_id (struct evsel *evsel,bool use_sample_identifier);void arch_evsel__set_sample_weight (struct evsel *evsel); -void arch_evsel__fixup_new_cycles (struct perf_event_attr *attr);void arch__post_evsel_config (struct evsel *evsel,struct perf_event_attr *attr);int evsel__set_filter (struct evsel *evsel,const char *filter);Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Signature · 1 line
- Arnaldo
Ian Rogers
reviewerRe: [PATCH v4 14/35] perf evlist: Remove __evlist__add_default
Quoted 29 lines · Arnaldo Carvalho de Melo
Em Fri, May 26, 2023 at 02:53:49PM -0700, Ian Rogers escreveu:
Quoted 5 lines
__evlist__add_defaultadds a cycles event to a typically empty evlist and was extended for hybrid withevlist__add_default_hybrid, as more than 1 PMU was necessary. Rather than have dedicated logic for the cycles event, this change switches to parsing 'cycles:P' which will handle wildcarding the PMUs appropriately for hybrid.I think I reported this earlier, but at this point 'perf test python' breaks, I fixed it in the tmp.perf-tools-next:
19: 'import perf' in python : FAILED!
⬢[acme@toolbox perf-tools-next]$ git log --oneline -1fe4f622c4fc7a02a(HEAD) perf evlist: Remove__evlist__add_default
⬢[acme@toolbox perf-tools-next]$
⬢[acme@toolbox perf-tools-next]$ perf test -v python
Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc
19: 'import perf' in python :
--- start ---
test child forked, pid 2976621
python usage test: "echo "import sys ;sys.path.append('/tmp/build/perf-tools-next/python'); import perf" | '/usr/bin/python3' " Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ImportError:/tmp/build/perf-tools-next/python/perf.cpython-310-x86_64-linux-gnu.so: undefined symbol:parse_eventtest child finished with -1
---- end ----
'import perf' in python: FAILED!
⬢[acme@toolbox perf-tools-next]$
Quoted 2 lines
Probably there will be a few more cases in the next patches, please check.
I'll rebase and resend. I needed to add:
https://lore.kernel.org/lkml/[email protected]/
to repro the failure. The test was passing without it.
Thanks,
Ian
Quoted 293 lines
- Arnaldo
Quoted 285 lines
Signed-off-by Ian Rogers <[email protected]>Reviewed-by Kan Liang <[email protected]>--- tools/perf/arch/x86/util/evsel.c | 20 -------------- tools/perf/builtin-record.c | 13 +++------ tools/perf/builtin-top.c | 10 ++++--- tools/perf/util/evlist-hybrid.c | 25 ----------------- tools/perf/util/evlist-hybrid.h | 1 - tools/perf/util/evlist.c | 22 ++++++--------- tools/perf/util/evlist.h | 7 ----- tools/perf/util/evsel.c | 46 -------------------------------- tools/perf/util/evsel.h | 3 --- 9 files changed, 17 insertions(+), 130 deletions(-)tools/perf/arch/x86/util/evsel.c +0
-20@@ -16,26 +16,6 @@ void arch_evsel__set_sample_weight(struct evsel *evsel)evsel__set_sample_bit (evsel,WEIGHT_STRUCT ); } -void arch_evsel__fixup_new_cycles (struct perf_event_attr *attr) -{ -struct perf_env env = { .total_mem =0 , } ; - -if (!perf_env__cpuid (&env)) -return ; - -/* -* On AMD, precise cycles event sampling internally uses IBS pmu. -* But IBS does not have filtering capabilities and perf by default -* sets exclude_guest = 1. This makes IBS pmu event init fail and -* thus perf ends up doing non-precise sampling. Avoid it by clearing -* exclude_guest. -*/ -if (env.cpuid &&strstarts (env.cpuid,"AuthenticAMD" )) - attr->exclude_guest =0 ; - -free (env.cpuid); -} -/* Check whether the evsel's PMU supports the perf metrics */ bool evsel__sys_has_perf_metrics (const struct evsel *evsel) {tools/perf/builtin-record.c +3
-10@@ -4161,18 +4161,11 @@ int cmd_record(int argc, const char **argv) record.opts.tail_synthesize = true;if (rec->evlist->core.nr_entries ==0 ) { -if (perf_pmu__has_hybrid ()) { - err =evlist__add_default_hybrid (rec->evlist, - !record.opts.no_samples); - }else { - err =__evlist__add_default (rec->evlist, - !record.opts.no_samples); - } +bool can_profile_kernel =perf_event_paranoid_check (1 ); -if (err <0 ) { -pr_err ("Not enough memory for event selector list\n" ); + err =parse_event (rec->evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); +if (err) goto out; - } }if (rec->opts.target.tid && !rec->opts.no_inherit_set)tools/perf/builtin-top.c +6
-4@@ -1653,10 +1653,12 @@ int cmd_top(int argc, const char **argv)if (annotate_check_args (&top.annotation_opts) <0 ) goto out_delete_evlist; -if (!top.evlist->core.nr_entries && -evlist__add_default (top.evlist) <0 ) { -pr_err ("Not enough memory for event selector list\n" ); - goto out_delete_evlist; +if (!top.evlist->core.nr_entries) { +bool can_profile_kernel =perf_event_paranoid_check (1 ); +int err =parse_event (top.evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); + +if (err) + goto out_delete_evlist; } status =evswitch__init (&top.evswitch, top.evlist, stderr);tools/perf/util/evlist-hybrid.c +0
-25@@ -16,31 +16,6 @@#include <perf/evsel.h> #include <perf/cpumap.h> -int evlist__add_default_hybrid (struct evlist *evlist,bool precise) -{ -struct evsel *evsel; -struct perf_pmu *pmu; -__u64 config; -struct perf_cpu_map *cpus; - -perf_pmu__for_each_hybrid_pmu (pmu) { - config =PERF_COUNT_HW_CPU_CYCLES | - ((__u64 )pmu->type <<PERF_PMU_TYPE_SHIFT ); - evsel =evsel__new_cycles (precise,PERF_TYPE_HARDWARE , - config); -if (!evsel) -return -ENOMEM ; - - cpus =perf_cpu_map__get (pmu->cpus); - evsel->core.cpus = cpus; - evsel->core.own_cpus =perf_cpu_map__get (cpus); - evsel->pmu_name =strdup (pmu->name); -evlist__add (evlist, evsel); - } - -return 0 ; -} -bool evlist__has_hybrid (struct evlist *evlist) {struct evsel *evsel;tools/perf/util/evlist-hybrid.h +0
-1@@ -7,7 +7,6 @@#include "evlist.h" #include <unistd.h> -int evlist__add_default_hybrid (struct evlist *evlist,bool precise);bool evlist__has_hybrid (struct evlist *evlist);#endif /* __PERF_EVLIST_HYBRID_H */ tools/perf/util/evlist.c +8
-14@@ -93,8 +93,15 @@ struct evlist *evlist__new(void)struct evlist *evlist__new_default (void ) {struct evlist *evlist =evlist__new (); +bool can_profile_kernel; +int err; + +if (!evlist) +return NULL ; -if (evlist &&evlist__add_default (evlist)) { + can_profile_kernel =perf_event_paranoid_check (1 ); + err =parse_event (evlist, can_profile_kernel ?"cycles:P" :"cycles:Pu" ); +if (err) {evlist__delete (evlist); evlist =NULL ; }@@ -237,19 +244,6 @@ static void evlist__set_leader(struct evlist *evlist)perf_evlist__set_leader (&evlist->core); } -int __evlist__add_default (struct evlist *evlist,bool precise) -{ -struct evsel *evsel; - - evsel =evsel__new_cycles (precise,PERF_TYPE_HARDWARE , -PERF_COUNT_HW_CPU_CYCLES ); -if (evsel ==NULL ) -return -ENOMEM ; - -evlist__add (evlist, evsel); -return 0 ; -} -static struct evsel *evlist__dummy_event (struct evlist *evlist) {struct perf_event_attr attr = {tools/perf/util/evlist.h +0
-7@@ -100,13 +100,6 @@ void evlist__delete(struct evlist *evlist);void evlist__add (struct evlist *evlist,struct evsel *entry);void evlist__remove (struct evlist *evlist,struct evsel *evsel); -int __evlist__add_default (struct evlist *evlist,bool precise); - -static inline int evlist__add_default (struct evlist *evlist) -{ -return __evlist__add_default (evlist, true); -} -int evlist__add_attrs (struct evlist *evlist,struct perf_event_attr *attrs,size_t nr_attrs); int __evlist__add_default_attrs(struct evlist *evlist,tools/perf/util/evsel.c +0
-46@@ -316,48 +316,6 @@ struct evsel *evsel__new_idx(struct perf_event_attr *attr, int idx)return evsel; } -static bool perf_event_can_profile_kernel (void ) -{ -return perf_event_paranoid_check (1 ); -} - -struct evsel *evsel__new_cycles (bool precise__maybe_unused ,__u32 type,__u64 config) -{ -struct perf_event_attr attr = { - .type = type, - .config = config, - .exclude_kernel = !perf_event_can_profile_kernel (), - }; -struct evsel *evsel; - -event_attr_init (&attr); - -/* -* Now let the usual logic to set up the perf_event_attr defaults -* to kick in when we return and before perf_evsel__open() is called. -*/ - evsel =evsel__new (&attr); -if (evsel ==NULL ) - goto out; - -arch_evsel__fixup_new_cycles (&evsel->core.attr); - - evsel->precise_max = true; - -/* use asprintf() because free(evsel) assumes name is allocated */ -if (asprintf (&evsel->name,"cycles%s%s%.*s" , - (attr.precise_ip || attr.exclude_kernel) ?":" :"" , - attr.exclude_kernel ?"u" :"" , - attr.precise_ip ? attr.precise_ip +1 :0 ,"ppp" ) <0 ) - goto error_free; -out: -return evsel; -error_free: -evsel__delete (evsel); - evsel =NULL ; - goto out; -} -int copy_config_terms (struct list_head *dst,struct list_head *src) {struct evsel_config_term *pos, *tmp;@@ -1131,10 +1089,6 @@ void __weak arch_evsel__set_sample_weight(struct evsel *evsel)evsel__set_sample_bit (evsel,WEIGHT ); } -void __weak arch_evsel__fixup_new_cycles (struct perf_event_attr *attr__maybe_unused ) -{ -} -void __weak arch__post_evsel_config (struct evsel *evsel__maybe_unused ,struct perf_event_attr *attr__maybe_unused ) {tools/perf/util/evsel.h +0
-4@@ -243,8 +243,6 @@ static inline struct evsel *evsel__newtp(const char *sys, const char *name) }#endif -struct evsel *evsel__new_cycles (bool precise,__u32 type,__u64 config); -#ifdef HAVE_LIBTRACEEVENT struct tep_event *event_format__new (const char *sys,const char *name);#endif @@ -312,7 +310,6 @@ void __evsel__reset_sample_bit(struct evsel *evsel, enum perf_event_sample_formavoid evsel__set_sample_id (struct evsel *evsel,bool use_sample_identifier);void arch_evsel__set_sample_weight (struct evsel *evsel); -void arch_evsel__fixup_new_cycles (struct perf_event_attr *attr);void arch__post_evsel_config (struct evsel *evsel,struct perf_event_attr *attr);int evsel__set_filter (struct evsel *evsel,const char *filter); --Quoted 2 lines
2.41.0.rc0.172.g3f132b7071-goog
--
- Arnaldo
Ian Rogers
reviewerPATCH v4 15/35 perf evlist: Reduce scope of evlist__has_hybrid
In reply to Ian RogersFunction is only used in printout, reduce scope to stat-display.c. Remove the now empty evlist-hybrid.c and evlist-hybrid.h.
--- tools/perf/builtin-record.c | 1 - tools/perf/builtin-stat.c | 1 - tools/perf/util/Build | 1 - tools/perf/util/evlist-hybrid.c | 31 ------------------------------- tools/perf/util/evlist-hybrid.h | 12 ------------ tools/perf/util/evlist.c | 1 - tools/perf/util/stat-display.c | 15 ++++++++++++++- 7 files changed, 14 insertions(+), 48 deletions(-) delete mode 100644 tools/perf/util/evlist-hybrid.c delete mode 100644 tools/perf/util/evlist-hybrid.h
tools/perf/builtin-record.c +0 -1
@@ -50,7 +50,6 @@#include "util/pfm.h" #include "util/clockid.h" #include "util/pmu-hybrid.h" -#include "util/evlist-hybrid.h" #include "util/off_cpu.h" #include "util/bpf-filter.h" #include "asm/bug.h"
tools/perf/builtin-stat.c +0 -1
@@ -48,7 +48,6 @@#include "util/pmu.h" #include "util/event.h" #include "util/evlist.h" -#include "util/evlist-hybrid.h" #include "util/evsel.h" #include "util/debug.h" #include "util/color.h"
tools/perf/util/Build +0 -1
@@ -11,7 +11,6 @@ perf-y += db-export.o perf-y += env.o perf-y += event.o perf-y += evlist.o -perf-y += evlist-hybrid.o perf-y += sideband_evlist.o perf-y += evsel.o perf-y += evsel_fprintf.o
tools/perf/util/evlist-hybrid.c deleted +0 -31
@@ -1,31 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0-only -#include <errno.h> -#include <inttypes.h> -#include "cpumap.h" -#include "evlist.h" -#include "evsel.h" -#include "../perf.h" -#include "util/pmu-hybrid.h" -#include "util/evlist-hybrid.h" -#include "debug.h" -#include <unistd.h> -#include <stdlib.h> -#include <linux/err.h> -#include <linux/string.h> -#include <perf/evlist.h> -#include <perf/evsel.h> -#include <perf/cpumap.h> - -bool evlist__has_hybrid (struct evlist *evlist) -{ -struct evsel *evsel; - -evlist__for_each_entry (evlist, evsel) { -if (evsel->pmu_name && -perf_pmu__is_hybrid (evsel->pmu_name)) { -return true; - } - } - -return false; -}
tools/perf/util/evlist-hybrid.h deleted +0 -12
@@ -1,12 +0,0 @@ -/* SPDX-License-Identifier: GPL-2.0 */ -#ifndef __PERF_EVLIST_HYBRID_H -#define __PERF_EVLIST_HYBRID_H - -#include <linux/compiler.h> -#include <linux/kernel.h> -#include "evlist.h" -#include <unistd.h> - -bool evlist__has_hybrid (struct evlist *evlist); - -#endif /* __PERF_EVLIST_HYBRID_H */
tools/perf/util/evlist.c +0 -1
@@ -28,7 +28,6 @@#include "util/string2.h" #include "util/perf_api_probe.h" #include "util/evsel_fprintf.h" -#include "util/evlist-hybrid.h" #include "util/pmu.h" #include "util/sample.h" #include "util/bpf-filter.h"
tools/perf/util/stat-display.c +14 -1
@@ -20,7 +20,6 @@#include "util.h" #include "iostat.h" #include "pmu-hybrid.h" -#include "evlist-hybrid.h" #define CNTR_NOT_SUPPORTED "<not supported>"#define CNTR_NOT_COUNTED "<not counted>"
@@ -677,6 +676,20 @@ static bool is_mixed_hw_group(struct evsel *counter)return false; } +static bool evlist__has_hybrid (struct evlist *evlist) +{ +struct evsel *evsel; + +evlist__for_each_entry (evlist, evsel) { +if (evsel->pmu_name && +perf_pmu__is_hybrid (evsel->pmu_name)) { +return true; + } + } + +return false; +} +static void printout (struct perf_stat_config *config,struct outstate *os,double uval,u64 run,u64 ena,double noise,int aggr_idx) {
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 16/35 perf pmu: Remove perf_pmu__hybrid_mounted
In reply to Ian Rogersperf_pmu__hybrid_mounted is used to detect whether cpu_core or cpu_atom is mounted with a non-empty cpus file by pmu_lookup. Discussion [1] showed the empty cpus file check to be redundant and so pmu_lookup needn't have a call to perf_pmu__hybrid_mounted.
Checking hybrid_mounted in pmu_is_uncore is redundant as the next cpumask read will fail returning false.
Reduce the scope of perf_pmu__find_hybrid_pmu by making it static.
[1] https://lore.kernel.org/lkml/[email protected]/
--- tools/perf/util/pmu-hybrid.c | 15 +-------------- tools/perf/util/pmu-hybrid.h | 3 --- tools/perf/util/pmu.c | 13 +------------ 3 files changed, 2 insertions(+), 29 deletions(-)
tools/perf/util/pmu-hybrid.c +1 -14
@@ -18,20 +18,7 @@LIST_HEAD (perf_pmu__hybrid_pmus); -bool perf_pmu__hybrid_mounted (const char *name) -{ -int cpu; -char pmu_name[PATH_MAX ]; -struct perf_pmu pmu = {.name = pmu_name}; - -if (strncmp (name,"cpu_" ,4 )) -return false; - -strlcpy (pmu_name, name,sizeof (pmu_name)); -return perf_pmu__scan_file (&pmu,"cpus" ,"%u" , &cpu) >0 ; -} - -struct perf_pmu *perf_pmu__find_hybrid_pmu (const char *name) +static struct perf_pmu *perf_pmu__find_hybrid_pmu (const char *name) {struct perf_pmu *pmu;
tools/perf/util/pmu-hybrid.h +0 -3
@@ -13,9 +13,6 @@ extern struct list_head perf_pmu__hybrid_pmus;#define perf_pmu__for_each_hybrid_pmu (pmu) \ list_for_each_entry(pmu, &perf_pmu__hybrid_pmus, hybrid_list) -bool perf_pmu__hybrid_mounted (const char *name); - -struct perf_pmu *perf_pmu__find_hybrid_pmu (const char *name);bool perf_pmu__is_hybrid (const char *name);static inline int perf_pmu__hybrid_pmu_num (void )
tools/perf/util/pmu.c +1 -12
@@ -617,9 +617,6 @@ static bool pmu_is_uncore(int dirfd, const char *name) {int fd; -if (perf_pmu__hybrid_mounted (name)) -return false; - fd =perf_pmu__pathname_fd (dirfd, name,"cpumask" ,O_PATH );if (fd <0 )return false;
@@ -907,15 +904,8 @@ static struct perf_pmu *pmu_lookup(int dirfd, const char *lookup_name)LIST_HEAD (aliases);__u32 type;char *name =pmu_find_real_name (lookup_name); -bool is_hybrid =perf_pmu__hybrid_mounted (name);char *alias_name; -/* -* Check pmu name for hybrid and the pmu may be invalid in sysfs -*/ -if (!strncmp (name,"cpu_" ,4 ) && !is_hybrid) -return NULL ; -/* * The pmu data we store & need consists of the pmu * type value and format definitions. Load both right
@@ -936,7 +926,6 @@ static struct perf_pmu *pmu_lookup(int dirfd, const char *lookup_name) pmu->cpus =pmu_cpumask (dirfd, name); pmu->name =strdup (name); -if (!pmu->name) goto err;
@@ -967,7 +956,7 @@ static struct perf_pmu *pmu_lookup(int dirfd, const char *lookup_name)list_splice (&aliases, &pmu->aliases);list_add_tail (&pmu->list, &pmus); -if (is_hybrid) +if (!strcmp (name,"cpu_core" ) || !strcmp (name,"cpu_atom" ))list_add_tail (&pmu->hybrid_list, &perf_pmu__hybrid_pmus);else INIT_LIST_HEAD (&pmu->hybrid_list);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 17/35 perf pmu: Rewrite perf_pmu__has_hybrid to avoid list
In reply to Ian RogersRather than list empty on perf_pmu__hybrid_pmus, detect if any core PMUs match the hybrid name. Computed values held in statics to avoid recomputation.
--- tools/perf/util/pmu.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-)
tools/perf/util/pmu.c +11 -5
@@ -60,8 +60,6 @@ struct perf_pmu_format {struct list_head list; }; -static bool hybrid_scanned; -static struct perf_pmu *perf_pmu__find2 (int dirfd,const char *name);/*
@@ -2013,12 +2011,20 @@ void perf_pmu__warn_invalid_config(struct perf_pmu *pmu, __u64 config,bool perf_pmu__has_hybrid (void ) { +static bool hybrid_scanned, has_hybrid; +if (!hybrid_scanned) { +struct perf_pmu *pmu =NULL ; + +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +if (pmu->is_core &&is_pmu_hybrid (pmu->name)) { + has_hybrid = true; +break ; + } + } hybrid_scanned = true; -perf_pmu__scan (NULL ); } - -return !list_empty (&perf_pmu__hybrid_pmus); +return has_hybrid; }int perf_pmu__match (char *pattern,char *name,char *tok)
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 18/35 perf x86: Iterate hybrid PMUs as core PMUs
In reply to Ian RogersRather than iterating over a separate hybrid list, iterate all PMUs with the hybrid ones having is_core as true.
--- tools/perf/arch/x86/tests/hybrid.c | 2 +- tools/perf/arch/x86/util/evlist.c | 25 +++++++++++++++++-------- tools/perf/arch/x86/util/perf_regs.c | 14 ++++++++++---- 3 files changed, 28 insertions(+), 13 deletions(-)
tools/perf/arch/x86/tests/hybrid.c +1 -1
@@ -3,7 +3,7 @@#include "debug.h" #include "evlist.h" #include "evsel.h" -#include "pmu-hybrid.h" +#include "pmu.h" #include "tests/tests.h" static bool test_config (const struct evsel *evsel,__u64 expected_config)
tools/perf/arch/x86/util/evlist.c +17 -8
@@ -4,7 +4,6 @@#include "util/evlist.h" #include "util/parse-events.h" #include "util/event.h" -#include "util/pmu-hybrid.h" #include "topdown.h" #include "evsel.h"
@@ -12,9 +11,6 @@ static int ___evlist__add_default_attrs(struct evlist *evlist,struct perf_event_attr *attrs, size_t nr_attrs) { -struct perf_cpu_map *cpus; -struct evsel *evsel, *n; -struct perf_pmu *pmu;LIST_HEAD (head );size_t i =0 ;
@@ -25,15 +21,24 @@ static int ___evlist__add_default_attrs(struct evlist *evlist,return evlist__add_attrs (evlist, attrs, nr_attrs);for (i =0 ; i < nr_attrs; i++) { +struct perf_pmu *pmu =NULL ; +if (attrs[i].type ==PERF_TYPE_SOFTWARE ) { - evsel =evsel__new (attrs + i); +struct evsel *evsel =evsel__new (attrs + i); +if (evsel ==NULL ) goto out_delete_partial_list;list_add_tail (&evsel->core.node, &head);continue ; } -perf_pmu__for_each_hybrid_pmu (pmu) { +while ((pmu =perf_pmu__scan (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;
@@ -51,8 +56,12 @@ static int ___evlist__add_default_attrs(struct evlist *evlist,return 0 ; out_delete_partial_list: -__evlist__for_each_entry_safe (&head, n, evsel) -evsel__delete (evsel); + { +struct evsel *evsel, *n; + +__evlist__for_each_entry_safe (&head, n, evsel) +evsel__delete (evsel); + }return -1 ; }
tools/perf/arch/x86/util/perf_regs.c +10 -4
@@ -10,7 +10,6 @@#include "../../../util/debug.h" #include "../../../util/event.h" #include "../../../util/pmu.h" -#include "../../../util/pmu-hybrid.h" const struct sample_reg sample_reg_masks[] = {SMPL_REG (AX ,PERF_REG_X86_AX ),
@@ -286,7 +285,6 @@ uint64_t arch__intr_reg_mask(void) .disabled = 1, .exclude_kernel = 1, }; -struct perf_pmu *pmu;int fd;/* * In an unnamed union, init it here to build on older gcc versions
@@ -294,12 +292,20 @@ uint64_t arch__intr_reg_mask(void) attr.sample_period =1 ;if (perf_pmu__has_hybrid ()) { +struct perf_pmu *pmu =NULL ; +__u64 type =PERF_TYPE_RAW ; +/* * The same register set is supported among different hybrid PMUs. * Only check the first available one. */ - pmu =list_first_entry (&perf_pmu__hybrid_pmus,typeof (*pmu), hybrid_list); - attr.config |= (__u64 )pmu->type <<PERF_PMU_TYPE_SHIFT ; +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +if (pmu->is_core) { + type = pmu->type; +break ; + } + } + attr.config |= type <<PERF_PMU_TYPE_SHIFT ; }event_attr_init (&attr);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 19/35 perf topology: Avoid hybrid list for hybrid topology
In reply to Ian RogersAvoid perf_pmu__for_each_hybrid_pmu in hybrid_topology__new by scanning all PMUs and processing the is_core ones. Add early exit for non-hybrid.
--- tools/perf/util/cputopo.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-)
tools/perf/util/cputopo.c +14 -5
@@ -12,7 +12,7 @@#include "cpumap.h" #include "debug.h" #include "env.h" -#include "pmu-hybrid.h" +#include "pmu.h" #define PACKAGE_CPUS_FMT \ "%s/devices/system/cpu/cpu%d/topology/package_cpus_list"
@@ -469,11 +469,17 @@ static int load_hybrid_node(struct hybrid_topology_node *node,struct hybrid_topology *hybrid_topology__new (void ) { -struct perf_pmu *pmu; +struct perf_pmu *pmu =NULL ;struct hybrid_topology *tp =NULL ; -u32 nr, i =0 ; +u32 nr =0 , i =0 ; - nr =perf_pmu__hybrid_pmu_num (); +if (!perf_pmu__has_hybrid ()) +return NULL ; + +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +if (pmu->is_core) + nr++; + }if (nr ==0 )return NULL ;
@@ -482,7 +488,10 @@ struct hybrid_topology *hybrid_topology__new(void)return NULL ; tp->nr = nr; -perf_pmu__for_each_hybrid_pmu (pmu) { +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +if (!pmu->is_core) +continue ; +if (load_hybrid_node (&tp->nodes[i], pmu)) {hybrid_topology__delete (tp);return NULL ;
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 20/35 perf evsel: Compute is_hybrid from PMU being core
In reply to Ian RogersShort-cut when has_hybrid is false, otherwise return if the evsel's PMU is core. Add a comment for the some what surprising no PMU cases of hardware and legacy cache events.
--- tools/perf/util/evsel.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +10 -2
@@ -46,8 +46,8 @@#include "memswap.h" #include "util.h" #include "util/hashmap.h" -#include "pmu-hybrid.h" #include "off_cpu.h" +#include "pmu.h" #include "../perf-sys.h" #include "util/parse-branch-options.h" #include "util/bpf-filter.h"
@@ -3132,9 +3132,17 @@ void evsel__zero_per_pkg(struct evsel *evsel) } } +/** +* evsel__is_hybrid - does the evsel have a known PMU that is hybrid. Note, this +* will be false on hybrid systems for hardware and legacy +* cache events. +*/ bool evsel__is_hybrid (const struct evsel *evsel) { -return evsel->pmu_name &&perf_pmu__is_hybrid (evsel->pmu_name); +if (!perf_pmu__has_hybrid ()) +return false; + +return evsel->core.is_pmu_core; }struct evsel *evsel__leader (const struct evsel *evsel)
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 21/35 perf header: Avoid hybrid PMU list in write_pmu_caps
In reply to Ian RogersAvoid perf_pmu__for_each_hybrid_pmu by iterating all PMUs are dumping the core ones. This will eventually allow removal of the hybrid PMU list.
--- tools/perf/util/header.c | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-)
tools/perf/util/header.c +11 -8
@@ -51,7 +51,6 @@#include "bpf-event.h" #include "bpf-utils.h" #include "clockid.h" -#include "pmu-hybrid.h" #include <linux/ctype.h> #include <internal/lib.h>
@@ -1589,17 +1588,21 @@ static int write_pmu_caps(struct feat_fd *ff,* Write hybrid pmu caps first to maintain compatibility with * older perf tool. */ - pmu =NULL ; -perf_pmu__for_each_hybrid_pmu (pmu) { - ret =__write_pmu_caps (ff, pmu, true); -if (ret <0 ) -return ret; +if (perf_pmu__has_hybrid ()) { + pmu =NULL ; +while ((pmu =perf_pmu__scan (pmu))) { +if (!pmu->is_core) +continue ; + + ret =__write_pmu_caps (ff, pmu, true); +if (ret <0 ) +return ret; + } } pmu =NULL ;while ((pmu =perf_pmu__scan (pmu))) { -if (!pmu->name || !strcmp (pmu->name,"cpu" ) || - !pmu->nr_caps ||perf_pmu__is_hybrid (pmu->name)) +if (pmu->is_core || !pmu->nr_caps)continue ; ret =__write_pmu_caps (ff, pmu, true);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 22/35 perf metrics: Remove perf_pmu__is_hybrid use
In reply to Ian RogersSwitch from perf_pmu__is_hybrid to avoid implicitly using the hybrid PMU list.
--- tools/perf/util/metricgroup.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-)
tools/perf/util/metricgroup.c +2 -4
@@ -11,7 +11,6 @@#include "evsel.h" #include "strbuf.h" #include "pmu.h" -#include "pmu-hybrid.h" #include "print-events.h" #include "smt.h" #include "expr.h"
@@ -274,7 +273,7 @@ static int setup_metric_events(const char *pmu, struct hashmap *ids,const char *metric_id;struct evsel *ev;size_t ids_size, matched_events, i; -bool all_pmus = !strcmp (pmu,"all" ) || !perf_pmu__is_hybrid (pmu); +bool all_pmus = !strcmp (pmu,"all" ) || !perf_pmu__has_hybrid () || !is_pmu_hybrid (pmu); *out_metric_events =NULL ; ids_size =hashmap__size (ids);
@@ -288,8 +287,7 @@ static int setup_metric_events(const char *pmu, struct hashmap *ids,struct expr_id_data *val_ptr;/* Don't match events for the wrong hybrid PMU. */ -if (!all_pmus && ev->pmu_name && -perf_pmu__is_hybrid (ev->pmu_name) && +if (!all_pmus && ev->pmu_name &&evsel__is_hybrid (ev) &&strcmp (ev->pmu_name, pmu))continue ;/*
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 23/35 perf stat: Avoid hybrid PMU list
In reply to Ian Rogersperf_pmu__is_hybrid implicitly uses the hybrid PMU list. Instead return false if hybrid isn't present, if it is then see if any evsel's PMUs are core.
--- tools/perf/util/stat-display.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
tools/perf/util/stat-display.c +5 -4
@@ -19,7 +19,7 @@#include <api/fs/fs.h> #include "util.h" #include "iostat.h" -#include "pmu-hybrid.h" +#include "pmu.h" #define CNTR_NOT_SUPPORTED "<not supported>"#define CNTR_NOT_COUNTED "<not counted>"
@@ -680,11 +680,12 @@ static bool evlist__has_hybrid(struct evlist *evlist) {struct evsel *evsel; +if (!perf_pmu__has_hybrid ()) +return false; +evlist__for_each_entry (evlist, evsel) { -if (evsel->pmu_name && -perf_pmu__is_hybrid (evsel->pmu_name)) { +if (evsel->core.is_pmu_core)return true; - } }return false;
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 24/35 perf mem: Avoid hybrid PMU list
In reply to Ian RogersAdd perf_pmu__num_mem_pmus that scans/counts the number of PMUs for mem events. Switch perf_pmu__for_each_hybrid_pmu to iterating all PMUs and only handling is_core ones.
--- tools/perf/builtin-c2c.c | 7 ++----- tools/perf/builtin-mem.c | 7 ++----- tools/perf/util/mem-events.c | 20 ++++++++++++++------ tools/perf/util/pmu.c | 17 +++++++++++++++++ tools/perf/util/pmu.h | 1 + 5 files changed, 36 insertions(+), 16 deletions(-)
tools/perf/builtin-c2c.c +2 -5
@@ -42,7 +42,6 @@#include "ui/ui.h" #include "ui/progress.h" #include "pmu.h" -#include "pmu-hybrid.h" #include "string2.h" #include "util/util.h"
@@ -3259,10 +3258,8 @@ static int perf_c2c__record(int argc, const char **argv) argc =parse_options (argc, argv, options, record_mem_usage,PARSE_OPT_KEEP_UNKNOWN ); -if (!perf_pmu__has_hybrid ()) - rec_argc = argc +11 ;/* max number of arguments */ -else - rec_argc = argc +11 *perf_pmu__hybrid_pmu_num (); +/* Max number of arguments multiplied by number of PMUs that can support them. */ + rec_argc = argc +11 *perf_pmu__num_mem_pmus (); rec_argv =calloc (rec_argc +1 ,sizeof (char *));if (!rec_argv)
tools/perf/builtin-mem.c +2 -5
@@ -18,7 +18,6 @@#include "util/map.h" #include "util/symbol.h" #include "util/pmu.h" -#include "util/pmu-hybrid.h" #include "util/sample.h" #include "util/string2.h" #include "util/util.h"
@@ -93,10 +92,8 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem) argc =parse_options (argc, argv, options, record_mem_usage,PARSE_OPT_KEEP_UNKNOWN ); -if (!perf_pmu__has_hybrid ()) - rec_argc = argc +9 ;/* max number of arguments */ -else - rec_argc = argc +9 *perf_pmu__hybrid_pmu_num (); +/* Max number of arguments multiplied by number of PMUs that can support them. */ + rec_argc = argc +9 *perf_pmu__num_mem_pmus ();if (mem->cpu_list) rec_argc +=2 ;
tools/perf/util/mem-events.c +14 -6
@@ -13,7 +13,6 @@#include "debug.h" #include "symbol.h" #include "pmu.h" -#include "pmu-hybrid.h" unsigned int perf_mem_events__loads_ldlat =30 ;
@@ -120,7 +119,6 @@ int perf_mem_events__init(void)for (j =0 ; j <PERF_MEM_EVENTS__MAX ; j++) {struct perf_mem_event *e =perf_mem_events__ptr (j); -struct perf_pmu *pmu;char sysfs_name[100 ];/*
@@ -135,7 +133,12 @@ int perf_mem_events__init(void) e->sysfs_name, "cpu"); e->supported =perf_mem_event__supported (mnt, sysfs_name); }else { -perf_pmu__for_each_hybrid_pmu (pmu) { +struct perf_pmu *pmu =NULL ; + +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +if (!pmu->is_core) +continue ; +scnprintf (sysfs_name,sizeof (sysfs_name), e->sysfs_name, pmu->name); e->supported |=perf_mem_event__supported (mnt, sysfs_name);
@@ -170,9 +173,12 @@ static void perf_mem_events__print_unsupport_hybrid(struct perf_mem_event *e, {const char *mnt =sysfs__mount ();char sysfs_name[100 ]; -struct perf_pmu *pmu; +struct perf_pmu *pmu =NULL ; + +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +if (!pmu->is_core) +continue ; -perf_pmu__for_each_hybrid_pmu (pmu) {scnprintf (sysfs_name,sizeof (sysfs_name), e->sysfs_name, pmu->name);if (!perf_mem_event__supported (mnt, sysfs_name)) {
@@ -210,7 +216,9 @@ int perf_mem_events__record_args(const char **rec_argv, int *argv_nr,return -1 ; } -perf_pmu__for_each_hybrid_pmu (pmu) { +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +if (!pmu->is_core) +continue ; rec_argv[i++] ="-e" ; s =perf_mem_events__name (j, pmu->name);if (s) {
tools/perf/util/pmu.c +17 -0
@@ -1660,6 +1660,23 @@ bool perf_pmu__auto_merge_stats(const struct perf_pmu *pmu)return !is_pmu_hybrid (pmu->name); } +static bool perf_pmu__is_mem_pmu (const struct perf_pmu *pmu) +{ +return pmu->is_core; +} + +int perf_pmu__num_mem_pmus (void ) +{ +struct perf_pmu *pmu =NULL ; +int count =0 ; + +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +if (perf_pmu__is_mem_pmu (pmu)) + count++; + } +return count; +} +static bool pmu_alias_is_duplicate (struct sevent *alias_a,struct sevent *alias_b) {
tools/perf/util/pmu.h +1 -0
@@ -231,6 +231,7 @@ bool is_pmu_hybrid(const char *name);bool perf_pmu__supports_legacy_cache (const struct perf_pmu *pmu);bool perf_pmu__supports_wildcard_numeric (const struct perf_pmu *pmu);bool perf_pmu__auto_merge_stats (const struct perf_pmu *pmu); +int perf_pmu__num_mem_pmus (void );void print_pmu_events (const struct print_callbacks *print_cb,void *print_state);bool pmu_have_event (const char *pname,const char *name);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 25/35 perf pmu: Remove perf_pmu__hybrid_pmus list
In reply to Ian RogersRather than iterate hybrid PMUs, inhererently Intel specific, iterate all PMUs checking whether they are core. To only get hybrid cores, first call perf_pmu__has_hybrid.
--- tools/perf/builtin-record.c | 1 - tools/perf/util/Build | 1 - tools/perf/util/pmu-hybrid.c | 39 ------------------------------ tools/perf/util/pmu-hybrid.h | 29 ---------------------- tools/perf/util/pmu.c | 7 ------ tools/perf/util/pmu.h | 2 -- tools/perf/util/print-events.c | 1 - tools/perf/util/python-ext-sources | 1 - 8 files changed, 81 deletions(-) delete mode 100644 tools/perf/util/pmu-hybrid.c delete mode 100644 tools/perf/util/pmu-hybrid.h
tools/perf/builtin-record.c +0 -1
@@ -49,7 +49,6 @@#include "util/util.h" #include "util/pfm.h" #include "util/clockid.h" -#include "util/pmu-hybrid.h" #include "util/off_cpu.h" #include "util/bpf-filter.h" #include "asm/bug.h"
tools/perf/util/Build +0 -1
@@ -73,7 +73,6 @@ perf-y += pmu.o perf-y += pmus.o perf-y += pmu-flex.o perf-y += pmu-bison.o -perf-y += pmu-hybrid.o perf-y += svghelper.o perf-$(CONFIG_LIBTRACEEVENT) += trace-event-info.o perf-y += trace-event-scripting.o
tools/perf/util/pmu-hybrid.c deleted +0 -39
@@ -1,39 +0,0 @@ -// SPDX-License-Identifier: GPL-2.0 -#include <linux/list.h> -#include <linux/compiler.h> -#include <linux/string.h> -#include <linux/zalloc.h> -#include <sys/types.h> -#include <errno.h> -#include <fcntl.h> -#include <sys/stat.h> -#include <unistd.h> -#include <stdio.h> -#include <stdbool.h> -#include <stdarg.h> -#include <locale.h> -#include <api/fs/fs.h> -#include "fncache.h" -#include "pmu-hybrid.h" - -LIST_HEAD (perf_pmu__hybrid_pmus); - -static struct perf_pmu *perf_pmu__find_hybrid_pmu (const char *name) -{ -struct perf_pmu *pmu; - -if (!name) -return NULL ; - -perf_pmu__for_each_hybrid_pmu (pmu) { -if (!strcmp (name, pmu->name)) -return pmu; - } - -return NULL ; -} - -bool perf_pmu__is_hybrid (const char *name) -{ -return perf_pmu__find_hybrid_pmu (name) !=NULL ; -}
tools/perf/util/pmu-hybrid.h deleted +0 -29
@@ -1,29 +0,0 @@ -/* SPDX-License-Identifier: GPL-2.0 */ -#ifndef __PMU_HYBRID_H -#define __PMU_HYBRID_H - -#include <linux/perf_event.h> -#include <linux/compiler.h> -#include <linux/list.h> -#include <stdbool.h> -#include "pmu.h" - -extern struct list_head perf_pmu__hybrid_pmus; - -#define perf_pmu__for_each_hybrid_pmu (pmu) \ - list_for_each_entry(pmu, &perf_pmu__hybrid_pmus, hybrid_list) - -bool perf_pmu__is_hybrid (const char *name); - -static inline int perf_pmu__hybrid_pmu_num (void ) -{ -struct perf_pmu *pmu; -int num =0 ; - -perf_pmu__for_each_hybrid_pmu (pmu) - num++; - -return num; -} - -#endif /* __PMU_HYBRID_H */
tools/perf/util/pmu.c +0 -7
@@ -32,7 +32,6 @@#include "string2.h" #include "strbuf.h" #include "fncache.h" -#include "pmu-hybrid.h" #include "util/evsel_config.h" struct perf_pmu perf_pmu__fake;
@@ -954,11 +953,6 @@ static struct perf_pmu *pmu_lookup(int dirfd, const char *lookup_name)list_splice (&aliases, &pmu->aliases);list_add_tail (&pmu->list, &pmus); -if (!strcmp (name,"cpu_core" ) || !strcmp (name,"cpu_atom" )) -list_add_tail (&pmu->hybrid_list, &perf_pmu__hybrid_pmus); -else -INIT_LIST_HEAD (&pmu->hybrid_list); - pmu->default_config =perf_pmu__get_default_config (pmu);return pmu;
@@ -2131,7 +2125,6 @@ void perf_pmu__destroy(void)list_for_each_entry_safe (pmu, tmp, &pmus, list) {list_del (&pmu->list); -list_del (&pmu->hybrid_list);perf_pmu__delete (pmu); }
tools/perf/util/pmu.h +0 -2
@@ -119,8 +119,6 @@ struct perf_pmu {struct list_head caps;/** @list: Element on pmus list in pmu.c. */ struct list_head list; -/** @hybrid_list: Element on perf_pmu__hybrid_pmus. */ -struct list_head hybrid_list;/** * @missing_features: Features to inhibit when events on this PMU are
tools/perf/util/print-events.c +0 -1
@@ -26,7 +26,6 @@#include "strlist.h" #include "tracepoint.h" #include "pfm.h" -#include "pmu-hybrid.h" #include "thread_map.h" #define MAX_NAME_LEN 100
tools/perf/util/python-ext-sources +0 -1
@@ -39,5 +39,4 @@ util/affinity.c util/rwsem.c util/hashmap.c util/perf_regs.c -util/pmu-hybrid.c util/fncache.c
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 26/35 perf pmus: Prefer perf_pmu__scan over perf_pmus__for_each_pmu
In reply to Ian Rogersperf_pmus__for_each_pmu doesn't lazily initialize pmus making its use error prone. Just use perf_pmu__scan as this only impacts non-performance critical tests.
--- tools/perf/bench/pmu-scan.c | 6 ++---- tools/perf/tests/event_groups.c | 7 ++----- tools/perf/tests/parse-events.c | 11 ++++------- tools/perf/util/pmus.h | 2 -- 4 files changed, 8 insertions(+), 18 deletions(-)
tools/perf/bench/pmu-scan.c +2 -4
@@ -40,13 +40,11 @@ static struct pmu_scan_result *results;static int save_result (void ) { -struct perf_pmu *pmu; +struct perf_pmu *pmu =NULL ;struct list_head *list;struct pmu_scan_result *r; -perf_pmu__scan (NULL ); - -perf_pmus__for_each_pmu (pmu) { +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { r =realloc (results, (nr_pmus +1 ) *sizeof (*r));if (r ==NULL )return -ENOMEM ;
tools/perf/tests/event_groups.c +2 -5
@@ -50,13 +50,10 @@ static int event_open(int type, unsigned long config, int group_fd)static int setup_uncore_event (void ) { -struct perf_pmu *pmu; +struct perf_pmu *pmu =NULL ;int i, fd; -if (list_empty (&pmus)) -perf_pmu__scan (NULL ); - -perf_pmus__for_each_pmu (pmu) { +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) {for (i =0 ; i <NR_UNCORE_PMUS ; i++) {if (!strcmp (uncore_pmus[i].name, pmu->name)) {pr_debug ("Using %s for uncore pmu event\n" , pmu->name);
tools/perf/tests/parse-events.c +4 -7
@@ -108,11 +108,11 @@ static int test__checkevent_raw(struct evlist *evlist)TEST_ASSERT_VAL ("wrong number of entries" ,0 != evlist->core.nr_entries);perf_evlist__for_each_evsel (&evlist->core, evsel) { -struct perf_pmu *pmu; +struct perf_pmu *pmu =NULL ;bool type_matched = false;TEST_ASSERT_VAL ("wrong config" ,test_perf_config (evsel,0x1a )); -perf_pmus__for_each_pmu (pmu) { +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) {if (pmu->type == evsel->attr.type) {TEST_ASSERT_VAL ("PMU type expected once" , !type_matched); type_matched = true;
@@ -2243,13 +2243,10 @@ static int test__terms2(struct test_suite *test __maybe_unused, int subtest __mastatic int test__pmu_events (struct test_suite *test__maybe_unused ,int subtest__maybe_unused ) { -struct perf_pmu *pmu; +struct perf_pmu *pmu =NULL ;int ret =TEST_OK ; -if (list_empty (&pmus)) -perf_pmu__scan (NULL ); - -perf_pmus__for_each_pmu (pmu) { +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) {struct stat st;char path[PATH_MAX ];struct dirent *ent;
tools/perf/util/pmus.h +0 -2
@@ -5,8 +5,6 @@extern struct list_head pmus;struct perf_pmu ; -#define perf_pmus__for_each_pmu (pmu) list_for_each_entry(pmu, &pmus, list) -const struct perf_pmu *perf_pmus__pmu_for_pmu_filter (const char *str);#endif /* __PMUS_H */
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 27/35 perf x86 mem: minor refactor to is_mem_loads_aux_event
In reply to Ian RogersFind the PMU and then the event off of it.
--- tools/perf/arch/x86/util/mem-events.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
tools/perf/arch/x86/util/mem-events.c +7 -7
@@ -55,13 +55,13 @@ struct perf_mem_event *perf_mem_events__ptr(int i)bool is_mem_loads_aux_event (struct evsel *leader) { -if (perf_pmu__find ("cpu" )) { -if (!pmu_have_event ("cpu" ,"mem-loads-aux" )) -return false; - }else if (perf_pmu__find ("cpu_core" )) { -if (!pmu_have_event ("cpu_core" ,"mem-loads-aux" )) -return false; - } +struct perf_pmu *pmu =perf_pmu__find ("cpu" ); + +if (!pmu) + pmu =perf_pmu__find ("cpu_core" ); + +if (pmu && !pmu_have_event (pmu->name,"mem-loads-aux" )) +return false;return leader->core.attr.config ==MEM_LOADS_AUX ; }
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 28/35 perf pmu: Separate pmu and pmus
In reply to Ian RogersSeparate and hide the pmus list in pmus.[ch]. Move pmus functionality out of pmu.[ch] into pmus.[ch] renaming pmus functions which were prefixed perf_pmu__ to perf_pmus__.
--- tools/perf/arch/arm/util/auxtrace.c | 7 +- tools/perf/arch/arm/util/cs-etm.c | 4 +- tools/perf/arch/arm64/util/pmu.c | 3 +- tools/perf/arch/x86/tests/hybrid.c | 5 +- tools/perf/arch/x86/util/auxtrace.c | 5 +- tools/perf/arch/x86/util/evlist.c | 5 +- tools/perf/arch/x86/util/evsel.c | 7 +- tools/perf/arch/x86/util/intel-bts.c | 4 +- tools/perf/arch/x86/util/intel-pt.c | 4 +- tools/perf/arch/x86/util/mem-events.c | 9 +- tools/perf/arch/x86/util/perf_regs.c | 5 +- tools/perf/arch/x86/util/topdown.c | 5 +- tools/perf/bench/pmu-scan.c | 10 +- tools/perf/builtin-c2c.c | 4 +- tools/perf/builtin-list.c | 4 +- tools/perf/builtin-mem.c | 4 +- tools/perf/builtin-record.c | 6 +- tools/perf/builtin-stat.c | 4 +- tools/perf/tests/attr.c | 4 +- tools/perf/tests/event_groups.c | 2 +- tools/perf/tests/parse-events.c | 8 +- tools/perf/tests/parse-metric.c | 4 +- tools/perf/tests/pmu-events.c | 3 +- tools/perf/tests/switch-tracking.c | 4 +- tools/perf/tests/topology.c | 4 +- tools/perf/util/cputopo.c | 7 +- tools/perf/util/env.c | 5 +- tools/perf/util/evsel.c | 3 +- tools/perf/util/header.c | 15 +- tools/perf/util/mem-events.c | 11 +- tools/perf/util/metricgroup.c | 5 +- tools/perf/util/parse-events.c | 15 +- tools/perf/util/parse-events.y | 3 +- tools/perf/util/pfm.c | 6 +- tools/perf/util/pmu.c | 411 +------------------------- tools/perf/util/pmu.h | 13 +- tools/perf/util/pmus.c | 396 ++++++++++++++++++++++++- tools/perf/util/pmus.h | 14 +- tools/perf/util/print-events.c | 5 +- tools/perf/util/stat-display.c | 3 +- 40 files changed, 531 insertions(+), 505 deletions(-)
tools/perf/arch/arm/util/auxtrace.c +4 -3
@@ -14,6 +14,7 @@#include "../../../util/debug.h" #include "../../../util/evlist.h" #include "../../../util/pmu.h" +#include "../../../util/pmus.h" #include "cs-etm.h" #include "arm-spe.h" #include "hisi-ptt.h"
@@ -40,7 +41,7 @@ static struct perf_pmu **find_all_arm_spe_pmus(int *nr_spes, int *err)return NULL ; } - arm_spe_pmus[*nr_spes] =perf_pmu__find (arm_spe_pmu_name); + arm_spe_pmus[*nr_spes] =perf_pmus__find (arm_spe_pmu_name);if (arm_spe_pmus[*nr_spes]) { pr_debug2("%s %d: arm_spe_pmu %d type %d name %s\n", __func__, __LINE__, *nr_spes,
@@ -87,7 +88,7 @@ static struct perf_pmu **find_all_hisi_ptt_pmus(int *nr_ptts, int *err)rewinddir (dir);while ((dent =readdir (dir))) {if (strstr (dent->d_name,HISI_PTT_PMU_NAME ) && idx < *nr_ptts) { - hisi_ptt_pmus[idx] =perf_pmu__find (dent->d_name); + hisi_ptt_pmus[idx] =perf_pmus__find (dent->d_name);if (hisi_ptt_pmus[idx]) idx++; }
@@ -131,7 +132,7 @@ struct auxtrace_recordif (!evlist)return NULL ; - cs_etm_pmu =perf_pmu__find (CORESIGHT_ETM_PMU_NAME ); + cs_etm_pmu =perf_pmus__find (CORESIGHT_ETM_PMU_NAME ); arm_spe_pmus =find_all_arm_spe_pmus (&nr_spes, err); hisi_ptt_pmus =find_all_hisi_ptt_pmus (&nr_ptts, err);
tools/perf/arch/arm/util/cs-etm.c +2 -2
@@ -25,7 +25,7 @@#include "../../../util/evsel.h" #include "../../../util/perf_api_probe.h" #include "../../../util/evsel_config.h" -#include "../../../util/pmu.h" +#include "../../../util/pmus.h" #include "../../../util/cs-etm.h" #include <internal/lib.h> // page_size #include "../../../util/session.h"
@@ -881,7 +881,7 @@ struct auxtrace_record *cs_etm_record_init(int *err)struct perf_pmu *cs_etm_pmu;struct cs_etm_recording *ptr; - cs_etm_pmu =perf_pmu__find (CORESIGHT_ETM_PMU_NAME ); + cs_etm_pmu =perf_pmus__find (CORESIGHT_ETM_PMU_NAME );if (!cs_etm_pmu) { *err = -EINVAL ;
tools/perf/arch/arm64/util/pmu.c +2 -1
@@ -3,6 +3,7 @@#include <internal/cpumap.h> #include "../../../util/cpumap.h" #include "../../../util/pmu.h" +#include "../../../util/pmus.h" #include <api/fs/fs.h> #include <math.h>
@@ -10,7 +11,7 @@ static struct perf_pmu *pmu__find_core_pmu(void) {struct perf_pmu *pmu =NULL ; -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (!is_pmu_core (pmu->name))continue ;
tools/perf/arch/x86/tests/hybrid.c +3 -2
@@ -4,6 +4,7 @@#include "evlist.h" #include "evsel.h" #include "pmu.h" +#include "pmus.h" #include "tests/tests.h" static bool test_config (const struct evsel *evsel,__u64 expected_config)
@@ -113,7 +114,7 @@ static int test__hybrid_raw1(struct evlist *evlist)struct perf_evsel *evsel;perf_evlist__for_each_evsel (&evlist->core, evsel) { -struct perf_pmu *pmu =perf_pmu__find_by_type (evsel->attr.type); +struct perf_pmu *pmu =perf_pmus__find_by_type (evsel->attr.type);TEST_ASSERT_VAL ("missing pmu" , pmu);TEST_ASSERT_VAL ("unexpected pmu" , !strncmp (pmu->name,"cpu_" ,4 ));
@@ -280,7 +281,7 @@ static int test_events(const struct evlist_test *events, int cnt)int test__hybrid (struct test_suite *test__maybe_unused ,int subtest__maybe_unused ) { -if (!perf_pmu__has_hybrid ()) +if (!perf_pmus__has_hybrid ())return TEST_SKIP ;return test_events (test__hybrid_events,ARRAY_SIZE (test__hybrid_events));
tools/perf/arch/x86/util/auxtrace.c +3 -2
@@ -10,6 +10,7 @@#include "../../../util/header.h" #include "../../../util/debug.h" #include "../../../util/pmu.h" +#include "../../../util/pmus.h" #include "../../../util/auxtrace.h" #include "../../../util/intel-pt.h" #include "../../../util/intel-bts.h"
@@ -25,8 +26,8 @@ struct auxtrace_record *auxtrace_record__init_intel(struct evlist *evlist,bool found_pt = false;bool found_bts = false; - intel_pt_pmu =perf_pmu__find (INTEL_PT_PMU_NAME ); - intel_bts_pmu =perf_pmu__find (INTEL_BTS_PMU_NAME ); + intel_pt_pmu =perf_pmus__find (INTEL_PT_PMU_NAME ); + intel_bts_pmu =perf_pmus__find (INTEL_BTS_PMU_NAME );evlist__for_each_entry (evlist, evsel) {if (intel_pt_pmu && evsel->core.attr.type == intel_pt_pmu->type)
tools/perf/arch/x86/util/evlist.c +3 -2
@@ -1,6 +1,7 @@// SPDX-License-Identifier: GPL-2.0 #include <stdio.h> #include "util/pmu.h" +#include "util/pmus.h" #include "util/evlist.h" #include "util/parse-events.h" #include "util/event.h"
@@ -17,7 +18,7 @@ static int ___evlist__add_default_attrs(struct evlist *evlist,for (i =0 ; i < nr_attrs; i++)event_attr_init (attrs + i); -if (!perf_pmu__has_hybrid ()) +if (!perf_pmus__has_hybrid ())return evlist__add_attrs (evlist, attrs, nr_attrs);for (i =0 ; i < nr_attrs; i++) {
@@ -32,7 +33,7 @@ static int ___evlist__add_default_attrs(struct evlist *evlist,continue ; } -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {struct perf_cpu_map *cpus;struct evsel *evsel;
tools/perf/arch/x86/util/evsel.c +4 -3
@@ -4,6 +4,7 @@#include "util/evsel.h" #include "util/env.h" #include "util/pmu.h" +#include "util/pmus.h" #include "linux/string.h" #include "evsel.h" #include "util/debug.h"
@@ -30,7 +31,7 @@ bool evsel__sys_has_perf_metrics(const struct evsel *evsel)* should be good enough to detect the perf metrics feature. */ if ((evsel->core.attr.type ==PERF_TYPE_RAW ) && -pmu_have_event (pmu_name,"slots" )) +perf_pmus__have_event (pmu_name,"slots" ))return true;return false;
@@ -98,8 +99,8 @@ void arch__post_evsel_config(struct evsel *evsel, struct perf_event_attr *attr)if (!evsel_pmu)return ; - ibs_fetch_pmu =perf_pmu__find ("ibs_fetch" ); - ibs_op_pmu =perf_pmu__find ("ibs_op" ); + ibs_fetch_pmu =perf_pmus__find ("ibs_fetch" ); + ibs_op_pmu =perf_pmus__find ("ibs_op" );if (ibs_fetch_pmu && ibs_fetch_pmu->type == evsel_pmu->type) {if (attr->config &IBS_FETCH_L3MISSONLY ) {
tools/perf/arch/x86/util/intel-bts.c +2 -2
@@ -17,7 +17,7 @@#include "../../../util/evlist.h" #include "../../../util/mmap.h" #include "../../../util/session.h" -#include "../../../util/pmu.h" +#include "../../../util/pmus.h" #include "../../../util/debug.h" #include "../../../util/record.h" #include "../../../util/tsc.h"
@@ -416,7 +416,7 @@ static int intel_bts_find_snapshot(struct auxtrace_record *itr, int idx,struct auxtrace_record *intel_bts_recording_init (int *err) { -struct perf_pmu *intel_bts_pmu =perf_pmu__find (INTEL_BTS_PMU_NAME ); +struct perf_pmu *intel_bts_pmu =perf_pmus__find (INTEL_BTS_PMU_NAME );struct intel_bts_recording *btsr;if (!intel_bts_pmu)
tools/perf/arch/x86/util/intel-pt.c +2 -2
@@ -23,7 +23,7 @@#include "../../../util/mmap.h" #include <subcmd/parse-options.h> #include "../../../util/parse-events.h" -#include "../../../util/pmu.h" +#include "../../../util/pmus.h" #include "../../../util/debug.h" #include "../../../util/auxtrace.h" #include "../../../util/perf_api_probe.h"
@@ -1185,7 +1185,7 @@ static u64 intel_pt_reference(struct auxtrace_record *itr __maybe_unused)struct auxtrace_record *intel_pt_recording_init (int *err) { -struct perf_pmu *intel_pt_pmu =perf_pmu__find (INTEL_PT_PMU_NAME ); +struct perf_pmu *intel_pt_pmu =perf_pmus__find (INTEL_PT_PMU_NAME );struct intel_pt_recording *ptr;if (!intel_pt_pmu)
tools/perf/arch/x86/util/mem-events.c +5 -4
@@ -1,5 +1,6 @@// SPDX-License-Identifier: GPL-2.0 #include "util/pmu.h" +#include "util/pmus.h" #include "util/env.h" #include "map_symbol.h" #include "mem-events.h"
@@ -55,12 +56,12 @@ struct perf_mem_event *perf_mem_events__ptr(int i)bool is_mem_loads_aux_event (struct evsel *leader) { -struct perf_pmu *pmu =perf_pmu__find ("cpu" ); +struct perf_pmu *pmu =perf_pmus__find ("cpu" );if (!pmu) - pmu =perf_pmu__find ("cpu_core" ); + pmu =perf_pmus__find ("cpu_core" ); -if (pmu && !pmu_have_event (pmu->name,"mem-loads-aux" )) +if (pmu && !perf_pmu__have_event (pmu,"mem-loads-aux" ))return false;return leader->core.attr.config ==MEM_LOADS_AUX ;
@@ -82,7 +83,7 @@ char *perf_mem_events__name(int i, char *pmu_name) pmu_name = (char *)"cpu" ; } -if (pmu_have_event (pmu_name,"mem-loads-aux" )) { +if (perf_pmus__have_event (pmu_name,"mem-loads-aux" )) {scnprintf (mem_loads_name,sizeof (mem_loads_name),MEM_LOADS_AUX_NAME , pmu_name, pmu_name, perf_mem_events__loads_ldlat);
tools/perf/arch/x86/util/perf_regs.c +3 -2
@@ -10,6 +10,7 @@#include "../../../util/debug.h" #include "../../../util/event.h" #include "../../../util/pmu.h" +#include "../../../util/pmus.h" const struct sample_reg sample_reg_masks[] = {SMPL_REG (AX ,PERF_REG_X86_AX ),
@@ -291,7 +292,7 @@ uint64_t arch__intr_reg_mask(void)*/ attr.sample_period =1 ; -if (perf_pmu__has_hybrid ()) { +if (perf_pmus__has_hybrid ()) {struct perf_pmu *pmu =NULL ;__u64 type =PERF_TYPE_RAW ;
@@ -299,7 +300,7 @@ 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_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {if (pmu->is_core) { type = pmu->type;break ;
tools/perf/arch/x86/util/topdown.c +3 -2
@@ -2,6 +2,7 @@#include "api/fs/fs.h" #include "util/evsel.h" #include "util/pmu.h" +#include "util/pmus.h" #include "util/topdown.h" #include "topdown.h" #include "evsel.h"
@@ -22,8 +23,8 @@ bool topdown_sys_has_perf_metrics(void)* The slots event is only available when the core PMU * supports the perf metrics feature. */ - pmu =perf_pmu__find_by_type (PERF_TYPE_RAW ); -if (pmu &&pmu_have_event (pmu->name,"slots" )) + pmu =perf_pmus__find_by_type (PERF_TYPE_RAW ); +if (pmu &&perf_pmu__have_event (pmu,"slots" )) has_perf_metrics = true; cached = true;
tools/perf/bench/pmu-scan.c +5 -5
@@ -44,7 +44,7 @@ static int save_result(void)struct list_head *list;struct pmu_scan_result *r; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) { r =realloc (results, (nr_pmus +1 ) *sizeof (*r));if (r ==NULL )return -ENOMEM ;
@@ -68,7 +68,7 @@ static int save_result(void) nr_pmus++; } -perf_pmu__destroy (); +perf_pmus__destroy ();return 0 ; }
@@ -81,7 +81,7 @@ static int check_result(void)for (int i =0 ; i < nr_pmus; i++) { r = &results[i]; - pmu =perf_pmu__find (r->name); + pmu =perf_pmus__find (r->name);if (pmu ==NULL ) {pr_err ("Cannot find PMU %s\n" , r->name);return -1 ;
@@ -144,7 +144,7 @@ static int run_pmu_scan(void)for (i =0 ; i < iterations; i++) {gettimeofday (&start,NULL ); -perf_pmu__scan (NULL ); +perf_pmus__scan (NULL );gettimeofday (&end,NULL );timersub (&end, &start, &diff);
@@ -152,7 +152,7 @@ static int run_pmu_scan(void)update_stats (&stats, runtime_us); ret =check_result (); -perf_pmu__destroy (); +perf_pmus__destroy ();if (ret <0 )break ; }
tools/perf/builtin-c2c.c +2 -2
@@ -41,7 +41,7 @@#include "symbol.h" #include "ui/ui.h" #include "ui/progress.h" -#include "pmu.h" +#include "pmus.h" #include "string2.h" #include "util/util.h"
@@ -3259,7 +3259,7 @@ static int perf_c2c__record(int argc, const char **argv) PARSE_OPT_KEEP_UNKNOWN);/* Max number of arguments multiplied by number of PMUs that can support them. */ - rec_argc = argc +11 *perf_pmu__num_mem_pmus (); + rec_argc = argc +11 *perf_pmus__num_mem_pmus (); rec_argv =calloc (rec_argc +1 ,sizeof (char *));if (!rec_argv)
tools/perf/builtin-list.c +2 -2
@@ -522,7 +522,7 @@ int cmd_list(int argc, const char **argv) strcmp(argv[i], "hwcache") == 0)print_hwcache_events (&print_cb, ps); else if (strcmp(argv[i], "pmu") == 0) - print_pmu_events(&print_cb, ps); + perf_pmus__print_pmu_events(&print_cb, ps); else if (strcmp(argv[i], "sdt") == 0) print_sdt_events(&print_cb, ps); else if (strcmp(argv[i], "metric") == 0 || strcmp(argv[i], "metrics") == 0) {
@@ -562,7 +562,7 @@ int cmd_list(int argc, const char **argv) event_symbols_sw, PERF_COUNT_SW_MAX);print_tool_events (&print_cb, ps);print_hwcache_events (&print_cb, ps); -print_pmu_events (&print_cb, ps); +perf_pmus__print_pmu_events (&print_cb, ps);print_tracepoint_events (&print_cb, ps);print_sdt_events (&print_cb, ps); default_ps.metrics = true;
tools/perf/builtin-mem.c +2 -2
@@ -17,7 +17,7 @@#include "util/dso.h" #include "util/map.h" #include "util/symbol.h" -#include "util/pmu.h" +#include "util/pmus.h" #include "util/sample.h" #include "util/string2.h" #include "util/util.h"
@@ -93,7 +93,7 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem) PARSE_OPT_KEEP_UNKNOWN);/* Max number of arguments multiplied by number of PMUs that can support them. */ - rec_argc = argc +9 *perf_pmu__num_mem_pmus (); + rec_argc = argc +9 *perf_pmus__num_mem_pmus ();if (mem->cpu_list) rec_argc +=2 ;
tools/perf/builtin-record.c +4 -2
@@ -48,6 +48,8 @@#include "util/bpf-event.h" #include "util/util.h" #include "util/pfm.h" +#include "util/pmu.h" +#include "util/pmus.h" #include "util/clockid.h" #include "util/off_cpu.h" #include "util/bpf-filter.h"
@@ -1292,7 +1294,7 @@ static int record__open(struct record *rec)* of waiting or event synthesis. */ if (opts->target.initial_delay ||target__has_cpu (&opts->target) || -perf_pmu__has_hybrid ()) { +perf_pmus__has_hybrid ()) { pos =evlist__get_tracking_event (evlist);if (!evsel__is_dummy_event (pos)) {/* Set up dummy event. */
@@ -2191,7 +2193,7 @@ static void record__uniquify_name(struct record *rec)char *new_name;int ret; -if (!perf_pmu__has_hybrid ()) +if (!perf_pmus__has_hybrid ())return ;evlist__for_each_entry (evlist, pos) {
tools/perf/builtin-stat.c +2 -2
@@ -1882,11 +1882,11 @@ static int add_default_attributes(void)if (evlist__add_default_attrs (evsel_list, default_attrs0) <0 )return -1 ; -if (pmu_have_event ("cpu" ,"stalled-cycles-frontend" )) { +if (perf_pmus__have_event ("cpu" ,"stalled-cycles-frontend" )) {if (evlist__add_default_attrs (evsel_list, frontend_attrs) <0 )return -1 ; } -if (pmu_have_event ("cpu" ,"stalled-cycles-backend" )) { +if (perf_pmus__have_event ("cpu" ,"stalled-cycles-backend" )) {if (evlist__add_default_attrs (evsel_list, backend_attrs) <0 )return -1 ; }
tools/perf/tests/attr.c +2 -2
@@ -34,7 +34,7 @@#include "event.h" #include "util.h" #include "tests.h" -#include "pmu.h" +#include "pmus.h" #define ENV "PERF_TEST_ATTR"
@@ -185,7 +185,7 @@ static int test__attr(struct test_suite *test __maybe_unused, int subtest __maybchar path_dir[PATH_MAX ];char *exec_path; -if (perf_pmu__has_hybrid ()) +if (perf_pmus__has_hybrid ())return TEST_SKIP ;/* First try development tree tests. */
tools/perf/tests/event_groups.c +1 -1
@@ -53,7 +53,7 @@ static int setup_uncore_event(void)struct perf_pmu *pmu =NULL ;int i, fd; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {for (i =0 ; i <NR_UNCORE_PMUS ; i++) {if (!strcmp (uncore_pmus[i].name, pmu->name)) {pr_debug ("Using %s for uncore pmu event\n" , pmu->name);
tools/perf/tests/parse-events.c +4 -4
@@ -112,7 +112,7 @@ static int test__checkevent_raw(struct evlist *evlist)bool type_matched = false;TEST_ASSERT_VAL ("wrong config" ,test_perf_config (evsel,0x1a )); -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {if (pmu->type == evsel->attr.type) {TEST_ASSERT_VAL ("PMU type expected once" , !type_matched); type_matched = true;
@@ -1443,12 +1443,12 @@ static int test__checkevent_config_cache(struct evlist *evlist)static bool test__pmu_cpu_valid (void ) { -return !!perf_pmu__find ("cpu" ); +return !!perf_pmus__find ("cpu" ); }static bool test__intel_pt_valid (void ) { -return !!perf_pmu__find ("intel_pt" ); +return !!perf_pmus__find ("intel_pt" ); }static int test__intel_pt (struct evlist *evlist)
@@ -2246,7 +2246,7 @@ static int test__pmu_events(struct test_suite *test __maybe_unused, int subteststruct perf_pmu *pmu =NULL ;int ret =TEST_OK ; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {struct stat st;char path[PATH_MAX ];struct dirent *ent;
tools/perf/tests/parse-metric.c +2 -2
@@ -11,7 +11,7 @@#include "debug.h" #include "expr.h" #include "stat.h" -#include "pmu.h" +#include "pmus.h" struct value {const char *event;
@@ -303,7 +303,7 @@ static int test__parse_metric(struct test_suite *test __maybe_unused, int subtesTEST_ASSERT_VAL ("recursion fail failed" ,test_recursion_fail () ==0 );TEST_ASSERT_VAL ("Memory bandwidth" ,test_memory_bandwidth () ==0 ); -if (!perf_pmu__has_hybrid ()) { +if (!perf_pmus__has_hybrid ()) {TEST_ASSERT_VAL ("cache_miss_cycles failed" ,test_cache_miss_cycles () ==0 );TEST_ASSERT_VAL ("test metric group" ,test_metric_group () ==0 ); }
tools/perf/tests/pmu-events.c +2 -1
@@ -2,6 +2,7 @@#include "math.h" #include "parse-events.h" #include "pmu.h" +#include "pmus.h" #include "tests.h" #include <errno.h> #include <stdio.h>
@@ -708,7 +709,7 @@ static int test__aliases(struct test_suite *test __maybe_unused,struct perf_pmu *pmu =NULL ;unsigned long i; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {int count =0 ;if (!is_pmu_core (pmu->name))
tools/perf/tests/switch-tracking.c +2 -2
@@ -20,7 +20,7 @@#include "tests.h" #include "util/mmap.h" #include "util/sample.h" -#include "pmu.h" +#include "pmus.h" static int spin_sleep (void ) {
@@ -375,7 +375,7 @@ static int test__switch_tracking(struct test_suite *test __maybe_unused, int sub cpu_clocks_evsel =evlist__last (evlist);/* Second event */ -if (perf_pmu__has_hybrid ()) { +if (perf_pmus__has_hybrid ()) { cycles ="cpu_core/cycles/u" ; err =parse_event (evlist, cycles);if (err) {
tools/perf/tests/topology.c +2 -2
@@ -8,7 +8,7 @@#include "session.h" #include "evlist.h" #include "debug.h" -#include "pmu.h" +#include "pmus.h" #include <linux/err.h> #define TEMPL "/tmp/perf-test-XXXXXX"
@@ -41,7 +41,7 @@ static int session_write_header(char *path) session =perf_session__new (&data,NULL );TEST_ASSERT_VAL ("can't get session" , !IS_ERR (session)); -if (!perf_pmu__has_hybrid ()) { +if (!perf_pmus__has_hybrid ()) { session->evlist =evlist__new_default ();TEST_ASSERT_VAL ("can't get evlist" , session->evlist); }else {
tools/perf/util/cputopo.c +4 -3
@@ -13,6 +13,7 @@#include "debug.h" #include "env.h" #include "pmu.h" +#include "pmus.h" #define PACKAGE_CPUS_FMT \ "%s/devices/system/cpu/cpu%d/topology/package_cpus_list"
@@ -473,10 +474,10 @@ struct hybrid_topology *hybrid_topology__new(void)struct hybrid_topology *tp =NULL ;u32 nr =0 , i =0 ; -if (!perf_pmu__has_hybrid ()) +if (!perf_pmus__has_hybrid ())return NULL ; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {if (pmu->is_core) nr++; }
@@ -488,7 +489,7 @@ struct hybrid_topology *hybrid_topology__new(void)return NULL ; tp->nr = nr; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {if (!pmu->is_core)continue ;
tools/perf/util/env.c +3 -2
@@ -10,6 +10,7 @@#include <sys/utsname.h> #include <stdlib.h> #include <string.h> +#include "pmus.h" #include "strbuf.h" struct perf_env perf_env;
@@ -323,7 +324,7 @@ int perf_env__read_pmu_mappings(struct perf_env *env)u32 pmu_num =0 ;struct strbuf sb; -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (!pmu->name)continue ; pmu_num++;
@@ -337,7 +338,7 @@ int perf_env__read_pmu_mappings(struct perf_env *env)if (strbuf_init (&sb,128 * pmu_num) <0 )return -ENOMEM ; -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (!pmu->name)continue ;if (strbuf_addf (&sb,"%u:%s" , pmu->type, pmu->name) <0 )
tools/perf/util/evsel.c +2 -1
@@ -48,6 +48,7 @@#include "util/hashmap.h" #include "off_cpu.h" #include "pmu.h" +#include "pmus.h" #include "../perf-sys.h" #include "util/parse-branch-options.h" #include "util/bpf-filter.h"
@@ -3139,7 +3140,7 @@ void evsel__zero_per_pkg(struct evsel *evsel)*/ bool evsel__is_hybrid (const struct evsel *evsel) { -if (!perf_pmu__has_hybrid ()) +if (!perf_pmus__has_hybrid ())return false;return evsel->core.is_pmu_core;
tools/perf/util/header.c +8 -7
@@ -37,6 +37,7 @@#include "debug.h" #include "cpumap.h" #include "pmu.h" +#include "pmus.h" #include "vdso.h" #include "strbuf.h" #include "build-id.h"
@@ -744,7 +745,7 @@ static int write_pmu_mappings(struct feat_fd *ff,* Do a first pass to count number of pmu to avoid lseek so this * works in pipe mode as well. */ -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (!pmu->name)continue ; pmu_num++;
@@ -754,7 +755,7 @@ static int write_pmu_mappings(struct feat_fd *ff,if (ret <0 )return ret; -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (!pmu->name)continue ;
@@ -1550,7 +1551,7 @@ static int __write_pmu_caps(struct feat_fd *ff, struct perf_pmu *pmu,static int write_cpu_pmu_caps (struct feat_fd *ff,struct evlist *evlist__maybe_unused ) { -struct perf_pmu *cpu_pmu =perf_pmu__find ("cpu" ); +struct perf_pmu *cpu_pmu =perf_pmus__find ("cpu" );int ret;if (!cpu_pmu)
@@ -1570,7 +1571,7 @@ static int write_pmu_caps(struct feat_fd *ff,int nr_pmu =0 ;int ret; -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (!pmu->name || !strcmp (pmu->name,"cpu" ) ||perf_pmu__caps_parse (pmu) <=0 )continue ;
@@ -1588,9 +1589,9 @@ static int write_pmu_caps(struct feat_fd *ff,* Write hybrid pmu caps first to maintain compatibility with * older perf tool. */ -if (perf_pmu__has_hybrid ()) { +if (perf_pmus__has_hybrid ()) { pmu =NULL ; -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (!pmu->is_core)continue ;
@@ -1601,7 +1602,7 @@ static int write_pmu_caps(struct feat_fd *ff, } pmu =NULL ; -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (pmu->is_core || !pmu->nr_caps)continue ;
tools/perf/util/mem-events.c +6 -5
@@ -13,6 +13,7 @@#include "debug.h" #include "symbol.h" #include "pmu.h" +#include "pmus.h" unsigned int perf_mem_events__loads_ldlat =30 ;
@@ -128,14 +129,14 @@ int perf_mem_events__init(void)if (!e->tag)continue ; -if (!perf_pmu__has_hybrid ()) { +if (!perf_pmus__has_hybrid ()) {scnprintf (sysfs_name,sizeof (sysfs_name), e->sysfs_name,"cpu" ); e->supported =perf_mem_event__supported (mnt, sysfs_name); }else {struct perf_pmu *pmu =NULL ; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {if (!pmu->is_core)continue ;
@@ -175,7 +176,7 @@ 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_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {if (!pmu->is_core)continue ;
@@ -201,7 +202,7 @@ int perf_mem_events__record_args(const char **rec_argv, int *argv_nr,if (!e->record)continue ; -if (!perf_pmu__has_hybrid ()) { +if (!perf_pmus__has_hybrid ()) {if (!e->supported) {pr_err ("failed: event '%s' not supported\n" ,perf_mem_events__name (j,NULL ));
@@ -216,7 +217,7 @@ int perf_mem_events__record_args(const char **rec_argv, int *argv_nr,return -1 ; } -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {if (!pmu->is_core)continue ; rec_argv[i++] ="-e" ;
tools/perf/util/metricgroup.c +3 -2
@@ -11,6 +11,7 @@#include "evsel.h" #include "strbuf.h" #include "pmu.h" +#include "pmus.h" #include "print-events.h" #include "smt.h" #include "expr.h"
@@ -273,7 +274,7 @@ static int setup_metric_events(const char *pmu, struct hashmap *ids,const char *metric_id;struct evsel *ev;size_t ids_size, matched_events, i; -bool all_pmus = !strcmp (pmu,"all" ) || !perf_pmu__has_hybrid () || !is_pmu_hybrid (pmu); +bool all_pmus = !strcmp (pmu,"all" ) || !perf_pmus__has_hybrid () || !is_pmu_hybrid (pmu); *out_metric_events =NULL ; ids_size =hashmap__size (ids);
@@ -488,7 +489,7 @@ static int metricgroup__sys_event_iter(const struct pmu_metric *pm,if (!pm->metric_expr || !pm->compat)return 0 ; -while ((pmu =perf_pmu__scan (pmu))) { +while ((pmu =perf_pmus__scan (pmu))) {if (!pmu->id ||strcmp (pmu->id, pm->compat))continue ;
tools/perf/util/parse-events.c +8 -7
@@ -21,6 +21,7 @@#include "parse-events-bison.h" #include "parse-events-flex.h" #include "pmu.h" +#include "pmus.h" #include "asm/bug.h" #include "util/parse-branch-options.h" #include "util/evsel_config.h"
@@ -452,7 +453,7 @@ 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_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {LIST_HEAD (config_terms);struct perf_event_attr attr;int ret;
@@ -1193,7 +1194,7 @@ static int config_term_pmu(struct perf_event_attr *attr, struct parse_events_error *err) {if (term->type_term ==PARSE_EVENTS__TERM_TYPE_LEGACY_CACHE ) { -const struct perf_pmu *pmu =perf_pmu__find_by_type (attr->type); +const struct perf_pmu *pmu =perf_pmus__find_by_type (attr->type);if (perf_pmu__supports_legacy_cache (pmu)) { attr->type =PERF_TYPE_HW_CACHE ;
@@ -1203,7 +1204,7 @@ static int config_term_pmu(struct perf_event_attr *attr, term->type_term =PARSE_EVENTS__TERM_TYPE_USER ; }if (term->type_term ==PARSE_EVENTS__TERM_TYPE_HARDWARE ) { -const struct perf_pmu *pmu =perf_pmu__find_by_type (attr->type); +const struct perf_pmu *pmu =perf_pmus__find_by_type (attr->type);if (!pmu) {pr_debug ("Failed to find PMU for type %d" , attr->type);
@@ -1480,7 +1481,7 @@ 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_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {int ret;if (!perf_pmu__supports_wildcard_numeric (pmu))
@@ -1529,7 +1530,7 @@ int parse_events_add_pmu(struct parse_events_state *parse_state,struct parse_events_error *err = parse_state->error;LIST_HEAD (config_terms); - pmu = parse_state->fake_pmu ?:perf_pmu__find (name); + pmu = parse_state->fake_pmu ?:perf_pmus__find (name);if (verbose >1 && !(pmu && pmu->selectable)) { fprintf(stderr, "Attempting to add event pmu '%s' with '",
@@ -1674,7 +1675,7 @@ int parse_events_multi_pmu_add(struct parse_events_state *parse_state,INIT_LIST_HEAD (list); -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {struct perf_pmu_alias *alias;bool auto_merge_stats;
@@ -2410,7 +2411,7 @@ static int set_filter(struct evsel *evsel, const void *arg)return 0 ; } -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) +while ((pmu =perf_pmus__scan (pmu)) !=NULL )if (pmu->type == evsel->core.attr.type) { found = true;break ;
tools/perf/util/parse-events.y +2 -1
@@ -15,6 +15,7 @@ #include <linux/types.h> #include <linux/zalloc.h> #include "pmu.h" +#include "pmus.h" #include "evsel.h" #include "parse-events.h" #include "parse-events-bison.h"
@@ -316,7 +317,7 @@ PE_NAME opt_pmu_config if (asprintf(&pattern, "%s*", $1) < 0) CLEANUP_YYABORT; - while ((pmu = perf_pmu__scan(pmu)) != NULL) { + while ((pmu = perf_pmus__scan(pmu)) != NULL) { char *name = pmu->name; if (parse_events__filter_pmu(parse_state, pmu))
tools/perf/util/pfm.c +3 -3
@@ -10,7 +10,7 @@#include "util/evlist.h" #include "util/evsel.h" #include "util/parse-events.h" -#include "util/pmu.h" +#include "util/pmus.h" #include "util/pfm.h" #include "util/strbuf.h"
@@ -49,7 +49,7 @@ int parse_libpfm_events_option(const struct option *opt, const char *str,/* * force loading of the PMU list */ -perf_pmu__scan (NULL ); +perf_pmus__scan (NULL );for (q = p;strsep (&p,",{}" ); q = p) { sep = p ? str + (p - p_orig -1 ) :"" ;
@@ -86,7 +86,7 @@ int parse_libpfm_events_option(const struct option *opt, const char *str, goto error; } - pmu =perf_pmu__find_by_type ((unsigned int )attr.type); + pmu =perf_pmus__find_by_type ((unsigned int )attr.type); evsel =parse_events__add_event (evlist->core.nr_entries, &attr, q,/*metric_id=*/ NULL , pmu);
tools/perf/util/pmu.c +8 -403
496 lines not on this page · raw message
tools/perf/util/pmu.h +4 -9
@@ -198,8 +198,6 @@ struct perf_pmu_alias {char *pmu_name; }; -struct perf_pmu *perf_pmu__find (const char *name); -struct perf_pmu *perf_pmu__find_by_type (unsigned int type);void pmu_add_sys_aliases (struct list_head *head,struct perf_pmu *pmu); int perf_pmu__config(struct perf_pmu *pmu, struct perf_event_attr *attr, struct list_head *head_terms,
@@ -222,16 +220,13 @@ void perf_pmu__set_format(unsigned long *bits, long from, long to);int perf_pmu__format_parse (int dirfd,struct list_head *head);void perf_pmu__del_formats (struct list_head *formats); -struct perf_pmu *perf_pmu__scan (struct perf_pmu *pmu); -bool is_pmu_core (const char *name);bool is_pmu_hybrid (const char *name);bool perf_pmu__supports_legacy_cache (const struct perf_pmu *pmu);bool perf_pmu__supports_wildcard_numeric (const struct perf_pmu *pmu);bool perf_pmu__auto_merge_stats (const struct perf_pmu *pmu); -int perf_pmu__num_mem_pmus (void ); -void print_pmu_events (const struct print_callbacks *print_cb,void *print_state); -bool pmu_have_event (const char *pname,const char *name); +bool perf_pmu__is_mem_pmu (const struct perf_pmu *pmu); +bool perf_pmu__have_event (const struct perf_pmu *pmu,const char *name);FILE *perf_pmu__open_file (struct perf_pmu *pmu,const char *name);FILE *perf_pmu__open_file_at (struct perf_pmu *pmu,int dirfd,const char *name);
@@ -261,7 +256,6 @@ void perf_pmu__warn_invalid_config(struct perf_pmu *pmu, __u64 config, const char *name);void perf_pmu__warn_invalid_formats (struct perf_pmu *pmu); -bool perf_pmu__has_hybrid (void );int perf_pmu__match (char *pattern,char *name,char *tok);char *pmu_find_real_name (const char *name);
@@ -273,6 +267,7 @@ int perf_pmu__pathname_scnprintf(char *buf, size_t size,int perf_pmu__event_source_devices_fd (void );int perf_pmu__pathname_fd (int dirfd,const char *pmu_name,const char *filename,int flags); -void perf_pmu__destroy (void ); +struct perf_pmu *perf_pmu__lookup (struct list_head *pmus,int dirfd,const char *lookup_name); +void perf_pmu__delete (struct perf_pmu *pmu);#endif /* __PMU_H */
tools/perf/util/pmus.c +394 -2
413 lines not on this page · raw message
tools/perf/util/pmus.h +13 -1
@@ -2,9 +2,21 @@#ifndef __PMUS_H#define __PMUS_H -extern struct list_head pmus;struct perf_pmu ; +struct print_callbacks ; + +void perf_pmus__destroy (void ); + +struct perf_pmu *perf_pmus__find (const char *name); +struct perf_pmu *perf_pmus__find_by_type (unsigned int type); + +struct perf_pmu *perf_pmus__scan (struct perf_pmu *pmu);const struct perf_pmu *perf_pmus__pmu_for_pmu_filter (const char *str); +int perf_pmus__num_mem_pmus (void ); +void perf_pmus__print_pmu_events (const struct print_callbacks *print_cb,void *print_state); +bool perf_pmus__have_event (const char *pname,const char *name); +bool perf_pmus__has_hybrid (void ); +#endif /* __PMUS_H */
tools/perf/util/print-events.c +3 -2
@@ -20,6 +20,7 @@#include "metricgroup.h" #include "parse-events.h" #include "pmu.h" +#include "pmus.h" #include "print-events.h" #include "probe-file.h" #include "string2.h"
@@ -271,7 +272,7 @@ int print_hwcache_events(const struct print_callbacks *print_cb, void *print_stastruct perf_pmu *pmu =NULL ;const char *event_type_descriptor = event_type_descriptors[PERF_TYPE_HW_CACHE ]; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) { +while ((pmu =perf_pmus__scan (pmu)) !=NULL ) {/* * Skip uncore PMUs for performance. PERF_TYPE_HW_CACHE type * attributes can accept software PMUs in the extended type, so
@@ -404,7 +405,7 @@ void print_events(const struct print_callbacks *print_cb, void *print_state)print_hwcache_events (print_cb, print_state); -print_pmu_events (print_cb, print_state); +perf_pmus__print_pmu_events (print_cb, print_state); print_cb->print_event(print_state, /*topic=*/NULL,
tools/perf/util/stat-display.c +2 -1
@@ -20,6 +20,7 @@#include "util.h" #include "iostat.h" #include "pmu.h" +#include "pmus.h" #define CNTR_NOT_SUPPORTED "<not supported>"#define CNTR_NOT_COUNTED "<not counted>"
@@ -680,7 +681,7 @@ static bool evlist__has_hybrid(struct evlist *evlist) {struct evsel *evsel; -if (!perf_pmu__has_hybrid ()) +if (!perf_pmus__has_hybrid ())return false;evlist__for_each_entry (evlist, evsel) {
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 29/35 perf pmus: Split pmus list into core and other
In reply to Ian RogersSplit the pmus list into core and other. This will later allow for the core and other pmus to be populated separately.
--- tools/perf/util/pmus.c | 52 ++++++++++++++++++++++++++++++------------ 1 file changed, 38 insertions(+), 14 deletions(-)
tools/perf/util/pmus.c +38 -14
@@ -12,13 +12,19 @@#include "pmu.h" #include "print-events.h" -static LIST_HEAD (pmus); +static LIST_HEAD (core_pmus); +static LIST_HEAD (other_pmus);void perf_pmus__destroy (void ) {struct perf_pmu *pmu, *tmp; -list_for_each_entry_safe (pmu, tmp, &pmus, list) { +list_for_each_entry_safe (pmu, tmp, &core_pmus, list) { +list_del (&pmu->list); + +perf_pmu__delete (pmu); + } +list_for_each_entry_safe (pmu, tmp, &other_pmus, list) {list_del (&pmu->list);perf_pmu__delete (pmu);
@@ -29,7 +35,12 @@ static struct perf_pmu *pmu_find(const char *name) {struct perf_pmu *pmu; -list_for_each_entry (pmu, &pmus, list) { +list_for_each_entry (pmu, &core_pmus, list) { +if (!strcmp (pmu->name, name) || + (pmu->alias_name && !strcmp (pmu->alias_name, name))) +return pmu; + } +list_for_each_entry (pmu, &other_pmus, list) {if (!strcmp (pmu->name, name) || (pmu->alias_name && !strcmp (pmu->alias_name, name)))return pmu;
@@ -53,7 +64,7 @@ struct perf_pmu *perf_pmus__find(const char *name)return pmu; dirfd =perf_pmu__event_source_devices_fd (); - pmu =perf_pmu__lookup (&pmus, dirfd, name); + pmu =perf_pmu__lookup (is_pmu_core (name) ? &core_pmus : &other_pmus, dirfd, name);close (dirfd);return pmu;
@@ -72,7 +83,7 @@ static struct perf_pmu *perf_pmu__find2(int dirfd, const char *name)if (pmu)return pmu; -return perf_pmu__lookup (&pmus, dirfd, name); +return perf_pmu__lookup (is_pmu_core (name) ? &core_pmus : &other_pmus, dirfd, name); }/* Add all pmus in sysfs to pmu list: */
@@ -93,7 +104,7 @@ static void pmu_read_sysfs(void)while ((dent =readdir (dir))) {if (!strcmp (dent->d_name,"." ) || !strcmp (dent->d_name,".." ))continue ; -/* add to static LIST_HEAD(pmus): */ +/* add to static LIST_HEAD(core_pmus) or LIST_HEAD(other_pmus): */ perf_pmu__find2 (fd, dent->d_name); }
@@ -104,24 +115,37 @@ struct perf_pmu *perf_pmus__find_by_type(unsigned int type) {struct perf_pmu *pmu; -list_for_each_entry (pmu, &pmus, list) +list_for_each_entry (pmu, &core_pmus, list) {if (pmu->type == type)return pmu; - + } +list_for_each_entry (pmu, &other_pmus, list) { +if (pmu->type == type) +return pmu; + }return NULL ; } +/* +* pmu iterator: If pmu is NULL, we start at the begin, otherwise return the +* next pmu. Returns NULL on end. +*/ struct perf_pmu *perf_pmus__scan (struct perf_pmu *pmu) { -/* -* pmu iterator: If pmu is NULL, we start at the begin, -* otherwise return the next pmu. Returns NULL on end. -*/ +bool use_core_pmus = !pmu || pmu->is_core; +if (!pmu) {pmu_read_sysfs (); - pmu =list_prepare_entry (pmu, &pmus, list); + pmu =list_prepare_entry (pmu, &core_pmus, list); + } +if (use_core_pmus) { +list_for_each_entry_continue (pmu, &core_pmus, list) +return pmu; + + pmu =NULL ; + pmu =list_prepare_entry (pmu, &other_pmus, list); } -list_for_each_entry_continue (pmu, &pmus, list) +list_for_each_entry_continue (pmu, &other_pmus, list)return pmu;return NULL ; }
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 30/35 perf pmus: Allow just core PMU scanning
In reply to Ian RogersScanning 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.
--- tools/perf/arch/arm64/util/pmu.c | 5 +-- tools/perf/arch/x86/util/evlist.c | 5 +-- tools/perf/arch/x86/util/perf_regs.c | 8 ++--- tools/perf/bench/pmu-scan.c | 50 ++++++++++++++++------------ tools/perf/tests/pmu-events.c | 5 +-- tools/perf/util/cputopo.c | 12 +++---- tools/perf/util/header.c | 5 +-- tools/perf/util/mem-events.c | 14 ++------ tools/perf/util/parse-events.c | 13 +++----- tools/perf/util/pmu.c | 10 ------ tools/perf/util/pmu.h | 2 -- tools/perf/util/pmus.c | 30 ++++++++++++----- tools/perf/util/pmus.h | 1 + tools/perf/util/print-events.c | 11 +++--- 14 files changed, 75 insertions(+), 96 deletions(-)
tools/perf/arch/arm64/util/pmu.c +1 -4
@@ -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
@@ -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
@@ -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 +29 -21
87 lines not on this page · raw message
tools/perf/tests/pmu-events.c +1 -4
@@ -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 +4 -8
@@ -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 +1 -4
@@ -1591,10 +1591,7 @@ 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 +3 -11
@@ -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 +4 -9
@@ -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 +0 -10
@@ -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;
tools/perf/util/pmu.h +0 -2
@@ -223,9 +223,7 @@ void perf_pmu__del_formats(struct list_head *formats);bool is_pmu_core (const char *name);bool is_pmu_hybrid (const char *name);bool perf_pmu__supports_legacy_cache (const struct perf_pmu *pmu); -bool perf_pmu__supports_wildcard_numeric (const struct perf_pmu *pmu);bool perf_pmu__auto_merge_stats (const struct perf_pmu *pmu); -bool perf_pmu__is_mem_pmu (const struct perf_pmu *pmu);bool perf_pmu__have_event (const struct perf_pmu *pmu,const char *name);FILE *perf_pmu__open_file (struct perf_pmu *pmu,const char *name);
tools/perf/util/pmus.c +22 -8
@@ -87,7 +87,7 @@ static struct perf_pmu *perf_pmu__find2(int dirfd, const char *name) }/* Add all pmus in sysfs to pmu list: */ -static void pmu_read_sysfs (void ) +static void pmu_read_sysfs (bool core_only) {int fd;DIR *dir;
@@ -104,6 +104,8 @@ static void pmu_read_sysfs(void)while ((dent =readdir (dir))) {if (!strcmp (dent->d_name,"." ) || !strcmp (dent->d_name,".." ))continue ; +if (core_only && !is_pmu_core (dent->d_name)) +continue ;/* add to static LIST_HEAD(core_pmus) or LIST_HEAD(other_pmus): */ perf_pmu__find2 (fd, dent->d_name); }
@@ -135,7 +137,7 @@ struct perf_pmu *perf_pmus__scan(struct perf_pmu *pmu)bool use_core_pmus = !pmu || pmu->is_core;if (!pmu) { -pmu_read_sysfs (); +pmu_read_sysfs (/*core_only=*/ false); pmu =list_prepare_entry (pmu, &core_pmus, list); }if (use_core_pmus) {
@@ -150,6 +152,18 @@ struct perf_pmu *perf_pmus__scan(struct perf_pmu *pmu)return NULL ; } +struct perf_pmu *perf_pmus__scan_core (struct perf_pmu *pmu) +{ +if (!pmu) { +pmu_read_sysfs (/*core_only=*/ true); + pmu =list_prepare_entry (pmu, &core_pmus, list); + } +list_for_each_entry_continue (pmu, &core_pmus, list) +return pmu; + +return NULL ; +} +const struct perf_pmu *perf_pmus__pmu_for_pmu_filter (const char *str) {struct perf_pmu *pmu =NULL ;
@@ -176,10 +190,10 @@ int perf_pmus__num_mem_pmus(void)struct perf_pmu *pmu =NULL ;int count =0 ; -while ((pmu =perf_pmus__scan (pmu)) !=NULL ) { -if (perf_pmu__is_mem_pmu (pmu)) - count++; - } +/* All core PMUs are for mem events. */ +while ((pmu =perf_pmus__scan_core (pmu)) !=NULL ) + count++; +return count; }
@@ -421,8 +435,8 @@ bool perf_pmus__has_hybrid(void)if (!hybrid_scanned) {struct perf_pmu *pmu =NULL ; -while ((pmu =perf_pmus__scan (pmu)) !=NULL ) { -if (pmu->is_core &&is_pmu_hybrid (pmu->name)) { +while ((pmu =perf_pmus__scan_core (pmu)) !=NULL ) { +if (is_pmu_hybrid (pmu->name)) { has_hybrid = true;break ; }
tools/perf/util/pmus.h +1 -0
@@ -11,6 +11,7 @@ struct perf_pmu *perf_pmus__find(const char *name);struct perf_pmu *perf_pmus__find_by_type (unsigned int type);struct perf_pmu *perf_pmus__scan (struct perf_pmu *pmu); +struct perf_pmu *perf_pmus__scan_core (struct perf_pmu *pmu);const struct perf_pmu *perf_pmus__pmu_for_pmu_filter (const char *str);
tools/perf/util/print-events.c +5 -6
@@ -272,12 +272,11 @@ int print_hwcache_events(const struct print_callbacks *print_cb, void *print_stastruct perf_pmu *pmu =NULL ;const char *event_type_descriptor = event_type_descriptors[PERF_TYPE_HW_CACHE ]; -while ((pmu =perf_pmus__scan (pmu)) !=NULL ) { -/* -* Skip uncore PMUs for performance. PERF_TYPE_HW_CACHE type -* attributes can accept software PMUs in the extended type, so -* also skip. -*/ +/* +* Only print core PMUs, skipping uncore for performance and +* PERF_TYPE_SOFTWARE that can succeed in opening legacy cache evenst. +*/ +while ((pmu =perf_pmus__scan_core (pmu)) !=NULL ) {if (pmu->is_uncore || pmu->type ==PERF_TYPE_SOFTWARE )continue ;
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 31/35 perf pmus: Avoid repeated sysfs scanning
In reply to Ian Rogersperf_pmus__scan will process every directory in sysfs to see if it is a PMU, attempting to add it if not already in the pmus list. Add two booleans to record whether this scanning has been done for core or all PMUs. Skip scanning in the event that scanning has already occurred.
--- tools/perf/util/pmus.c | 33 +++++++++++++++++++++++++++++++-- 1 file changed, 31 insertions(+), 2 deletions(-)
tools/perf/util/pmus.c +31 -2
@@ -14,6 +14,8 @@static LIST_HEAD (core_pmus);static LIST_HEAD (other_pmus); +static bool read_sysfs_core_pmus; +static bool read_sysfs_all_pmus;void perf_pmus__destroy (void ) {
@@ -29,6 +31,8 @@ void perf_pmus__destroy(void)perf_pmu__delete (pmu); } + read_sysfs_core_pmus = false; + read_sysfs_all_pmus = false; }static struct perf_pmu *pmu_find (const char *name)
@@ -53,6 +57,7 @@ struct perf_pmu *perf_pmus__find(const char *name) {struct perf_pmu *pmu;int dirfd; +bool core_pmu;/* * Once PMU is loaded it stays in the list,
@@ -63,8 +68,15 @@ struct perf_pmu *perf_pmus__find(const char *name)if (pmu)return pmu; +if (read_sysfs_all_pmus) +return NULL ; + + core_pmu =is_pmu_core (name); +if (core_pmu && read_sysfs_core_pmus) +return NULL ; + dirfd =perf_pmu__event_source_devices_fd (); - pmu =perf_pmu__lookup (is_pmu_core (name) ? &core_pmus : &other_pmus, dirfd, name); + pmu =perf_pmu__lookup (core_pmu ? &core_pmus : &other_pmus, dirfd, name);close (dirfd);return pmu;
@@ -73,6 +85,7 @@ struct perf_pmu *perf_pmus__find(const char *name)static struct perf_pmu *perf_pmu__find2 (int dirfd,const char *name) {struct perf_pmu *pmu; +bool core_pmu;/* * Once PMU is loaded it stays in the list,
@@ -83,7 +96,14 @@ static struct perf_pmu *perf_pmu__find2(int dirfd, const char *name)if (pmu)return pmu; -return perf_pmu__lookup (is_pmu_core (name) ? &core_pmus : &other_pmus, dirfd, name); +if (read_sysfs_all_pmus) +return NULL ; + + core_pmu =is_pmu_core (name); +if (core_pmu && read_sysfs_core_pmus) +return NULL ; + +return perf_pmu__lookup (core_pmu ? &core_pmus : &other_pmus, dirfd, name); }/* Add all pmus in sysfs to pmu list: */
@@ -93,6 +113,9 @@ static void pmu_read_sysfs(bool core_only)DIR *dir;struct dirent *dent; +if (read_sysfs_all_pmus || (core_only && read_sysfs_core_pmus)) +return ; + fd =perf_pmu__event_source_devices_fd ();if (fd <0 )return ;
@@ -111,6 +134,12 @@ static void pmu_read_sysfs(bool core_only) }closedir (dir); +if (core_only) { + read_sysfs_core_pmus = true; + }else { + read_sysfs_core_pmus = true; + read_sysfs_all_pmus = true; + } }struct perf_pmu *perf_pmus__find_by_type (unsigned int type)
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 32/35 perf pmus: Ensure all PMUs are read for find_by_type
In reply to Ian Rogersperf_pmus__find_by_type may be called for something like a raw event, in which case the PMU isn't guaranteed to have been looked up. Add a second check to make sure all PMUs are loaded.
--- tools/perf/util/pmus.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-)
tools/perf/util/pmus.c +14 -1
@@ -142,7 +142,7 @@ static void pmu_read_sysfs(bool core_only) } } -struct perf_pmu *perf_pmus__find_by_type (unsigned int type) +static struct perf_pmu *__perf_pmus__find_by_type (unsigned int type) {struct perf_pmu *pmu;
@@ -150,6 +150,7 @@ struct perf_pmu *perf_pmus__find_by_type(unsigned int type)if (pmu->type == type)return pmu; } +list_for_each_entry (pmu, &other_pmus, list) {if (pmu->type == type)return pmu;
@@ -157,6 +158,18 @@ struct perf_pmu *perf_pmus__find_by_type(unsigned int type)return NULL ; } +struct perf_pmu *perf_pmus__find_by_type (unsigned int type) +{ +struct perf_pmu *pmu =__perf_pmus__find_by_type (type); + +if (pmu || read_sysfs_all_pmus) +return pmu; + +pmu_read_sysfs (/*core_only=*/ false); + pmu =__perf_pmus__find_by_type (type); +return pmu; +} +/* * pmu iterator: If pmu is NULL, we start at the begin, otherwise return the * next pmu. Returns NULL on end.
Signature · 7 lines
2.41.0.rc0.172.g3f132b7071-goog _______________________________________________ linux-arm-kernel mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Ian Rogers
reviewerPATCH v4 33/35 perf pmus: Add function to return count of core PMUs
In reply to Ian RogersAdd perf_pmus__num_core_pmus that will count core PMUs holding the result in a static. Reuse for perf_pmus__num_mem_pmus.
--- tools/perf/util/pmus.c | 21 ++++++++++++++------- tools/perf/util/pmus.h | 1 + 2 files changed, 15 insertions(+), 7 deletions(-)
tools/perf/util/pmus.c +14 -7
@@ -229,14 +229,8 @@ const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str)int perf_pmus__num_mem_pmus (void ) { -struct perf_pmu *pmu =NULL ; -int count =0 ; -/* All core PMUs are for mem events. */ -while ((pmu =perf_pmus__scan_core (pmu)) !=NULL ) - count++; - -return count; +return perf_pmus__num_core_pmus (); }/** Struct for ordering events as output in perf list. */
@@ -488,6 +482,19 @@ bool perf_pmus__has_hybrid(void)return has_hybrid; } +int perf_pmus__num_core_pmus (void ) +{ +static int count; + +if (!count) { +struct perf_pmu *pmu =NULL ; + +while ((pmu =perf_pmus__scan_core (pmu)) !=NULL ) + count++; + } +return count; +} +struct perf_pmu *evsel__find_pmu (const struct evsel *evsel) {struct perf_pmu *pmu = evsel->pmu;
tools/perf/util/pmus.h +1 -0
@@ -19,5 +19,6 @@ int perf_pmus__num_mem_pmus(void);void perf_pmus__print_pmu_events (const struct print_callbacks *print_cb,void *print_state);bool perf_pmus__have_event (const char *pname,const char *name);bool perf_pmus__has_hybrid (void ); +int perf_pmus__num_core_pmus (void );#endif /* __PMUS_H */
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 34/35 perf pmus: Remove perf_pmus__has_hybrid
In reply to Ian Rogersperf_pmus__has_hybrid was used to detect when there was >1 core PMU, this can be achieved with perf_pmus__num_core_pmus that doesn't depend upon is_pmu_hybrid and PMU name comparisons. When modifying the function calls take the opportunity to improve comments, enable/simplify tests that were previously failing for hybrid but now pass and to simplify generic code.
--- tools/perf/arch/x86/tests/hybrid.c | 2 +- tools/perf/arch/x86/util/evlist.c | 2 +- tools/perf/arch/x86/util/perf_regs.c | 2 +- tools/perf/builtin-record.c | 4 ++-- tools/perf/tests/attr.c | 9 ++++++++- tools/perf/tests/parse-metric.c | 7 ++----- tools/perf/tests/switch-tracking.c | 12 +----------- tools/perf/tests/topology.c | 14 ++------------ tools/perf/util/cputopo.c | 10 ++-------- tools/perf/util/evsel.c | 2 +- tools/perf/util/header.c | 2 +- tools/perf/util/mem-events.c | 18 +++++------------- tools/perf/util/metricgroup.c | 2 +- tools/perf/util/pmus.c | 18 ------------------ tools/perf/util/pmus.h | 1 - tools/perf/util/stat-display.c | 2 +- 16 files changed, 29 insertions(+), 78 deletions(-)
tools/perf/arch/x86/tests/hybrid.c +1 -1
@@ -281,7 +281,7 @@ static int test_events(const struct evlist_test *events, int cnt)int test__hybrid (struct test_suite *test__maybe_unused ,int subtest__maybe_unused ) { -if (!perf_pmus__has_hybrid ()) +if (perf_pmus__num_core_pmus () ==1 )return TEST_SKIP ;return test_events (test__hybrid_events,ARRAY_SIZE (test__hybrid_events));
tools/perf/arch/x86/util/evlist.c +1 -1
@@ -18,7 +18,7 @@ static int ___evlist__add_default_attrs(struct evlist *evlist,for (i =0 ; i < nr_attrs; i++)event_attr_init (attrs + i); -if (!perf_pmus__has_hybrid ()) +if (perf_pmus__num_core_pmus () ==1 )return evlist__add_attrs (evlist, attrs, nr_attrs);for (i =0 ; i < nr_attrs; i++) {
tools/perf/arch/x86/util/perf_regs.c +1 -1
@@ -292,7 +292,7 @@ uint64_t arch__intr_reg_mask(void)*/ attr.sample_period =1 ; -if (perf_pmus__has_hybrid ()) { +if (perf_pmus__num_core_pmus () >1 ) {struct perf_pmu *pmu =NULL ;__u64 type =PERF_TYPE_RAW ;
tools/perf/builtin-record.c +2 -2
@@ -1294,7 +1294,7 @@ static int record__open(struct record *rec)* of waiting or event synthesis. */ if (opts->target.initial_delay ||target__has_cpu (&opts->target) || -perf_pmus__has_hybrid ()) { +perf_pmus__num_core_pmus () >1 ) { pos =evlist__get_tracking_event (evlist);if (!evsel__is_dummy_event (pos)) {/* Set up dummy event. */
@@ -2193,7 +2193,7 @@ static void record__uniquify_name(struct record *rec)char *new_name;int ret; -if (!perf_pmus__has_hybrid ()) +if (perf_pmus__num_core_pmus () ==1 )return ;evlist__for_each_entry (evlist, pos) {
tools/perf/tests/attr.c +8 -1
@@ -185,8 +185,15 @@ static int test__attr(struct test_suite *test __maybe_unused, int subtest __maybchar path_dir[PATH_MAX ];char *exec_path; -if (perf_pmus__has_hybrid ()) +if (perf_pmus__num_core_pmus () >1 ) { +/* +* TODO: Attribute tests hard code the PMU type. If there are >1 +* core PMU then each PMU will have a different type whic +* requires additional support. +*/ +pr_debug ("Skip test on hybrid systems" );return TEST_SKIP ; + }/* First try development tree tests. */ if (!lstat ("./tests" , &st))
tools/perf/tests/parse-metric.c +2 -5
@@ -302,11 +302,8 @@ static int test__parse_metric(struct test_suite *test __maybe_unused, int subtesTEST_ASSERT_VAL ("DCache_L2 failed" ,test_dcache_l2 () ==0 );TEST_ASSERT_VAL ("recursion fail failed" ,test_recursion_fail () ==0 );TEST_ASSERT_VAL ("Memory bandwidth" ,test_memory_bandwidth () ==0 ); - -if (!perf_pmus__has_hybrid ()) { -TEST_ASSERT_VAL ("cache_miss_cycles failed" ,test_cache_miss_cycles () ==0 ); -TEST_ASSERT_VAL ("test metric group" ,test_metric_group () ==0 ); - } +TEST_ASSERT_VAL ("cache_miss_cycles failed" ,test_cache_miss_cycles () ==0 ); +TEST_ASSERT_VAL ("test metric group" ,test_metric_group () ==0 );return 0 ; }
tools/perf/tests/switch-tracking.c +1 -11
@@ -375,17 +375,7 @@ static int test__switch_tracking(struct test_suite *test __maybe_unused, int sub cpu_clocks_evsel =evlist__last (evlist);/* Second event */ -if (perf_pmus__has_hybrid ()) { - cycles ="cpu_core/cycles/u" ; - err =parse_event (evlist, cycles); -if (err) { - cycles ="cpu_atom/cycles/u" ; -pr_debug ("Trying %s\n" , cycles); - err =parse_event (evlist, cycles); - } - }else { - err =parse_event (evlist, cycles); - } + err =parse_event (evlist, cycles);if (err) {pr_debug ("Failed to parse event %s\n" , cycles); goto out_err;
tools/perf/tests/topology.c +2 -12
@@ -41,18 +41,8 @@ static int session_write_header(char *path) session =perf_session__new (&data,NULL );TEST_ASSERT_VAL ("can't get session" , !IS_ERR (session)); -if (!perf_pmus__has_hybrid ()) { - session->evlist =evlist__new_default (); -TEST_ASSERT_VAL ("can't get evlist" , session->evlist); - }else { -struct parse_events_error err; - - session->evlist =evlist__new (); -TEST_ASSERT_VAL ("can't get evlist" , session->evlist); -parse_events_error__init (&err); -parse_events (session->evlist,"cpu_core/cycles/" , &err); -parse_events_error__exit (&err); - } + session->evlist =evlist__new_default (); +TEST_ASSERT_VAL ("can't get evlist" , session->evlist);perf_header__set_feat (&session->header,HEADER_CPU_TOPOLOGY );perf_header__set_feat (&session->header,HEADER_NRCPUS );
tools/perf/util/cputopo.c +2 -8
@@ -472,15 +472,9 @@ struct hybrid_topology *hybrid_topology__new(void) {struct perf_pmu *pmu =NULL ;struct hybrid_topology *tp =NULL ; -u32 nr =0 , i =0 ; +int nr =perf_pmus__num_core_pmus (), i =0 ; -if (!perf_pmus__has_hybrid ()) -return NULL ; - -while ((pmu =perf_pmus__scan_core (pmu)) !=NULL ) - nr++; - -if (nr ==0 ) +if (nr <=1 )return NULL ; tp =zalloc (sizeof (*tp) +sizeof (tp->nodes[0 ]) * nr);
tools/perf/util/evsel.c +1 -1
@@ -3140,7 +3140,7 @@ void evsel__zero_per_pkg(struct evsel *evsel)*/ bool evsel__is_hybrid (const struct evsel *evsel) { -if (!perf_pmus__has_hybrid ()) +if (perf_pmus__num_core_pmus () ==1 )return false;return evsel->core.is_pmu_core;
tools/perf/util/header.c +1 -1
@@ -1589,7 +1589,7 @@ static int write_pmu_caps(struct feat_fd *ff,* Write hybrid pmu caps first to maintain compatibility with * older perf tool. */ -if (perf_pmus__has_hybrid ()) { +if (perf_pmus__num_core_pmus () >1 ) { pmu =NULL ;while ((pmu =perf_pmus__scan_core (pmu))) { ret =__write_pmu_caps (ff, pmu, true);
tools/perf/util/mem-events.c +5 -13
@@ -121,6 +121,7 @@ int perf_mem_events__init(void)for (j =0 ; j <PERF_MEM_EVENTS__MAX ; j++) {struct perf_mem_event *e =perf_mem_events__ptr (j);char sysfs_name[100 ]; +struct perf_pmu *pmu =NULL ;/* * If the event entry isn't valid, skip initialization
@@ -129,18 +130,9 @@ int perf_mem_events__init(void)if (!e->tag)continue ; -if (!perf_pmus__has_hybrid ()) { -scnprintf (sysfs_name,sizeof (sysfs_name), - e->sysfs_name,"cpu" ); - e->supported =perf_mem_event__supported (mnt, sysfs_name); - }else { -struct perf_pmu *pmu =NULL ; - -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); - } +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); }if (e->supported)
@@ -196,7 +188,7 @@ int perf_mem_events__record_args(const char **rec_argv, int *argv_nr,if (!e->record)continue ; -if (!perf_pmus__has_hybrid ()) { +if (perf_pmus__num_core_pmus () ==1 ) {if (!e->supported) {pr_err ("failed: event '%s' not supported\n" ,perf_mem_events__name (j,NULL ));
tools/perf/util/metricgroup.c +1 -1
@@ -274,7 +274,7 @@ static int setup_metric_events(const char *pmu, struct hashmap *ids,const char *metric_id;struct evsel *ev;size_t ids_size, matched_events, i; -bool all_pmus = !strcmp (pmu,"all" ) || !perf_pmus__has_hybrid () || !is_pmu_hybrid (pmu); +bool all_pmus = !strcmp (pmu,"all" ) ||perf_pmus__num_core_pmus () ==1 || !is_pmu_core (pmu); *out_metric_events =NULL ; ids_size =hashmap__size (ids);
tools/perf/util/pmus.c +0 -18
@@ -464,24 +464,6 @@ bool perf_pmus__have_event(const char *pname, const char *name)return pmu &&perf_pmu__have_event (pmu, name); } -bool perf_pmus__has_hybrid (void ) -{ -static bool hybrid_scanned, has_hybrid; - -if (!hybrid_scanned) { -struct perf_pmu *pmu =NULL ; - -while ((pmu =perf_pmus__scan_core (pmu)) !=NULL ) { -if (is_pmu_hybrid (pmu->name)) { - has_hybrid = true; -break ; - } - } - hybrid_scanned = true; - } -return has_hybrid; -} -int perf_pmus__num_core_pmus (void ) {static int count;
tools/perf/util/pmus.h +0 -1
@@ -18,7 +18,6 @@ const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str);int perf_pmus__num_mem_pmus (void );void perf_pmus__print_pmu_events (const struct print_callbacks *print_cb,void *print_state);bool perf_pmus__have_event (const char *pname,const char *name); -bool perf_pmus__has_hybrid (void );int perf_pmus__num_core_pmus (void );#endif /* __PMUS_H */
tools/perf/util/stat-display.c +1 -1
@@ -681,7 +681,7 @@ static bool evlist__has_hybrid(struct evlist *evlist) {struct evsel *evsel; -if (!perf_pmus__has_hybrid ()) +if (perf_pmus__num_core_pmus () ==1 )return false;evlist__for_each_entry (evlist, evsel) {
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog
Ian Rogers
reviewerPATCH v4 35/35 perf pmu: Remove is_pmu_hybrid
In reply to Ian RogersUsers have been removed or switched to using pmu->is_core with perf_pmus__num_core_pmus() > 1.
--- tools/perf/util/pmu.c | 7 +------ tools/perf/util/pmu.h | 1 - 2 files changed, 1 insertion(+), 7 deletions(-)
tools/perf/util/pmu.c +1 -6
@@ -1417,11 +1417,6 @@ bool is_pmu_core(const char *name)return !strcmp (name,"cpu" ) ||is_sysfs_pmu_core (name); } -bool is_pmu_hybrid (const char *name) -{ -return !strcmp (name,"cpu_atom" ) || !strcmp (name,"cpu_core" ); -} -bool perf_pmu__supports_legacy_cache (const struct perf_pmu *pmu) {return pmu->is_core;
@@ -1429,7 +1424,7 @@ bool perf_pmu__supports_legacy_cache(const struct perf_pmu *pmu)bool perf_pmu__auto_merge_stats (const struct perf_pmu *pmu) { -return !is_pmu_hybrid (pmu->name); +return pmu->is_core &&perf_pmus__num_core_pmus () >1 ; }bool perf_pmu__have_event (const struct perf_pmu *pmu,const char *name)
tools/perf/util/pmu.h +0 -1
@@ -221,7 +221,6 @@ int perf_pmu__format_parse(int dirfd, struct list_head *head);void perf_pmu__del_formats (struct list_head *formats);bool is_pmu_core (const char *name); -bool is_pmu_hybrid (const char *name);bool perf_pmu__supports_legacy_cache (const struct perf_pmu *pmu);bool perf_pmu__auto_merge_stats (const struct perf_pmu *pmu);bool perf_pmu__have_event (const struct perf_pmu *pmu,const char *name);
Signature · 1 line
2.41.0.rc0.172.g3f132b7071-goog