Re: [PATCH v4 3/4] perf-stat: introduce config stat.bpf-counter-events
> On Apr 20, 2021, at 10:31 AM, Jiri Olsa wrote: > > On Mon, Apr 19, 2021 at 01:36:48PM -0700, Song Liu wrote: > > SNIP > >> if (stat_config.initial_delay < 0) { >> @@ -784,11 +790,11 @@ static int __run_perf_stat(int argc, const char >> **argv, int run_idx) >> if (affinity__setup() < 0) >> return -1; >> >> -if (target__has_bpf()) { >> -evlist__for_each_entry(evsel_list, counter) { >> -if (bpf_counter__load(counter, )) >> -return -1; >> -} >> +evlist__for_each_entry(evsel_list, counter) { >> +if (bpf_counter__load(counter, )) >> +return -1; >> +if (!evsel__is_bpf(counter)) >> +all_counters_use_bpf = false; > > could be done in bpf_counter__load, check below: > >> } >> >> evlist__for_each_cpu (evsel_list, i, cpu) { >> diff --git a/tools/perf/util/bpf_counter.c b/tools/perf/util/bpf_counter.c >> index 5de991ab46af9..33b1888103dfa 100644 >> --- a/tools/perf/util/bpf_counter.c >> +++ b/tools/perf/util/bpf_counter.c >> @@ -790,7 +790,8 @@ int bpf_counter__load(struct evsel *evsel, struct target >> *target) >> { >> if (target->bpf_str) >> evsel->bpf_counter_ops = _program_profiler_ops; >> -else if (target->use_bpf) >> +else if (target->use_bpf || >> + evsel__match_bpf_counter_events(evsel->name)) >> evsel->bpf_counter_ops = _ops; > > with: > else > all_counters_use_bpf = false; > > I was also thinking of oving it to evlist, but it's sat specific, > so I think it's good as static.. thanks for changing the implementation Hmm... then we need to somehow make all_counters_use_bpf visible in bpf_counter.c, which won't be very clean. Also, since this is stat specific, I guess it is better to keep it inside builtin-stat.c? The runtime overhead should be minimal. Thanks, Song
Re: [PATCH v4 3/4] perf-stat: introduce config stat.bpf-counter-events
On Mon, Apr 19, 2021 at 01:36:48PM -0700, Song Liu wrote: SNIP > if (stat_config.initial_delay < 0) { > @@ -784,11 +790,11 @@ static int __run_perf_stat(int argc, const char **argv, > int run_idx) > if (affinity__setup() < 0) > return -1; > > - if (target__has_bpf()) { > - evlist__for_each_entry(evsel_list, counter) { > - if (bpf_counter__load(counter, )) > - return -1; > - } > + evlist__for_each_entry(evsel_list, counter) { > + if (bpf_counter__load(counter, )) > + return -1; > + if (!evsel__is_bpf(counter)) > + all_counters_use_bpf = false; could be done in bpf_counter__load, check below: > } > > evlist__for_each_cpu (evsel_list, i, cpu) { > diff --git a/tools/perf/util/bpf_counter.c b/tools/perf/util/bpf_counter.c > index 5de991ab46af9..33b1888103dfa 100644 > --- a/tools/perf/util/bpf_counter.c > +++ b/tools/perf/util/bpf_counter.c > @@ -790,7 +790,8 @@ int bpf_counter__load(struct evsel *evsel, struct target > *target) > { > if (target->bpf_str) > evsel->bpf_counter_ops = _program_profiler_ops; > - else if (target->use_bpf) > + else if (target->use_bpf || > + evsel__match_bpf_counter_events(evsel->name)) > evsel->bpf_counter_ops = _ops; with: else all_counters_use_bpf = false; I was also thinking of oving it to evlist, but it's sat specific, so I think it's good as static.. thanks for changing the implementation jirka > > if (evsel->bpf_counter_ops) > diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c > index 6bcb5ef221f8c..63d472b336de2 100644 > --- a/tools/perf/util/config.c > +++ b/tools/perf/util/config.c > @@ -18,6 +18,7 @@ > #include "util/hist.h" /* perf_hist_config */ > #include "util/llvm-utils.h" /* perf_llvm_config */ > #include "util/stat.h" /* perf_stat__set_big_num */ > +#include "util/evsel.h" /* evsel__hw_names, evsel__use_bpf_counters */ > #include "build-id.h" > #include "debug.h" > #include "config.h" > @@ -460,6 +461,9 @@ static int perf_stat_config(const char *var, const char > *value) > if (!strcmp(var, "stat.no-csv-summary")) > perf_stat__set_no_csv_summary(perf_config_bool(var, value)); > > + if (!strcmp(var, "stat.bpf-counter-events")) > + evsel__bpf_counter_events = strdup(value); > + > /* Add other config variables here. */ > return 0; > } SNIP
[PATCH v4 3/4] perf-stat: introduce config stat.bpf-counter-events
Currently, to use BPF to aggregate perf event counters, the user uses --bpf-counters option. Enable "use bpf by default" events with a config option, stat.bpf-counter-events. Events with name in the option will use BPF. This also enables mixed BPF event and regular event in the same sesssion. For example: perf config stat.bpf-counter-events=instructions perf stat -e instructions,cs The second command will use BPF for "instructions" but not "cs". Signed-off-by: Song Liu --- tools/perf/Documentation/perf-stat.txt | 2 ++ tools/perf/builtin-stat.c | 40 +++--- tools/perf/util/bpf_counter.c | 3 +- tools/perf/util/config.c | 4 +++ tools/perf/util/evsel.c| 22 ++ tools/perf/util/evsel.h| 8 ++ tools/perf/util/target.h | 5 7 files changed, 61 insertions(+), 23 deletions(-) diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt index 6ec5960b08c3d..f10e24da23e90 100644 --- a/tools/perf/Documentation/perf-stat.txt +++ b/tools/perf/Documentation/perf-stat.txt @@ -97,6 +97,8 @@ report:: Use BPF programs to aggregate readings from perf_events. This allows multiple perf-stat sessions that are counting the same metric (cycles, instructions, etc.) to share hardware counters. + To use BPF programs on common events by default, use + "perf config stat.bpf-counter-events=". --bpf-attr-map:: With option "--bpf-counters", different perf-stat sessions share diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 2a2c15cac80a3..157105e792eaf 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -160,6 +160,7 @@ static const char *smi_cost_attrs = { }; static struct evlist *evsel_list; +static bool all_counters_use_bpf = true; static struct target target = { .uid= UINT_MAX, @@ -399,6 +400,9 @@ static int read_affinity_counters(struct timespec *rs) struct affinity affinity; int i, ncpus, cpu; + if (all_counters_use_bpf) + return 0; + if (affinity__setup() < 0) return -1; @@ -413,6 +417,8 @@ static int read_affinity_counters(struct timespec *rs) evlist__for_each_entry(evsel_list, counter) { if (evsel__cpu_iter_skip(counter, cpu)) continue; + if (evsel__is_bpf(counter)) + continue; if (!counter->err) { counter->err = read_counter_cpu(counter, rs, counter->cpu_iter - 1); @@ -429,6 +435,9 @@ static int read_bpf_map_counters(void) int err; evlist__for_each_entry(evsel_list, counter) { + if (!evsel__is_bpf(counter)) + continue; + err = bpf_counter__read(counter); if (err) return err; @@ -439,14 +448,10 @@ static int read_bpf_map_counters(void) static void read_counters(struct timespec *rs) { struct evsel *counter; - int err; if (!stat_config.stop_read_counter) { - if (target__has_bpf()) - err = read_bpf_map_counters(); - else - err = read_affinity_counters(rs); - if (err < 0) + if (read_bpf_map_counters() || + read_affinity_counters(rs)) return; } @@ -535,12 +540,13 @@ static int enable_counters(void) struct evsel *evsel; int err; - if (target__has_bpf()) { - evlist__for_each_entry(evsel_list, evsel) { - err = bpf_counter__enable(evsel); - if (err) - return err; - } + evlist__for_each_entry(evsel_list, evsel) { + if (!evsel__is_bpf(evsel)) + continue; + + err = bpf_counter__enable(evsel); + if (err) + return err; } if (stat_config.initial_delay < 0) { @@ -784,11 +790,11 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx) if (affinity__setup() < 0) return -1; - if (target__has_bpf()) { - evlist__for_each_entry(evsel_list, counter) { - if (bpf_counter__load(counter, )) - return -1; - } + evlist__for_each_entry(evsel_list, counter) { + if (bpf_counter__load(counter, )) + return -1; + if (!evsel__is_bpf(counter)) + all_counters_use_bpf = false; } evlist__for_each_cpu (evsel_list, i, cpu) { diff --git