GitHub-Mirror/Star64_linux
2
0
Fork 0
mirror of https://github.com/Fishwaldo/Star64_linux.git synced 2025-06-23 15:11:16 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

perf stat: Introduce struct runtime_stat_data

Browse source 
To pass more info to the saved_value in the runtime_stat, add a new
struct runtime_stat_data.  Currently it only has 'ctx' field but later
patch will add more.

Note that we intentionally pass 0 as ctx to clock-related events for
compatibility.  It was already there in a few places.  So move the code
into the saved_value_lookup() explicitly and add a comment.

Suggested-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20210115071139.257042-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This commit is contained in:
Namhyung Kim 2021-01-15 16:11:38 +09:00 • committed by Arnaldo Carvalho de Melo
parent 66dd86b2a2
commit 3ff1e7180a
1 changed files with 173 additions and 173 deletions
Download patch file Download diff file Expand all files Collapse all files

346
tools/perf/util/stat-shadow.c
View file

@ -114,6 +114,10 @@ static struct saved_value *saved_value_lookup(struct evsel *evsel,
rblist = &st->value_list; rblist = &st->value_list;
/* don't use context info for clock events */
if (type == STAT_NSECS)
dm.ctx = 0;
nd = rblist__find(rblist, &dm); nd = rblist__find(rblist, &dm);
if (nd) if (nd)
return container_of(nd, struct saved_value, rb_node); return container_of(nd, struct saved_value, rb_node);
@ -191,12 +195,17 @@ void perf_stat__reset_shadow_per_stat(struct runtime_stat *st)
reset_stat(st); reset_stat(st);
} }
struct runtime_stat_data {
int ctx;
};
static void update_runtime_stat(struct runtime_stat *st, static void update_runtime_stat(struct runtime_stat *st,
enum stat_type type, enum stat_type type,
int ctx, int cpu, u64 count) int cpu, u64 count,
struct runtime_stat_data *rsd)
{ {
struct saved_value *v = saved_value_lookup(NULL, cpu, true, struct saved_value *v = saved_value_lookup(NULL, cpu, true, type,
type, ctx, st); rsd->ctx, st);
if (v) if (v)
update_stats(&v->stats, count); update_stats(&v->stats, count);
@ -210,73 +219,75 @@ static void update_runtime_stat(struct runtime_stat *st,
void perf_stat__update_shadow_stats(struct evsel *counter, u64 count, void perf_stat__update_shadow_stats(struct evsel *counter, u64 count,
int cpu, struct runtime_stat *st) int cpu, struct runtime_stat *st)
{ {
int ctx = evsel_context(counter);
u64 count_ns = count; u64 count_ns = count;
struct saved_value *v; struct saved_value *v;
struct runtime_stat_data rsd = {
.ctx = evsel_context(counter),
};
count *= counter->scale; count *= counter->scale;
if (evsel__is_clock(counter)) if (evsel__is_clock(counter))
update_runtime_stat(st, STAT_NSECS, 0, cpu, count_ns); update_runtime_stat(st, STAT_NSECS, cpu, count_ns, &rsd);
else if (evsel__match(counter, HARDWARE, HW_CPU_CYCLES)) else if (evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
update_runtime_stat(st, STAT_CYCLES, ctx, cpu, count); update_runtime_stat(st, STAT_CYCLES, cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, CYCLES_IN_TX)) else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
update_runtime_stat(st, STAT_CYCLES_IN_TX, ctx, cpu, count); update_runtime_stat(st, STAT_CYCLES_IN_TX, cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TRANSACTION_START)) else if (perf_stat_evsel__is(counter, TRANSACTION_START))
update_runtime_stat(st, STAT_TRANSACTION, ctx, cpu, count); update_runtime_stat(st, STAT_TRANSACTION, cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, ELISION_START)) else if (perf_stat_evsel__is(counter, ELISION_START))
update_runtime_stat(st, STAT_ELISION, ctx, cpu, count); update_runtime_stat(st, STAT_ELISION, cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_TOTAL_SLOTS)) else if (perf_stat_evsel__is(counter, TOPDOWN_TOTAL_SLOTS))
update_runtime_stat(st, STAT_TOPDOWN_TOTAL_SLOTS, update_runtime_stat(st, STAT_TOPDOWN_TOTAL_SLOTS,
ctx, cpu, count); cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_SLOTS_ISSUED)) else if (perf_stat_evsel__is(counter, TOPDOWN_SLOTS_ISSUED))
update_runtime_stat(st, STAT_TOPDOWN_SLOTS_ISSUED, update_runtime_stat(st, STAT_TOPDOWN_SLOTS_ISSUED,
ctx, cpu, count); cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_SLOTS_RETIRED)) else if (perf_stat_evsel__is(counter, TOPDOWN_SLOTS_RETIRED))
update_runtime_stat(st, STAT_TOPDOWN_SLOTS_RETIRED, update_runtime_stat(st, STAT_TOPDOWN_SLOTS_RETIRED,
ctx, cpu, count); cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_FETCH_BUBBLES)) else if (perf_stat_evsel__is(counter, TOPDOWN_FETCH_BUBBLES))
update_runtime_stat(st, STAT_TOPDOWN_FETCH_BUBBLES, update_runtime_stat(st, STAT_TOPDOWN_FETCH_BUBBLES,
ctx, cpu, count); cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_RECOVERY_BUBBLES)) else if (perf_stat_evsel__is(counter, TOPDOWN_RECOVERY_BUBBLES))
update_runtime_stat(st, STAT_TOPDOWN_RECOVERY_BUBBLES, update_runtime_stat(st, STAT_TOPDOWN_RECOVERY_BUBBLES,
ctx, cpu, count); cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_RETIRING)) else if (perf_stat_evsel__is(counter, TOPDOWN_RETIRING))
update_runtime_stat(st, STAT_TOPDOWN_RETIRING, update_runtime_stat(st, STAT_TOPDOWN_RETIRING,
ctx, cpu, count); cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_BAD_SPEC)) else if (perf_stat_evsel__is(counter, TOPDOWN_BAD_SPEC))
update_runtime_stat(st, STAT_TOPDOWN_BAD_SPEC, update_runtime_stat(st, STAT_TOPDOWN_BAD_SPEC,
ctx, cpu, count); cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_FE_BOUND)) else if (perf_stat_evsel__is(counter, TOPDOWN_FE_BOUND))
update_runtime_stat(st, STAT_TOPDOWN_FE_BOUND, update_runtime_stat(st, STAT_TOPDOWN_FE_BOUND,
ctx, cpu, count); cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, TOPDOWN_BE_BOUND)) else if (perf_stat_evsel__is(counter, TOPDOWN_BE_BOUND))
update_runtime_stat(st, STAT_TOPDOWN_BE_BOUND, update_runtime_stat(st, STAT_TOPDOWN_BE_BOUND,
ctx, cpu, count); cpu, count, &rsd);
else if (evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) else if (evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
update_runtime_stat(st, STAT_STALLED_CYCLES_FRONT, update_runtime_stat(st, STAT_STALLED_CYCLES_FRONT,
ctx, cpu, count); cpu, count, &rsd);
else if (evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND)) else if (evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
update_runtime_stat(st, STAT_STALLED_CYCLES_BACK, update_runtime_stat(st, STAT_STALLED_CYCLES_BACK,
ctx, cpu, count); cpu, count, &rsd);
else if (evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS)) else if (evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
update_runtime_stat(st, STAT_BRANCHES, ctx, cpu, count); update_runtime_stat(st, STAT_BRANCHES, cpu, count, &rsd);
else if (evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES)) else if (evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
update_runtime_stat(st, STAT_CACHEREFS, ctx, cpu, count); update_runtime_stat(st, STAT_CACHEREFS, cpu, count, &rsd);
else if (evsel__match(counter, HW_CACHE, HW_CACHE_L1D)) else if (evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
update_runtime_stat(st, STAT_L1_DCACHE, ctx, cpu, count); update_runtime_stat(st, STAT_L1_DCACHE, cpu, count, &rsd);
else if (evsel__match(counter, HW_CACHE, HW_CACHE_L1I)) else if (evsel__match(counter, HW_CACHE, HW_CACHE_L1I))
update_runtime_stat(st, STAT_L1_ICACHE, ctx, cpu, count); update_runtime_stat(st, STAT_L1_ICACHE, cpu, count, &rsd);
else if (evsel__match(counter, HW_CACHE, HW_CACHE_LL)) else if (evsel__match(counter, HW_CACHE, HW_CACHE_LL))
update_runtime_stat(st, STAT_LL_CACHE, ctx, cpu, count); update_runtime_stat(st, STAT_LL_CACHE, cpu, count, &rsd);
else if (evsel__match(counter, HW_CACHE, HW_CACHE_DTLB)) else if (evsel__match(counter, HW_CACHE, HW_CACHE_DTLB))
update_runtime_stat(st, STAT_DTLB_CACHE, ctx, cpu, count); update_runtime_stat(st, STAT_DTLB_CACHE, cpu, count, &rsd);
else if (evsel__match(counter, HW_CACHE, HW_CACHE_ITLB)) else if (evsel__match(counter, HW_CACHE, HW_CACHE_ITLB))
update_runtime_stat(st, STAT_ITLB_CACHE, ctx, cpu, count); update_runtime_stat(st, STAT_ITLB_CACHE, cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, SMI_NUM)) else if (perf_stat_evsel__is(counter, SMI_NUM))
update_runtime_stat(st, STAT_SMI_NUM, ctx, cpu, count); update_runtime_stat(st, STAT_SMI_NUM, cpu, count, &rsd);
else if (perf_stat_evsel__is(counter, APERF)) else if (perf_stat_evsel__is(counter, APERF))
update_runtime_stat(st, STAT_APERF, ctx, cpu, count); update_runtime_stat(st, STAT_APERF, cpu, count, &rsd);
if (counter->collect_stat) { if (counter->collect_stat) {
v = saved_value_lookup(counter, cpu, true, STAT_NONE, 0, st); v = saved_value_lookup(counter, cpu, true, STAT_NONE, 0, st);
@ -422,11 +433,12 @@ void perf_stat__collect_metric_expr(struct evlist *evsel_list)
} }
static double runtime_stat_avg(struct runtime_stat *st, static double runtime_stat_avg(struct runtime_stat *st,
enum stat_type type, int ctx, int cpu) enum stat_type type, int cpu,
struct runtime_stat_data *rsd)
{ {
struct saved_value *v; struct saved_value *v;
v = saved_value_lookup(NULL, cpu, false, type, ctx, st); v = saved_value_lookup(NULL, cpu, false, type, rsd->ctx, st);
if (!v) if (!v)
return 0.0; return 0.0;
@ -434,11 +446,12 @@ static double runtime_stat_avg(struct runtime_stat *st,
} }
static double runtime_stat_n(struct runtime_stat *st, static double runtime_stat_n(struct runtime_stat *st,
enum stat_type type, int ctx, int cpu) enum stat_type type, int cpu,
struct runtime_stat_data *rsd)
{ {
struct saved_value *v; struct saved_value *v;
v = saved_value_lookup(NULL, cpu, false, type, ctx, st); v = saved_value_lookup(NULL, cpu, false, type, rsd->ctx, st);
if (!v) if (!v)
return 0.0; return 0.0;
@ -446,16 +459,15 @@ static double runtime_stat_n(struct runtime_stat *st,
} }
static void print_stalled_cycles_frontend(struct perf_stat_config *config, static void print_stalled_cycles_frontend(struct perf_stat_config *config,
int cpu, int cpu, double avg,
struct evsel *evsel, double avg,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double total, ratio = 0.0; double total, ratio = 0.0;
const char *color; const char *color;
int ctx = evsel_context(evsel);
total = runtime_stat_avg(st, STAT_CYCLES, ctx, cpu); total = runtime_stat_avg(st, STAT_CYCLES, cpu, rsd);
if (total) if (total)
ratio = avg / total * 100.0; ratio = avg / total * 100.0;
@ -470,16 +482,15 @@ static void print_stalled_cycles_frontend(struct perf_stat_config *config,
} }
static void print_stalled_cycles_backend(struct perf_stat_config *config, static void print_stalled_cycles_backend(struct perf_stat_config *config,
int cpu, int cpu, double avg,
struct evsel *evsel, double avg,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double total, ratio = 0.0; double total, ratio = 0.0;
const char *color; const char *color;
int ctx = evsel_context(evsel);
total = runtime_stat_avg(st, STAT_CYCLES, ctx, cpu); total = runtime_stat_avg(st, STAT_CYCLES, cpu, rsd);
if (total) if (total)
ratio = avg / total * 100.0; ratio = avg / total * 100.0;
@ -490,17 +501,15 @@ static void print_stalled_cycles_backend(struct perf_stat_config *config,
} }
static void print_branch_misses(struct perf_stat_config *config, static void print_branch_misses(struct perf_stat_config *config,
int cpu, int cpu, double avg,
struct evsel *evsel,
double avg,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double total, ratio = 0.0; double total, ratio = 0.0;
const char *color; const char *color;
int ctx = evsel_context(evsel);
total = runtime_stat_avg(st, STAT_BRANCHES, ctx, cpu); total = runtime_stat_avg(st, STAT_BRANCHES, cpu, rsd);
if (total) if (total)
ratio = avg / total * 100.0; ratio = avg / total * 100.0;
@ -511,18 +520,15 @@ static void print_branch_misses(struct perf_stat_config *config,
} }
static void print_l1_dcache_misses(struct perf_stat_config *config, static void print_l1_dcache_misses(struct perf_stat_config *config,
int cpu, int cpu, double avg,
struct evsel *evsel,
double avg,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double total, ratio = 0.0; double total, ratio = 0.0;
const char *color; const char *color;
int ctx = evsel_context(evsel);
total = runtime_stat_avg(st, STAT_L1_DCACHE, ctx, cpu); total = runtime_stat_avg(st, STAT_L1_DCACHE, cpu, rsd);
if (total) if (total)
ratio = avg / total * 100.0; ratio = avg / total * 100.0;
@ -533,18 +539,15 @@ static void print_l1_dcache_misses(struct perf_stat_config *config,
} }
static void print_l1_icache_misses(struct perf_stat_config *config, static void print_l1_icache_misses(struct perf_stat_config *config,
int cpu, int cpu, double avg,
struct evsel *evsel,
double avg,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double total, ratio = 0.0; double total, ratio = 0.0;
const char *color; const char *color;
int ctx = evsel_context(evsel);
total = runtime_stat_avg(st, STAT_L1_ICACHE, ctx, cpu); total = runtime_stat_avg(st, STAT_L1_ICACHE, cpu, rsd);
if (total) if (total)
ratio = avg / total * 100.0; ratio = avg / total * 100.0;
@ -554,17 +557,15 @@ static void print_l1_icache_misses(struct perf_stat_config *config,
} }
static void print_dtlb_cache_misses(struct perf_stat_config *config, static void print_dtlb_cache_misses(struct perf_stat_config *config,
int cpu, int cpu, double avg,
struct evsel *evsel,
double avg,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double total, ratio = 0.0; double total, ratio = 0.0;
const char *color; const char *color;
int ctx = evsel_context(evsel);
total = runtime_stat_avg(st, STAT_DTLB_CACHE, ctx, cpu); total = runtime_stat_avg(st, STAT_DTLB_CACHE, cpu, rsd);
if (total) if (total)
ratio = avg / total * 100.0; ratio = avg / total * 100.0;
@ -574,17 +575,15 @@ static void print_dtlb_cache_misses(struct perf_stat_config *config,
} }
static void print_itlb_cache_misses(struct perf_stat_config *config, static void print_itlb_cache_misses(struct perf_stat_config *config,
int cpu, int cpu, double avg,
struct evsel *evsel,
double avg,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double total, ratio = 0.0; double total, ratio = 0.0;
const char *color; const char *color;
int ctx = evsel_context(evsel);
total = runtime_stat_avg(st, STAT_ITLB_CACHE, ctx, cpu); total = runtime_stat_avg(st, STAT_ITLB_CACHE, cpu, rsd);
if (total) if (total)
ratio = avg / total * 100.0; ratio = avg / total * 100.0;
@ -594,17 +593,15 @@ static void print_itlb_cache_misses(struct perf_stat_config *config,
} }
static void print_ll_cache_misses(struct perf_stat_config *config, static void print_ll_cache_misses(struct perf_stat_config *config,
int cpu, int cpu, double avg,
struct evsel *evsel,
double avg,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double total, ratio = 0.0; double total, ratio = 0.0;
const char *color; const char *color;
int ctx = evsel_context(evsel);
total = runtime_stat_avg(st, STAT_LL_CACHE, ctx, cpu); total = runtime_stat_avg(st, STAT_LL_CACHE, cpu, rsd);
if (total) if (total)
ratio = avg / total * 100.0; ratio = avg / total * 100.0;
@ -662,56 +659,61 @@ static double sanitize_val(double x)
return x; return x;
} }
static double td_total_slots(int ctx, int cpu, struct runtime_stat *st) static double td_total_slots(int cpu, struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
return runtime_stat_avg(st, STAT_TOPDOWN_TOTAL_SLOTS, ctx, cpu); return runtime_stat_avg(st, STAT_TOPDOWN_TOTAL_SLOTS, cpu, rsd);
} }
static double td_bad_spec(int ctx, int cpu, struct runtime_stat *st) static double td_bad_spec(int cpu, struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double bad_spec = 0; double bad_spec = 0;
double total_slots; double total_slots;
double total; double total;
total = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_ISSUED, ctx, cpu) - total = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_ISSUED, cpu, rsd) -
runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED, ctx, cpu) + runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED, cpu, rsd) +
runtime_stat_avg(st, STAT_TOPDOWN_RECOVERY_BUBBLES, ctx, cpu); runtime_stat_avg(st, STAT_TOPDOWN_RECOVERY_BUBBLES, cpu, rsd);
total_slots = td_total_slots(ctx, cpu, st); total_slots = td_total_slots(cpu, st, rsd);
if (total_slots) if (total_slots)
bad_spec = total / total_slots; bad_spec = total / total_slots;
return sanitize_val(bad_spec); return sanitize_val(bad_spec);
} }
static double td_retiring(int ctx, int cpu, struct runtime_stat *st) static double td_retiring(int cpu, struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double retiring = 0; double retiring = 0;
double total_slots = td_total_slots(ctx, cpu, st); double total_slots = td_total_slots(cpu, st, rsd);
double ret_slots = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED, double ret_slots = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED,
ctx, cpu); cpu, rsd);
if (total_slots) if (total_slots)
retiring = ret_slots / total_slots; retiring = ret_slots / total_slots;
return retiring; return retiring;
} }
static double td_fe_bound(int ctx, int cpu, struct runtime_stat *st) static double td_fe_bound(int cpu, struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double fe_bound = 0; double fe_bound = 0;
double total_slots = td_total_slots(ctx, cpu, st); double total_slots = td_total_slots(cpu, st, rsd);
double fetch_bub = runtime_stat_avg(st, STAT_TOPDOWN_FETCH_BUBBLES, double fetch_bub = runtime_stat_avg(st, STAT_TOPDOWN_FETCH_BUBBLES,
ctx, cpu); cpu, rsd);
if (total_slots) if (total_slots)
fe_bound = fetch_bub / total_slots; fe_bound = fetch_bub / total_slots;
return fe_bound; return fe_bound;
} }
static double td_be_bound(int ctx, int cpu, struct runtime_stat *st) static double td_be_bound(int cpu, struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double sum = (td_fe_bound(ctx, cpu, st) + double sum = (td_fe_bound(cpu, st, rsd) +
td_bad_spec(ctx, cpu, st) + td_bad_spec(cpu, st, rsd) +
td_retiring(ctx, cpu, st)); td_retiring(cpu, st, rsd));
if (sum == 0) if (sum == 0)
return 0; return 0;
return sanitize_val(1.0 - sum); return sanitize_val(1.0 - sum);
@ -722,15 +724,15 @@ static double td_be_bound(int ctx, int cpu, struct runtime_stat *st)
* the ratios we need to recreate the sum. * the ratios we need to recreate the sum.
*/ */
static double td_metric_ratio(int ctx, int cpu, static double td_metric_ratio(int cpu, enum stat_type type,
enum stat_type type, struct runtime_stat *stat,
struct runtime_stat *stat) struct runtime_stat_data *rsd)
{ {
double sum = runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, ctx, cpu) + double sum = runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, cpu, rsd) +
runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, ctx, cpu) + runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, cpu, rsd) +
runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, ctx, cpu) + runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, cpu, rsd) +
runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, ctx, cpu); runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, cpu, rsd);
double d = runtime_stat_avg(stat, type, ctx, cpu); double d = runtime_stat_avg(stat, type, cpu, rsd);
if (sum) if (sum)
return d / sum; return d / sum;
@ -742,34 +744,33 @@ static double td_metric_ratio(int ctx, int cpu,
* We allow two missing. * We allow two missing.
*/ */
static bool full_td(int ctx, int cpu, static bool full_td(int cpu, struct runtime_stat *stat,
struct runtime_stat *stat) struct runtime_stat_data *rsd)
{ {
int c = 0; int c = 0;
if (runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, ctx, cpu) > 0) if (runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, cpu, rsd) > 0)
c++; c++;
if (runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, ctx, cpu) > 0) if (runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, cpu, rsd) > 0)
c++; c++;
if (runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, ctx, cpu) > 0) if (runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, cpu, rsd) > 0)
c++; c++;
if (runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, ctx, cpu) > 0) if (runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, cpu, rsd) > 0)
c++; c++;
return c >= 2; return c >= 2;
} }
static void print_smi_cost(struct perf_stat_config *config, static void print_smi_cost(struct perf_stat_config *config, int cpu,
int cpu, struct evsel *evsel,
struct perf_stat_output_ctx *out, struct perf_stat_output_ctx *out,
struct runtime_stat *st) struct runtime_stat *st,
struct runtime_stat_data *rsd)
{ {
double smi_num, aperf, cycles, cost = 0.0; double smi_num, aperf, cycles, cost = 0.0;
int ctx = evsel_context(evsel);
const char *color = NULL; const char *color = NULL;
smi_num = runtime_stat_avg(st, STAT_SMI_NUM, ctx, cpu); smi_num = runtime_stat_avg(st, STAT_SMI_NUM, cpu, rsd);
aperf = runtime_stat_avg(st, STAT_APERF, ctx, cpu); aperf = runtime_stat_avg(st, STAT_APERF, cpu, rsd);
cycles = runtime_stat_avg(st, STAT_CYCLES, ctx, cpu); cycles = runtime_stat_avg(st, STAT_CYCLES, cpu, rsd);
if ((cycles == 0) || (aperf == 0)) if ((cycles == 0) || (aperf == 0))
return; return;
@ -930,12 +931,14 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
print_metric_t print_metric = out->print_metric; print_metric_t print_metric = out->print_metric;
double total, ratio = 0.0, total2; double total, ratio = 0.0, total2;
const char *color = NULL; const char *color = NULL;
int ctx = evsel_context(evsel); struct runtime_stat_data rsd = {
.ctx = evsel_context(evsel),
};
struct metric_event *me; struct metric_event *me;
int num = 1; int num = 1;
if (evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) { if (evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
total = runtime_stat_avg(st, STAT_CYCLES, ctx, cpu); total = runtime_stat_avg(st, STAT_CYCLES, cpu, &rsd);
if (total) { if (total) {
ratio = avg / total; ratio = avg / total;
@ -945,12 +948,11 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
print_metric(config, ctxp, NULL, NULL, "insn per cycle", 0); print_metric(config, ctxp, NULL, NULL, "insn per cycle", 0);
} }
total = runtime_stat_avg(st, STAT_STALLED_CYCLES_FRONT, total = runtime_stat_avg(st, STAT_STALLED_CYCLES_FRONT, cpu, &rsd);
ctx, cpu);
total = max(total, runtime_stat_avg(st, total = max(total, runtime_stat_avg(st,
STAT_STALLED_CYCLES_BACK, STAT_STALLED_CYCLES_BACK,
ctx, cpu)); cpu, &rsd));
if (total && avg) { if (total && avg) {
out->new_line(config, ctxp); out->new_line(config, ctxp);
@ -960,8 +962,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
ratio); ratio);
} }
} else if (evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES)) { } else if (evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES)) {
if (runtime_stat_n(st, STAT_BRANCHES, ctx, cpu) != 0) if (runtime_stat_n(st, STAT_BRANCHES, cpu, &rsd) != 0)
print_branch_misses(config, cpu, evsel, avg, out, st); print_branch_misses(config, cpu, avg, out, st, &rsd);
else else
print_metric(config, ctxp, NULL, NULL, "of all branches", 0); print_metric(config, ctxp, NULL, NULL, "of all branches", 0);
} else if ( } else if (
@ -970,8 +972,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
if (runtime_stat_n(st, STAT_L1_DCACHE, ctx, cpu) != 0) if (runtime_stat_n(st, STAT_L1_DCACHE, cpu, &rsd) != 0)
print_l1_dcache_misses(config, cpu, evsel, avg, out, st); print_l1_dcache_misses(config, cpu, avg, out, st, &rsd);
else else
print_metric(config, ctxp, NULL, NULL, "of all L1-dcache accesses", 0); print_metric(config, ctxp, NULL, NULL, "of all L1-dcache accesses", 0);
} else if ( } else if (
@ -980,8 +982,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
if (runtime_stat_n(st, STAT_L1_ICACHE, ctx, cpu) != 0) if (runtime_stat_n(st, STAT_L1_ICACHE, cpu, &rsd) != 0)
print_l1_icache_misses(config, cpu, evsel, avg, out, st); print_l1_icache_misses(config, cpu, avg, out, st, &rsd);
else else
print_metric(config, ctxp, NULL, NULL, "of all L1-icache accesses", 0); print_metric(config, ctxp, NULL, NULL, "of all L1-icache accesses", 0);
} else if ( } else if (
@ -990,8 +992,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
if (runtime_stat_n(st, STAT_DTLB_CACHE, ctx, cpu) != 0) if (runtime_stat_n(st, STAT_DTLB_CACHE, cpu, &rsd) != 0)
print_dtlb_cache_misses(config, cpu, evsel, avg, out, st); print_dtlb_cache_misses(config, cpu, avg, out, st, &rsd);
else else
print_metric(config, ctxp, NULL, NULL, "of all dTLB cache accesses", 0); print_metric(config, ctxp, NULL, NULL, "of all dTLB cache accesses", 0);
} else if ( } else if (
@ -1000,8 +1002,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
if (runtime_stat_n(st, STAT_ITLB_CACHE, ctx, cpu) != 0) if (runtime_stat_n(st, STAT_ITLB_CACHE, cpu, &rsd) != 0)
print_itlb_cache_misses(config, cpu, evsel, avg, out, st); print_itlb_cache_misses(config, cpu, avg, out, st, &rsd);
else else
print_metric(config, ctxp, NULL, NULL, "of all iTLB cache accesses", 0); print_metric(config, ctxp, NULL, NULL, "of all iTLB cache accesses", 0);
} else if ( } else if (
@ -1010,27 +1012,27 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
((PERF_COUNT_HW_CACHE_OP_READ) << 8) | ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) { ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
if (runtime_stat_n(st, STAT_LL_CACHE, ctx, cpu) != 0) if (runtime_stat_n(st, STAT_LL_CACHE, cpu, &rsd) != 0)
print_ll_cache_misses(config, cpu, evsel, avg, out, st); print_ll_cache_misses(config, cpu, avg, out, st, &rsd);
else else
print_metric(config, ctxp, NULL, NULL, "of all LL-cache accesses", 0); print_metric(config, ctxp, NULL, NULL, "of all LL-cache accesses", 0);
} else if (evsel__match(evsel, HARDWARE, HW_CACHE_MISSES)) { } else if (evsel__match(evsel, HARDWARE, HW_CACHE_MISSES)) {
total = runtime_stat_avg(st, STAT_CACHEREFS, ctx, cpu); total = runtime_stat_avg(st, STAT_CACHEREFS, cpu, &rsd);
if (total) if (total)
ratio = avg * 100 / total; ratio = avg * 100 / total;
if (runtime_stat_n(st, STAT_CACHEREFS, ctx, cpu) != 0) if (runtime_stat_n(st, STAT_CACHEREFS, cpu, &rsd) != 0)
print_metric(config, ctxp, NULL, "%8.3f %%", print_metric(config, ctxp, NULL, "%8.3f %%",
"of all cache refs", ratio); "of all cache refs", ratio);
else else
print_metric(config, ctxp, NULL, NULL, "of all cache refs", 0); print_metric(config, ctxp, NULL, NULL, "of all cache refs", 0);
} else if (evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) { } else if (evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
print_stalled_cycles_frontend(config, cpu, evsel, avg, out, st); print_stalled_cycles_frontend(config, cpu, avg, out, st, &rsd);
} else if (evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) { } else if (evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
print_stalled_cycles_backend(config, cpu, evsel, avg, out, st); print_stalled_cycles_backend(config, cpu, avg, out, st, &rsd);
} else if (evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) { } else if (evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
total = runtime_stat_avg(st, STAT_NSECS, 0, cpu); total = runtime_stat_avg(st, STAT_NSECS, cpu, &rsd);
if (total) { if (total) {
ratio = avg / total; ratio = avg / total;
@ -1039,7 +1041,7 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
print_metric(config, ctxp, NULL, NULL, "Ghz", 0); print_metric(config, ctxp, NULL, NULL, "Ghz", 0);
} }
} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) { } else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
total = runtime_stat_avg(st, STAT_CYCLES, ctx, cpu); total = runtime_stat_avg(st, STAT_CYCLES, cpu, &rsd);
if (total) if (total)
print_metric(config, ctxp, NULL, print_metric(config, ctxp, NULL,
@ -1049,8 +1051,8 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
print_metric(config, ctxp, NULL, NULL, "transactional cycles", print_metric(config, ctxp, NULL, NULL, "transactional cycles",
0); 0);
} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) { } else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
total = runtime_stat_avg(st, STAT_CYCLES, ctx, cpu); total = runtime_stat_avg(st, STAT_CYCLES, cpu, &rsd);
total2 = runtime_stat_avg(st, STAT_CYCLES_IN_TX, ctx, cpu); total2 = runtime_stat_avg(st, STAT_CYCLES_IN_TX, cpu, &rsd);
if (total2 < avg) if (total2 < avg)
total2 = avg; total2 = avg;
@ -1060,21 +1062,19 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
else else
print_metric(config, ctxp, NULL, NULL, "aborted cycles", 0); print_metric(config, ctxp, NULL, NULL, "aborted cycles", 0);
} else if (perf_stat_evsel__is(evsel, TRANSACTION_START)) { } else if (perf_stat_evsel__is(evsel, TRANSACTION_START)) {
total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, cpu, &rsd);
ctx, cpu);
if (avg) if (avg)
ratio = total / avg; ratio = total / avg;
if (runtime_stat_n(st, STAT_CYCLES_IN_TX, ctx, cpu) != 0) if (runtime_stat_n(st, STAT_CYCLES_IN_TX, cpu, &rsd) != 0)
print_metric(config, ctxp, NULL, "%8.0f", print_metric(config, ctxp, NULL, "%8.0f",
"cycles / transaction", ratio); "cycles / transaction", ratio);
else else
print_metric(config, ctxp, NULL, NULL, "cycles / transaction", print_metric(config, ctxp, NULL, NULL, "cycles / transaction",
0); 0);
} else if (perf_stat_evsel__is(evsel, ELISION_START)) { } else if (perf_stat_evsel__is(evsel, ELISION_START)) {
total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, cpu, &rsd);
ctx, cpu);
if (avg) if (avg)
ratio = total / avg; ratio = total / avg;
@ -1087,28 +1087,28 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
else else
print_metric(config, ctxp, NULL, NULL, "CPUs utilized", 0); print_metric(config, ctxp, NULL, NULL, "CPUs utilized", 0);
} else if (perf_stat_evsel__is(evsel, TOPDOWN_FETCH_BUBBLES)) { } else if (perf_stat_evsel__is(evsel, TOPDOWN_FETCH_BUBBLES)) {
double fe_bound = td_fe_bound(ctx, cpu, st); double fe_bound = td_fe_bound(cpu, st, &rsd);
if (fe_bound > 0.2) if (fe_bound > 0.2)
color = PERF_COLOR_RED; color = PERF_COLOR_RED;
print_metric(config, ctxp, color, "%8.1f%%", "frontend bound", print_metric(config, ctxp, color, "%8.1f%%", "frontend bound",
fe_bound * 100.); fe_bound * 100.);
} else if (perf_stat_evsel__is(evsel, TOPDOWN_SLOTS_RETIRED)) { } else if (perf_stat_evsel__is(evsel, TOPDOWN_SLOTS_RETIRED)) {
double retiring = td_retiring(ctx, cpu, st); double retiring = td_retiring(cpu, st, &rsd);
if (retiring > 0.7) if (retiring > 0.7)
color = PERF_COLOR_GREEN; color = PERF_COLOR_GREEN;
print_metric(config, ctxp, color, "%8.1f%%", "retiring", print_metric(config, ctxp, color, "%8.1f%%", "retiring",
retiring * 100.); retiring * 100.);
} else if (perf_stat_evsel__is(evsel, TOPDOWN_RECOVERY_BUBBLES)) { } else if (perf_stat_evsel__is(evsel, TOPDOWN_RECOVERY_BUBBLES)) {
double bad_spec = td_bad_spec(ctx, cpu, st); double bad_spec = td_bad_spec(cpu, st, &rsd);
if (bad_spec > 0.1) if (bad_spec > 0.1)
color = PERF_COLOR_RED; color = PERF_COLOR_RED;
print_metric(config, ctxp, color, "%8.1f%%", "bad speculation", print_metric(config, ctxp, color, "%8.1f%%", "bad speculation",
bad_spec * 100.); bad_spec * 100.);
} else if (perf_stat_evsel__is(evsel, TOPDOWN_SLOTS_ISSUED)) { } else if (perf_stat_evsel__is(evsel, TOPDOWN_SLOTS_ISSUED)) {
double be_bound = td_be_bound(ctx, cpu, st); double be_bound = td_be_bound(cpu, st, &rsd);
const char *name = "backend bound"; const char *name = "backend bound";
static int have_recovery_bubbles = -1; static int have_recovery_bubbles = -1;
@ -1121,43 +1121,43 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
if (be_bound > 0.2) if (be_bound > 0.2)
color = PERF_COLOR_RED; color = PERF_COLOR_RED;
if (td_total_slots(ctx, cpu, st) > 0) if (td_total_slots(cpu, st, &rsd) > 0)
print_metric(config, ctxp, color, "%8.1f%%", name, print_metric(config, ctxp, color, "%8.1f%%", name,
be_bound * 100.); be_bound * 100.);
else else
print_metric(config, ctxp, NULL, NULL, name, 0); print_metric(config, ctxp, NULL, NULL, name, 0);
} else if (perf_stat_evsel__is(evsel, TOPDOWN_RETIRING) && } else if (perf_stat_evsel__is(evsel, TOPDOWN_RETIRING) &&
full_td(ctx, cpu, st)) { full_td(cpu, st, &rsd)) {
double retiring = td_metric_ratio(ctx, cpu, double retiring = td_metric_ratio(cpu,
STAT_TOPDOWN_RETIRING, st); STAT_TOPDOWN_RETIRING, st,
&rsd);
if (retiring > 0.7) if (retiring > 0.7)
color = PERF_COLOR_GREEN; color = PERF_COLOR_GREEN;
print_metric(config, ctxp, color, "%8.1f%%", "retiring", print_metric(config, ctxp, color, "%8.1f%%", "retiring",
retiring * 100.); retiring * 100.);
} else if (perf_stat_evsel__is(evsel, TOPDOWN_FE_BOUND) && } else if (perf_stat_evsel__is(evsel, TOPDOWN_FE_BOUND) &&
full_td(ctx, cpu, st)) { full_td(cpu, st, &rsd)) {
double fe_bound = td_metric_ratio(ctx, cpu, double fe_bound = td_metric_ratio(cpu,
STAT_TOPDOWN_FE_BOUND, st); STAT_TOPDOWN_FE_BOUND, st,
&rsd);
if (fe_bound > 0.2) if (fe_bound > 0.2)
color = PERF_COLOR_RED; color = PERF_COLOR_RED;
print_metric(config, ctxp, color, "%8.1f%%", "frontend bound", print_metric(config, ctxp, color, "%8.1f%%", "frontend bound",
fe_bound * 100.); fe_bound * 100.);
} else if (perf_stat_evsel__is(evsel, TOPDOWN_BE_BOUND) && } else if (perf_stat_evsel__is(evsel, TOPDOWN_BE_BOUND) &&
full_td(ctx, cpu, st)) { full_td(cpu, st, &rsd)) {
double be_bound = td_metric_ratio(ctx, cpu, double be_bound = td_metric_ratio(cpu,
STAT_TOPDOWN_BE_BOUND, st); STAT_TOPDOWN_BE_BOUND, st,
&rsd);
if (be_bound > 0.2) if (be_bound > 0.2)
color = PERF_COLOR_RED; color = PERF_COLOR_RED;
print_metric(config, ctxp, color, "%8.1f%%", "backend bound", print_metric(config, ctxp, color, "%8.1f%%", "backend bound",
be_bound * 100.); be_bound * 100.);
} else if (perf_stat_evsel__is(evsel, TOPDOWN_BAD_SPEC) && } else if (perf_stat_evsel__is(evsel, TOPDOWN_BAD_SPEC) &&
full_td(ctx, cpu, st)) { full_td(cpu, st, &rsd)) {
double bad_spec = td_metric_ratio(ctx, cpu, double bad_spec = td_metric_ratio(cpu,
STAT_TOPDOWN_BAD_SPEC, st); STAT_TOPDOWN_BAD_SPEC, st,
&rsd);
if (bad_spec > 0.1) if (bad_spec > 0.1)
color = PERF_COLOR_RED; color = PERF_COLOR_RED;
print_metric(config, ctxp, color, "%8.1f%%", "bad speculation", print_metric(config, ctxp, color, "%8.1f%%", "bad speculation",
@ -1165,11 +1165,11 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
} else if (evsel->metric_expr) { } else if (evsel->metric_expr) {
generic_metric(config, evsel->metric_expr, evsel->metric_events, NULL, generic_metric(config, evsel->metric_expr, evsel->metric_events, NULL,
evsel->name, evsel->metric_name, NULL, 1, cpu, out, st); evsel->name, evsel->metric_name, NULL, 1, cpu, out, st);
} else if (runtime_stat_n(st, STAT_NSECS, 0, cpu) != 0) { } else if (runtime_stat_n(st, STAT_NSECS, cpu, &rsd) != 0) {
char unit = 'M'; char unit = 'M';
char unit_buf[10]; char unit_buf[10];
total = runtime_stat_avg(st, STAT_NSECS, 0, cpu); total = runtime_stat_avg(st, STAT_NSECS, cpu, &rsd);
if (total) if (total)
ratio = 1000.0 * avg / total; ratio = 1000.0 * avg / total;
@ -1180,7 +1180,7 @@ void perf_stat__print_shadow_stats(struct perf_stat_config *config,
snprintf(unit_buf, sizeof(unit_buf), "%c/sec", unit); snprintf(unit_buf, sizeof(unit_buf), "%c/sec", unit);
print_metric(config, ctxp, NULL, "%8.3f", unit_buf, ratio); print_metric(config, ctxp, NULL, "%8.3f", unit_buf, ratio);
} else if (perf_stat_evsel__is(evsel, SMI_NUM)) { } else if (perf_stat_evsel__is(evsel, SMI_NUM)) {
print_smi_cost(config, cpu, evsel, out, st); print_smi_cost(config, cpu, out, st, &rsd);
} else { } else {
num = 0; num = 0;
} }