1 // SPDX-License-Identifier: GPL-2.0
2 #include <math.h>
3 #include <stdio.h>
4 #include "evsel.h"
5 #include "stat.h"
6 #include "color.h"
7 #include "debug.h"
8 #include "pmu.h"
9 #include "rblist.h"
10 #include "evlist.h"
11 #include "expr.h"
12 #include "metricgroup.h"
13 #include "cgroup.h"
14 #include "units.h"
15 #include <linux/zalloc.h>
16 #include "iostat.h"
17 #include "util/hashmap.h"
18
19 /*
20 * AGGR_GLOBAL: Use CPU 0
21 * AGGR_SOCKET: Use first CPU of socket
22 * AGGR_DIE: Use first CPU of die
23 * AGGR_CORE: Use first CPU of core
24 * AGGR_NONE: Use matching CPU
25 * AGGR_THREAD: Not supported?
26 */
27
28 struct runtime_stat rt_stat;
29 struct stats walltime_nsecs_stats;
30 struct rusage_stats ru_stats;
31
32 struct saved_value {
33 struct rb_node rb_node;
34 struct evsel *evsel;
35 enum stat_type type;
36 int ctx;
37 int map_idx; /* cpu or thread map index */
38 struct cgroup *cgrp;
39 struct stats stats;
40 u64 metric_total;
41 int metric_other;
42 };
43
saved_value_cmp(struct rb_node * rb_node,const void * entry)44 static int saved_value_cmp(struct rb_node *rb_node, const void *entry)
45 {
46 struct saved_value *a = container_of(rb_node,
47 struct saved_value,
48 rb_node);
49 const struct saved_value *b = entry;
50
51 if (a->map_idx != b->map_idx)
52 return a->map_idx - b->map_idx;
53
54 /*
55 * Previously the rbtree was used to link generic metrics.
56 * The keys were evsel/cpu. Now the rbtree is extended to support
57 * per-thread shadow stats. For shadow stats case, the keys
58 * are cpu/type/ctx/stat (evsel is NULL). For generic metrics
59 * case, the keys are still evsel/cpu (type/ctx/stat are 0 or NULL).
60 */
61 if (a->type != b->type)
62 return a->type - b->type;
63
64 if (a->ctx != b->ctx)
65 return a->ctx - b->ctx;
66
67 if (a->cgrp != b->cgrp)
68 return (char *)a->cgrp < (char *)b->cgrp ? -1 : +1;
69
70 if (a->evsel == b->evsel)
71 return 0;
72 if ((char *)a->evsel < (char *)b->evsel)
73 return -1;
74 return +1;
75 }
76
saved_value_new(struct rblist * rblist __maybe_unused,const void * entry)77 static struct rb_node *saved_value_new(struct rblist *rblist __maybe_unused,
78 const void *entry)
79 {
80 struct saved_value *nd = malloc(sizeof(struct saved_value));
81
82 if (!nd)
83 return NULL;
84 memcpy(nd, entry, sizeof(struct saved_value));
85 return &nd->rb_node;
86 }
87
saved_value_delete(struct rblist * rblist __maybe_unused,struct rb_node * rb_node)88 static void saved_value_delete(struct rblist *rblist __maybe_unused,
89 struct rb_node *rb_node)
90 {
91 struct saved_value *v;
92
93 BUG_ON(!rb_node);
94 v = container_of(rb_node, struct saved_value, rb_node);
95 free(v);
96 }
97
saved_value_lookup(struct evsel * evsel,int map_idx,bool create,enum stat_type type,int ctx,struct runtime_stat * st,struct cgroup * cgrp)98 static struct saved_value *saved_value_lookup(struct evsel *evsel,
99 int map_idx,
100 bool create,
101 enum stat_type type,
102 int ctx,
103 struct runtime_stat *st,
104 struct cgroup *cgrp)
105 {
106 struct rblist *rblist;
107 struct rb_node *nd;
108 struct saved_value dm = {
109 .map_idx = map_idx,
110 .evsel = evsel,
111 .type = type,
112 .ctx = ctx,
113 .cgrp = cgrp,
114 };
115
116 rblist = &st->value_list;
117
118 /* don't use context info for clock events */
119 if (type == STAT_NSECS)
120 dm.ctx = 0;
121
122 nd = rblist__find(rblist, &dm);
123 if (nd)
124 return container_of(nd, struct saved_value, rb_node);
125 if (create) {
126 rblist__add_node(rblist, &dm);
127 nd = rblist__find(rblist, &dm);
128 if (nd)
129 return container_of(nd, struct saved_value, rb_node);
130 }
131 return NULL;
132 }
133
runtime_stat__init(struct runtime_stat * st)134 void runtime_stat__init(struct runtime_stat *st)
135 {
136 struct rblist *rblist = &st->value_list;
137
138 rblist__init(rblist);
139 rblist->node_cmp = saved_value_cmp;
140 rblist->node_new = saved_value_new;
141 rblist->node_delete = saved_value_delete;
142 }
143
runtime_stat__exit(struct runtime_stat * st)144 void runtime_stat__exit(struct runtime_stat *st)
145 {
146 rblist__exit(&st->value_list);
147 }
148
perf_stat__init_shadow_stats(void)149 void perf_stat__init_shadow_stats(void)
150 {
151 runtime_stat__init(&rt_stat);
152 }
153
evsel_context(struct evsel * evsel)154 static int evsel_context(struct evsel *evsel)
155 {
156 int ctx = 0;
157
158 if (evsel->core.attr.exclude_kernel)
159 ctx |= CTX_BIT_KERNEL;
160 if (evsel->core.attr.exclude_user)
161 ctx |= CTX_BIT_USER;
162 if (evsel->core.attr.exclude_hv)
163 ctx |= CTX_BIT_HV;
164 if (evsel->core.attr.exclude_host)
165 ctx |= CTX_BIT_HOST;
166 if (evsel->core.attr.exclude_idle)
167 ctx |= CTX_BIT_IDLE;
168
169 return ctx;
170 }
171
reset_stat(struct runtime_stat * st)172 static void reset_stat(struct runtime_stat *st)
173 {
174 struct rblist *rblist;
175 struct rb_node *pos, *next;
176
177 rblist = &st->value_list;
178 next = rb_first_cached(&rblist->entries);
179 while (next) {
180 pos = next;
181 next = rb_next(pos);
182 memset(&container_of(pos, struct saved_value, rb_node)->stats,
183 0,
184 sizeof(struct stats));
185 }
186 }
187
perf_stat__reset_shadow_stats(void)188 void perf_stat__reset_shadow_stats(void)
189 {
190 reset_stat(&rt_stat);
191 memset(&walltime_nsecs_stats, 0, sizeof(walltime_nsecs_stats));
192 memset(&ru_stats, 0, sizeof(ru_stats));
193 }
194
perf_stat__reset_shadow_per_stat(struct runtime_stat * st)195 void perf_stat__reset_shadow_per_stat(struct runtime_stat *st)
196 {
197 reset_stat(st);
198 }
199
200 struct runtime_stat_data {
201 int ctx;
202 struct cgroup *cgrp;
203 };
204
update_runtime_stat(struct runtime_stat * st,enum stat_type type,int map_idx,u64 count,struct runtime_stat_data * rsd)205 static void update_runtime_stat(struct runtime_stat *st,
206 enum stat_type type,
207 int map_idx, u64 count,
208 struct runtime_stat_data *rsd)
209 {
210 struct saved_value *v = saved_value_lookup(NULL, map_idx, true, type,
211 rsd->ctx, st, rsd->cgrp);
212
213 if (v)
214 update_stats(&v->stats, count);
215 }
216
217 /*
218 * Update various tracking values we maintain to print
219 * more semantic information such as miss/hit ratios,
220 * instruction rates, etc:
221 */
perf_stat__update_shadow_stats(struct evsel * counter,u64 count,int map_idx,struct runtime_stat * st)222 void perf_stat__update_shadow_stats(struct evsel *counter, u64 count,
223 int map_idx, struct runtime_stat *st)
224 {
225 u64 count_ns = count;
226 struct saved_value *v;
227 struct runtime_stat_data rsd = {
228 .ctx = evsel_context(counter),
229 .cgrp = counter->cgrp,
230 };
231
232 count *= counter->scale;
233
234 if (evsel__is_clock(counter))
235 update_runtime_stat(st, STAT_NSECS, map_idx, count_ns, &rsd);
236 else if (evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
237 update_runtime_stat(st, STAT_CYCLES, map_idx, count, &rsd);
238 else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
239 update_runtime_stat(st, STAT_CYCLES_IN_TX, map_idx, count, &rsd);
240 else if (perf_stat_evsel__is(counter, TRANSACTION_START))
241 update_runtime_stat(st, STAT_TRANSACTION, map_idx, count, &rsd);
242 else if (perf_stat_evsel__is(counter, ELISION_START))
243 update_runtime_stat(st, STAT_ELISION, map_idx, count, &rsd);
244 else if (perf_stat_evsel__is(counter, TOPDOWN_TOTAL_SLOTS))
245 update_runtime_stat(st, STAT_TOPDOWN_TOTAL_SLOTS,
246 map_idx, count, &rsd);
247 else if (perf_stat_evsel__is(counter, TOPDOWN_SLOTS_ISSUED))
248 update_runtime_stat(st, STAT_TOPDOWN_SLOTS_ISSUED,
249 map_idx, count, &rsd);
250 else if (perf_stat_evsel__is(counter, TOPDOWN_SLOTS_RETIRED))
251 update_runtime_stat(st, STAT_TOPDOWN_SLOTS_RETIRED,
252 map_idx, count, &rsd);
253 else if (perf_stat_evsel__is(counter, TOPDOWN_FETCH_BUBBLES))
254 update_runtime_stat(st, STAT_TOPDOWN_FETCH_BUBBLES,
255 map_idx, count, &rsd);
256 else if (perf_stat_evsel__is(counter, TOPDOWN_RECOVERY_BUBBLES))
257 update_runtime_stat(st, STAT_TOPDOWN_RECOVERY_BUBBLES,
258 map_idx, count, &rsd);
259 else if (perf_stat_evsel__is(counter, TOPDOWN_RETIRING))
260 update_runtime_stat(st, STAT_TOPDOWN_RETIRING,
261 map_idx, count, &rsd);
262 else if (perf_stat_evsel__is(counter, TOPDOWN_BAD_SPEC))
263 update_runtime_stat(st, STAT_TOPDOWN_BAD_SPEC,
264 map_idx, count, &rsd);
265 else if (perf_stat_evsel__is(counter, TOPDOWN_FE_BOUND))
266 update_runtime_stat(st, STAT_TOPDOWN_FE_BOUND,
267 map_idx, count, &rsd);
268 else if (perf_stat_evsel__is(counter, TOPDOWN_BE_BOUND))
269 update_runtime_stat(st, STAT_TOPDOWN_BE_BOUND,
270 map_idx, count, &rsd);
271 else if (perf_stat_evsel__is(counter, TOPDOWN_HEAVY_OPS))
272 update_runtime_stat(st, STAT_TOPDOWN_HEAVY_OPS,
273 map_idx, count, &rsd);
274 else if (perf_stat_evsel__is(counter, TOPDOWN_BR_MISPREDICT))
275 update_runtime_stat(st, STAT_TOPDOWN_BR_MISPREDICT,
276 map_idx, count, &rsd);
277 else if (perf_stat_evsel__is(counter, TOPDOWN_FETCH_LAT))
278 update_runtime_stat(st, STAT_TOPDOWN_FETCH_LAT,
279 map_idx, count, &rsd);
280 else if (perf_stat_evsel__is(counter, TOPDOWN_MEM_BOUND))
281 update_runtime_stat(st, STAT_TOPDOWN_MEM_BOUND,
282 map_idx, count, &rsd);
283 else if (evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
284 update_runtime_stat(st, STAT_STALLED_CYCLES_FRONT,
285 map_idx, count, &rsd);
286 else if (evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
287 update_runtime_stat(st, STAT_STALLED_CYCLES_BACK,
288 map_idx, count, &rsd);
289 else if (evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
290 update_runtime_stat(st, STAT_BRANCHES, map_idx, count, &rsd);
291 else if (evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
292 update_runtime_stat(st, STAT_CACHEREFS, map_idx, count, &rsd);
293 else if (evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
294 update_runtime_stat(st, STAT_L1_DCACHE, map_idx, count, &rsd);
295 else if (evsel__match(counter, HW_CACHE, HW_CACHE_L1I))
296 update_runtime_stat(st, STAT_L1_ICACHE, map_idx, count, &rsd);
297 else if (evsel__match(counter, HW_CACHE, HW_CACHE_LL))
298 update_runtime_stat(st, STAT_LL_CACHE, map_idx, count, &rsd);
299 else if (evsel__match(counter, HW_CACHE, HW_CACHE_DTLB))
300 update_runtime_stat(st, STAT_DTLB_CACHE, map_idx, count, &rsd);
301 else if (evsel__match(counter, HW_CACHE, HW_CACHE_ITLB))
302 update_runtime_stat(st, STAT_ITLB_CACHE, map_idx, count, &rsd);
303 else if (perf_stat_evsel__is(counter, SMI_NUM))
304 update_runtime_stat(st, STAT_SMI_NUM, map_idx, count, &rsd);
305 else if (perf_stat_evsel__is(counter, APERF))
306 update_runtime_stat(st, STAT_APERF, map_idx, count, &rsd);
307
308 if (counter->collect_stat) {
309 v = saved_value_lookup(counter, map_idx, true, STAT_NONE, 0, st,
310 rsd.cgrp);
311 update_stats(&v->stats, count);
312 if (counter->metric_leader)
313 v->metric_total += count;
314 } else if (counter->metric_leader && !counter->merged_stat) {
315 v = saved_value_lookup(counter->metric_leader,
316 map_idx, true, STAT_NONE, 0, st, rsd.cgrp);
317 v->metric_total += count;
318 v->metric_other++;
319 }
320 }
321
322 /* used for get_ratio_color() */
323 enum grc_type {
324 GRC_STALLED_CYCLES_FE,
325 GRC_STALLED_CYCLES_BE,
326 GRC_CACHE_MISSES,
327 GRC_MAX_NR
328 };
329
get_ratio_color(enum grc_type type,double ratio)330 static const char *get_ratio_color(enum grc_type type, double ratio)
331 {
332 static const double grc_table[GRC_MAX_NR][3] = {
333 [GRC_STALLED_CYCLES_FE] = { 50.0, 30.0, 10.0 },
334 [GRC_STALLED_CYCLES_BE] = { 75.0, 50.0, 20.0 },
335 [GRC_CACHE_MISSES] = { 20.0, 10.0, 5.0 },
336 };
337 const char *color = PERF_COLOR_NORMAL;
338
339 if (ratio > grc_table[type][0])
340 color = PERF_COLOR_RED;
341 else if (ratio > grc_table[type][1])
342 color = PERF_COLOR_MAGENTA;
343 else if (ratio > grc_table[type][2])
344 color = PERF_COLOR_YELLOW;
345
346 return color;
347 }
348
runtime_stat_avg(struct runtime_stat * st,enum stat_type type,int map_idx,struct runtime_stat_data * rsd)349 static double runtime_stat_avg(struct runtime_stat *st,
350 enum stat_type type, int map_idx,
351 struct runtime_stat_data *rsd)
352 {
353 struct saved_value *v;
354
355 v = saved_value_lookup(NULL, map_idx, false, type, rsd->ctx, st, rsd->cgrp);
356 if (!v)
357 return 0.0;
358
359 return avg_stats(&v->stats);
360 }
361
runtime_stat_n(struct runtime_stat * st,enum stat_type type,int map_idx,struct runtime_stat_data * rsd)362 static double runtime_stat_n(struct runtime_stat *st,
363 enum stat_type type, int map_idx,
364 struct runtime_stat_data *rsd)
365 {
366 struct saved_value *v;
367
368 v = saved_value_lookup(NULL, map_idx, false, type, rsd->ctx, st, rsd->cgrp);
369 if (!v)
370 return 0.0;
371
372 return v->stats.n;
373 }
374
print_stalled_cycles_frontend(struct perf_stat_config * config,int map_idx,double avg,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)375 static void print_stalled_cycles_frontend(struct perf_stat_config *config,
376 int map_idx, double avg,
377 struct perf_stat_output_ctx *out,
378 struct runtime_stat *st,
379 struct runtime_stat_data *rsd)
380 {
381 double total, ratio = 0.0;
382 const char *color;
383
384 total = runtime_stat_avg(st, STAT_CYCLES, map_idx, rsd);
385
386 if (total)
387 ratio = avg / total * 100.0;
388
389 color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);
390
391 if (ratio)
392 out->print_metric(config, out->ctx, color, "%7.2f%%", "frontend cycles idle",
393 ratio);
394 else
395 out->print_metric(config, out->ctx, NULL, NULL, "frontend cycles idle", 0);
396 }
397
print_stalled_cycles_backend(struct perf_stat_config * config,int map_idx,double avg,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)398 static void print_stalled_cycles_backend(struct perf_stat_config *config,
399 int map_idx, double avg,
400 struct perf_stat_output_ctx *out,
401 struct runtime_stat *st,
402 struct runtime_stat_data *rsd)
403 {
404 double total, ratio = 0.0;
405 const char *color;
406
407 total = runtime_stat_avg(st, STAT_CYCLES, map_idx, rsd);
408
409 if (total)
410 ratio = avg / total * 100.0;
411
412 color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);
413
414 out->print_metric(config, out->ctx, color, "%7.2f%%", "backend cycles idle", ratio);
415 }
416
print_branch_misses(struct perf_stat_config * config,int map_idx,double avg,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)417 static void print_branch_misses(struct perf_stat_config *config,
418 int map_idx, double avg,
419 struct perf_stat_output_ctx *out,
420 struct runtime_stat *st,
421 struct runtime_stat_data *rsd)
422 {
423 double total, ratio = 0.0;
424 const char *color;
425
426 total = runtime_stat_avg(st, STAT_BRANCHES, map_idx, rsd);
427
428 if (total)
429 ratio = avg / total * 100.0;
430
431 color = get_ratio_color(GRC_CACHE_MISSES, ratio);
432
433 out->print_metric(config, out->ctx, color, "%7.2f%%", "of all branches", ratio);
434 }
435
print_l1_dcache_misses(struct perf_stat_config * config,int map_idx,double avg,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)436 static void print_l1_dcache_misses(struct perf_stat_config *config,
437 int map_idx, double avg,
438 struct perf_stat_output_ctx *out,
439 struct runtime_stat *st,
440 struct runtime_stat_data *rsd)
441 {
442 double total, ratio = 0.0;
443 const char *color;
444
445 total = runtime_stat_avg(st, STAT_L1_DCACHE, map_idx, rsd);
446
447 if (total)
448 ratio = avg / total * 100.0;
449
450 color = get_ratio_color(GRC_CACHE_MISSES, ratio);
451
452 out->print_metric(config, out->ctx, color, "%7.2f%%", "of all L1-dcache accesses", ratio);
453 }
454
print_l1_icache_misses(struct perf_stat_config * config,int map_idx,double avg,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)455 static void print_l1_icache_misses(struct perf_stat_config *config,
456 int map_idx, double avg,
457 struct perf_stat_output_ctx *out,
458 struct runtime_stat *st,
459 struct runtime_stat_data *rsd)
460 {
461 double total, ratio = 0.0;
462 const char *color;
463
464 total = runtime_stat_avg(st, STAT_L1_ICACHE, map_idx, rsd);
465
466 if (total)
467 ratio = avg / total * 100.0;
468
469 color = get_ratio_color(GRC_CACHE_MISSES, ratio);
470 out->print_metric(config, out->ctx, color, "%7.2f%%", "of all L1-icache accesses", ratio);
471 }
472
print_dtlb_cache_misses(struct perf_stat_config * config,int map_idx,double avg,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)473 static void print_dtlb_cache_misses(struct perf_stat_config *config,
474 int map_idx, double avg,
475 struct perf_stat_output_ctx *out,
476 struct runtime_stat *st,
477 struct runtime_stat_data *rsd)
478 {
479 double total, ratio = 0.0;
480 const char *color;
481
482 total = runtime_stat_avg(st, STAT_DTLB_CACHE, map_idx, rsd);
483
484 if (total)
485 ratio = avg / total * 100.0;
486
487 color = get_ratio_color(GRC_CACHE_MISSES, ratio);
488 out->print_metric(config, out->ctx, color, "%7.2f%%", "of all dTLB cache accesses", ratio);
489 }
490
print_itlb_cache_misses(struct perf_stat_config * config,int map_idx,double avg,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)491 static void print_itlb_cache_misses(struct perf_stat_config *config,
492 int map_idx, double avg,
493 struct perf_stat_output_ctx *out,
494 struct runtime_stat *st,
495 struct runtime_stat_data *rsd)
496 {
497 double total, ratio = 0.0;
498 const char *color;
499
500 total = runtime_stat_avg(st, STAT_ITLB_CACHE, map_idx, rsd);
501
502 if (total)
503 ratio = avg / total * 100.0;
504
505 color = get_ratio_color(GRC_CACHE_MISSES, ratio);
506 out->print_metric(config, out->ctx, color, "%7.2f%%", "of all iTLB cache accesses", ratio);
507 }
508
print_ll_cache_misses(struct perf_stat_config * config,int map_idx,double avg,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)509 static void print_ll_cache_misses(struct perf_stat_config *config,
510 int map_idx, double avg,
511 struct perf_stat_output_ctx *out,
512 struct runtime_stat *st,
513 struct runtime_stat_data *rsd)
514 {
515 double total, ratio = 0.0;
516 const char *color;
517
518 total = runtime_stat_avg(st, STAT_LL_CACHE, map_idx, rsd);
519
520 if (total)
521 ratio = avg / total * 100.0;
522
523 color = get_ratio_color(GRC_CACHE_MISSES, ratio);
524 out->print_metric(config, out->ctx, color, "%7.2f%%", "of all LL-cache accesses", ratio);
525 }
526
527 /*
528 * High level "TopDown" CPU core pipe line bottleneck break down.
529 *
530 * Basic concept following
531 * Yasin, A Top Down Method for Performance analysis and Counter architecture
532 * ISPASS14
533 *
534 * The CPU pipeline is divided into 4 areas that can be bottlenecks:
535 *
536 * Frontend -> Backend -> Retiring
537 * BadSpeculation in addition means out of order execution that is thrown away
538 * (for example branch mispredictions)
539 * Frontend is instruction decoding.
540 * Backend is execution, like computation and accessing data in memory
541 * Retiring is good execution that is not directly bottlenecked
542 *
543 * The formulas are computed in slots.
544 * A slot is an entry in the pipeline each for the pipeline width
545 * (for example a 4-wide pipeline has 4 slots for each cycle)
546 *
547 * Formulas:
548 * BadSpeculation = ((SlotsIssued - SlotsRetired) + RecoveryBubbles) /
549 * TotalSlots
550 * Retiring = SlotsRetired / TotalSlots
551 * FrontendBound = FetchBubbles / TotalSlots
552 * BackendBound = 1.0 - BadSpeculation - Retiring - FrontendBound
553 *
554 * The kernel provides the mapping to the low level CPU events and any scaling
555 * needed for the CPU pipeline width, for example:
556 *
557 * TotalSlots = Cycles * 4
558 *
559 * The scaling factor is communicated in the sysfs unit.
560 *
561 * In some cases the CPU may not be able to measure all the formulas due to
562 * missing events. In this case multiple formulas are combined, as possible.
563 *
564 * Full TopDown supports more levels to sub-divide each area: for example
565 * BackendBound into computing bound and memory bound. For now we only
566 * support Level 1 TopDown.
567 */
568
sanitize_val(double x)569 static double sanitize_val(double x)
570 {
571 if (x < 0 && x >= -0.02)
572 return 0.0;
573 return x;
574 }
575
td_total_slots(int map_idx,struct runtime_stat * st,struct runtime_stat_data * rsd)576 static double td_total_slots(int map_idx, struct runtime_stat *st,
577 struct runtime_stat_data *rsd)
578 {
579 return runtime_stat_avg(st, STAT_TOPDOWN_TOTAL_SLOTS, map_idx, rsd);
580 }
581
td_bad_spec(int map_idx,struct runtime_stat * st,struct runtime_stat_data * rsd)582 static double td_bad_spec(int map_idx, struct runtime_stat *st,
583 struct runtime_stat_data *rsd)
584 {
585 double bad_spec = 0;
586 double total_slots;
587 double total;
588
589 total = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_ISSUED, map_idx, rsd) -
590 runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED, map_idx, rsd) +
591 runtime_stat_avg(st, STAT_TOPDOWN_RECOVERY_BUBBLES, map_idx, rsd);
592
593 total_slots = td_total_slots(map_idx, st, rsd);
594 if (total_slots)
595 bad_spec = total / total_slots;
596 return sanitize_val(bad_spec);
597 }
598
td_retiring(int map_idx,struct runtime_stat * st,struct runtime_stat_data * rsd)599 static double td_retiring(int map_idx, struct runtime_stat *st,
600 struct runtime_stat_data *rsd)
601 {
602 double retiring = 0;
603 double total_slots = td_total_slots(map_idx, st, rsd);
604 double ret_slots = runtime_stat_avg(st, STAT_TOPDOWN_SLOTS_RETIRED,
605 map_idx, rsd);
606
607 if (total_slots)
608 retiring = ret_slots / total_slots;
609 return retiring;
610 }
611
td_fe_bound(int map_idx,struct runtime_stat * st,struct runtime_stat_data * rsd)612 static double td_fe_bound(int map_idx, struct runtime_stat *st,
613 struct runtime_stat_data *rsd)
614 {
615 double fe_bound = 0;
616 double total_slots = td_total_slots(map_idx, st, rsd);
617 double fetch_bub = runtime_stat_avg(st, STAT_TOPDOWN_FETCH_BUBBLES,
618 map_idx, rsd);
619
620 if (total_slots)
621 fe_bound = fetch_bub / total_slots;
622 return fe_bound;
623 }
624
td_be_bound(int map_idx,struct runtime_stat * st,struct runtime_stat_data * rsd)625 static double td_be_bound(int map_idx, struct runtime_stat *st,
626 struct runtime_stat_data *rsd)
627 {
628 double sum = (td_fe_bound(map_idx, st, rsd) +
629 td_bad_spec(map_idx, st, rsd) +
630 td_retiring(map_idx, st, rsd));
631 if (sum == 0)
632 return 0;
633 return sanitize_val(1.0 - sum);
634 }
635
636 /*
637 * Kernel reports metrics multiplied with slots. To get back
638 * the ratios we need to recreate the sum.
639 */
640
td_metric_ratio(int map_idx,enum stat_type type,struct runtime_stat * stat,struct runtime_stat_data * rsd)641 static double td_metric_ratio(int map_idx, enum stat_type type,
642 struct runtime_stat *stat,
643 struct runtime_stat_data *rsd)
644 {
645 double sum = runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, map_idx, rsd) +
646 runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, map_idx, rsd) +
647 runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, map_idx, rsd) +
648 runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, map_idx, rsd);
649 double d = runtime_stat_avg(stat, type, map_idx, rsd);
650
651 if (sum)
652 return d / sum;
653 return 0;
654 }
655
656 /*
657 * ... but only if most of the values are actually available.
658 * We allow two missing.
659 */
660
full_td(int map_idx,struct runtime_stat * stat,struct runtime_stat_data * rsd)661 static bool full_td(int map_idx, struct runtime_stat *stat,
662 struct runtime_stat_data *rsd)
663 {
664 int c = 0;
665
666 if (runtime_stat_avg(stat, STAT_TOPDOWN_RETIRING, map_idx, rsd) > 0)
667 c++;
668 if (runtime_stat_avg(stat, STAT_TOPDOWN_BE_BOUND, map_idx, rsd) > 0)
669 c++;
670 if (runtime_stat_avg(stat, STAT_TOPDOWN_FE_BOUND, map_idx, rsd) > 0)
671 c++;
672 if (runtime_stat_avg(stat, STAT_TOPDOWN_BAD_SPEC, map_idx, rsd) > 0)
673 c++;
674 return c >= 2;
675 }
676
print_smi_cost(struct perf_stat_config * config,int map_idx,struct perf_stat_output_ctx * out,struct runtime_stat * st,struct runtime_stat_data * rsd)677 static void print_smi_cost(struct perf_stat_config *config, int map_idx,
678 struct perf_stat_output_ctx *out,
679 struct runtime_stat *st,
680 struct runtime_stat_data *rsd)
681 {
682 double smi_num, aperf, cycles, cost = 0.0;
683 const char *color = NULL;
684
685 smi_num = runtime_stat_avg(st, STAT_SMI_NUM, map_idx, rsd);
686 aperf = runtime_stat_avg(st, STAT_APERF, map_idx, rsd);
687 cycles = runtime_stat_avg(st, STAT_CYCLES, map_idx, rsd);
688
689 if ((cycles == 0) || (aperf == 0))
690 return;
691
692 if (smi_num)
693 cost = (aperf - cycles) / aperf * 100.00;
694
695 if (cost > 10)
696 color = PERF_COLOR_RED;
697 out->print_metric(config, out->ctx, color, "%8.1f%%", "SMI cycles%", cost);
698 out->print_metric(config, out->ctx, NULL, "%4.0f", "SMI#", smi_num);
699 }
700
prepare_metric(struct evsel ** metric_events,struct metric_ref * metric_refs,struct expr_parse_ctx * pctx,int map_idx,struct runtime_stat * st)701 static int prepare_metric(struct evsel **metric_events,
702 struct metric_ref *metric_refs,
703 struct expr_parse_ctx *pctx,
704 int map_idx,
705 struct runtime_stat *st)
706 {
707 double scale;
708 char *n;
709 int i, j, ret;
710
711 for (i = 0; metric_events[i]; i++) {
712 struct saved_value *v;
713 struct stats *stats;
714 u64 metric_total = 0;
715 int source_count;
716
717 if (evsel__is_tool(metric_events[i])) {
718 source_count = 1;
719 switch (metric_events[i]->tool_event) {
720 case PERF_TOOL_DURATION_TIME:
721 stats = &walltime_nsecs_stats;
722 scale = 1e-9;
723 break;
724 case PERF_TOOL_USER_TIME:
725 stats = &ru_stats.ru_utime_usec_stat;
726 scale = 1e-6;
727 break;
728 case PERF_TOOL_SYSTEM_TIME:
729 stats = &ru_stats.ru_stime_usec_stat;
730 scale = 1e-6;
731 break;
732 case PERF_TOOL_NONE:
733 pr_err("Invalid tool event 'none'");
734 abort();
735 case PERF_TOOL_MAX:
736 pr_err("Invalid tool event 'max'");
737 abort();
738 default:
739 pr_err("Unknown tool event '%s'", evsel__name(metric_events[i]));
740 abort();
741 }
742 } else {
743 v = saved_value_lookup(metric_events[i], map_idx, false,
744 STAT_NONE, 0, st,
745 metric_events[i]->cgrp);
746 if (!v)
747 break;
748 stats = &v->stats;
749 /*
750 * If an event was scaled during stat gathering, reverse
751 * the scale before computing the metric.
752 */
753 scale = 1.0 / metric_events[i]->scale;
754
755 source_count = evsel__source_count(metric_events[i]);
756
757 if (v->metric_other)
758 metric_total = v->metric_total * scale;
759 }
760 n = strdup(evsel__metric_id(metric_events[i]));
761 if (!n)
762 return -ENOMEM;
763
764 expr__add_id_val_source_count(pctx, n,
765 metric_total ? : avg_stats(stats) * scale,
766 source_count);
767 }
768
769 for (j = 0; metric_refs && metric_refs[j].metric_name; j++) {
770 ret = expr__add_ref(pctx, &metric_refs[j]);
771 if (ret)
772 return ret;
773 }
774
775 return i;
776 }
777
generic_metric(struct perf_stat_config * config,const char * metric_expr,struct evsel ** metric_events,struct metric_ref * metric_refs,char * name,const char * metric_name,const char * metric_unit,int runtime,int map_idx,struct perf_stat_output_ctx * out,struct runtime_stat * st)778 static void generic_metric(struct perf_stat_config *config,
779 const char *metric_expr,
780 struct evsel **metric_events,
781 struct metric_ref *metric_refs,
782 char *name,
783 const char *metric_name,
784 const char *metric_unit,
785 int runtime,
786 int map_idx,
787 struct perf_stat_output_ctx *out,
788 struct runtime_stat *st)
789 {
790 print_metric_t print_metric = out->print_metric;
791 struct expr_parse_ctx *pctx;
792 double ratio, scale;
793 int i;
794 void *ctxp = out->ctx;
795
796 pctx = expr__ctx_new();
797 if (!pctx)
798 return;
799
800 if (config->user_requested_cpu_list)
801 pctx->sctx.user_requested_cpu_list = strdup(config->user_requested_cpu_list);
802 pctx->sctx.runtime = runtime;
803 pctx->sctx.system_wide = config->system_wide;
804 i = prepare_metric(metric_events, metric_refs, pctx, map_idx, st);
805 if (i < 0) {
806 expr__ctx_free(pctx);
807 return;
808 }
809 if (!metric_events[i]) {
810 if (expr__parse(&ratio, pctx, metric_expr) == 0) {
811 char *unit;
812 char metric_bf[64];
813
814 if (metric_unit && metric_name) {
815 if (perf_pmu__convert_scale(metric_unit,
816 &unit, &scale) >= 0) {
817 ratio *= scale;
818 }
819 if (strstr(metric_expr, "?"))
820 scnprintf(metric_bf, sizeof(metric_bf),
821 "%s %s_%d", unit, metric_name, runtime);
822 else
823 scnprintf(metric_bf, sizeof(metric_bf),
824 "%s %s", unit, metric_name);
825
826 print_metric(config, ctxp, NULL, "%8.1f",
827 metric_bf, ratio);
828 } else {
829 print_metric(config, ctxp, NULL, "%8.2f",
830 metric_name ?
831 metric_name :
832 out->force_header ? name : "",
833 ratio);
834 }
835 } else {
836 print_metric(config, ctxp, NULL, NULL,
837 out->force_header ?
838 (metric_name ? metric_name : name) : "", 0);
839 }
840 } else {
841 print_metric(config, ctxp, NULL, NULL,
842 out->force_header ?
843 (metric_name ? metric_name : name) : "", 0);
844 }
845
846 expr__ctx_free(pctx);
847 }
848
test_generic_metric(struct metric_expr * mexp,int map_idx,struct runtime_stat * st)849 double test_generic_metric(struct metric_expr *mexp, int map_idx, struct runtime_stat *st)
850 {
851 struct expr_parse_ctx *pctx;
852 double ratio = 0.0;
853
854 pctx = expr__ctx_new();
855 if (!pctx)
856 return NAN;
857
858 if (prepare_metric(mexp->metric_events, mexp->metric_refs, pctx, map_idx, st) < 0)
859 goto out;
860
861 if (expr__parse(&ratio, pctx, mexp->metric_expr))
862 ratio = 0.0;
863
864 out:
865 expr__ctx_free(pctx);
866 return ratio;
867 }
868
perf_stat__print_shadow_stats(struct perf_stat_config * config,struct evsel * evsel,double avg,int map_idx,struct perf_stat_output_ctx * out,struct rblist * metric_events,struct runtime_stat * st)869 void perf_stat__print_shadow_stats(struct perf_stat_config *config,
870 struct evsel *evsel,
871 double avg, int map_idx,
872 struct perf_stat_output_ctx *out,
873 struct rblist *metric_events,
874 struct runtime_stat *st)
875 {
876 void *ctxp = out->ctx;
877 print_metric_t print_metric = out->print_metric;
878 double total, ratio = 0.0, total2;
879 const char *color = NULL;
880 struct runtime_stat_data rsd = {
881 .ctx = evsel_context(evsel),
882 .cgrp = evsel->cgrp,
883 };
884 struct metric_event *me;
885 int num = 1;
886
887 if (config->iostat_run) {
888 iostat_print_metric(config, evsel, out);
889 } else if (evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
890 total = runtime_stat_avg(st, STAT_CYCLES, map_idx, &rsd);
891
892 if (total) {
893 ratio = avg / total;
894 print_metric(config, ctxp, NULL, "%7.2f ",
895 "insn per cycle", ratio);
896 } else {
897 print_metric(config, ctxp, NULL, NULL, "insn per cycle", 0);
898 }
899
900 total = runtime_stat_avg(st, STAT_STALLED_CYCLES_FRONT, map_idx, &rsd);
901
902 total = max(total, runtime_stat_avg(st,
903 STAT_STALLED_CYCLES_BACK,
904 map_idx, &rsd));
905
906 if (total && avg) {
907 out->new_line(config, ctxp);
908 ratio = total / avg;
909 print_metric(config, ctxp, NULL, "%7.2f ",
910 "stalled cycles per insn",
911 ratio);
912 }
913 } else if (evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES)) {
914 if (runtime_stat_n(st, STAT_BRANCHES, map_idx, &rsd) != 0)
915 print_branch_misses(config, map_idx, avg, out, st, &rsd);
916 else
917 print_metric(config, ctxp, NULL, NULL, "of all branches", 0);
918 } else if (
919 evsel->core.attr.type == PERF_TYPE_HW_CACHE &&
920 evsel->core.attr.config == ( PERF_COUNT_HW_CACHE_L1D |
921 ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
922 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
923
924 if (runtime_stat_n(st, STAT_L1_DCACHE, map_idx, &rsd) != 0)
925 print_l1_dcache_misses(config, map_idx, avg, out, st, &rsd);
926 else
927 print_metric(config, ctxp, NULL, NULL, "of all L1-dcache accesses", 0);
928 } else if (
929 evsel->core.attr.type == PERF_TYPE_HW_CACHE &&
930 evsel->core.attr.config == ( PERF_COUNT_HW_CACHE_L1I |
931 ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
932 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
933
934 if (runtime_stat_n(st, STAT_L1_ICACHE, map_idx, &rsd) != 0)
935 print_l1_icache_misses(config, map_idx, avg, out, st, &rsd);
936 else
937 print_metric(config, ctxp, NULL, NULL, "of all L1-icache accesses", 0);
938 } else if (
939 evsel->core.attr.type == PERF_TYPE_HW_CACHE &&
940 evsel->core.attr.config == ( PERF_COUNT_HW_CACHE_DTLB |
941 ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
942 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
943
944 if (runtime_stat_n(st, STAT_DTLB_CACHE, map_idx, &rsd) != 0)
945 print_dtlb_cache_misses(config, map_idx, avg, out, st, &rsd);
946 else
947 print_metric(config, ctxp, NULL, NULL, "of all dTLB cache accesses", 0);
948 } else if (
949 evsel->core.attr.type == PERF_TYPE_HW_CACHE &&
950 evsel->core.attr.config == ( PERF_COUNT_HW_CACHE_ITLB |
951 ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
952 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
953
954 if (runtime_stat_n(st, STAT_ITLB_CACHE, map_idx, &rsd) != 0)
955 print_itlb_cache_misses(config, map_idx, avg, out, st, &rsd);
956 else
957 print_metric(config, ctxp, NULL, NULL, "of all iTLB cache accesses", 0);
958 } else if (
959 evsel->core.attr.type == PERF_TYPE_HW_CACHE &&
960 evsel->core.attr.config == ( PERF_COUNT_HW_CACHE_LL |
961 ((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
962 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
963
964 if (runtime_stat_n(st, STAT_LL_CACHE, map_idx, &rsd) != 0)
965 print_ll_cache_misses(config, map_idx, avg, out, st, &rsd);
966 else
967 print_metric(config, ctxp, NULL, NULL, "of all LL-cache accesses", 0);
968 } else if (evsel__match(evsel, HARDWARE, HW_CACHE_MISSES)) {
969 total = runtime_stat_avg(st, STAT_CACHEREFS, map_idx, &rsd);
970
971 if (total)
972 ratio = avg * 100 / total;
973
974 if (runtime_stat_n(st, STAT_CACHEREFS, map_idx, &rsd) != 0)
975 print_metric(config, ctxp, NULL, "%8.3f %%",
976 "of all cache refs", ratio);
977 else
978 print_metric(config, ctxp, NULL, NULL, "of all cache refs", 0);
979 } else if (evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
980 print_stalled_cycles_frontend(config, map_idx, avg, out, st, &rsd);
981 } else if (evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
982 print_stalled_cycles_backend(config, map_idx, avg, out, st, &rsd);
983 } else if (evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
984 total = runtime_stat_avg(st, STAT_NSECS, map_idx, &rsd);
985
986 if (total) {
987 ratio = avg / total;
988 print_metric(config, ctxp, NULL, "%8.3f", "GHz", ratio);
989 } else {
990 print_metric(config, ctxp, NULL, NULL, "Ghz", 0);
991 }
992 } else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
993 total = runtime_stat_avg(st, STAT_CYCLES, map_idx, &rsd);
994
995 if (total)
996 print_metric(config, ctxp, NULL,
997 "%7.2f%%", "transactional cycles",
998 100.0 * (avg / total));
999 else
1000 print_metric(config, ctxp, NULL, NULL, "transactional cycles",
1001 0);
1002 } else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
1003 total = runtime_stat_avg(st, STAT_CYCLES, map_idx, &rsd);
1004 total2 = runtime_stat_avg(st, STAT_CYCLES_IN_TX, map_idx, &rsd);
1005
1006 if (total2 < avg)
1007 total2 = avg;
1008 if (total)
1009 print_metric(config, ctxp, NULL, "%7.2f%%", "aborted cycles",
1010 100.0 * ((total2-avg) / total));
1011 else
1012 print_metric(config, ctxp, NULL, NULL, "aborted cycles", 0);
1013 } else if (perf_stat_evsel__is(evsel, TRANSACTION_START)) {
1014 total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, map_idx, &rsd);
1015
1016 if (avg)
1017 ratio = total / avg;
1018
1019 if (runtime_stat_n(st, STAT_CYCLES_IN_TX, map_idx, &rsd) != 0)
1020 print_metric(config, ctxp, NULL, "%8.0f",
1021 "cycles / transaction", ratio);
1022 else
1023 print_metric(config, ctxp, NULL, NULL, "cycles / transaction",
1024 0);
1025 } else if (perf_stat_evsel__is(evsel, ELISION_START)) {
1026 total = runtime_stat_avg(st, STAT_CYCLES_IN_TX, map_idx, &rsd);
1027
1028 if (avg)
1029 ratio = total / avg;
1030
1031 print_metric(config, ctxp, NULL, "%8.0f", "cycles / elision", ratio);
1032 } else if (evsel__is_clock(evsel)) {
1033 if ((ratio = avg_stats(&walltime_nsecs_stats)) != 0)
1034 print_metric(config, ctxp, NULL, "%8.3f", "CPUs utilized",
1035 avg / (ratio * evsel->scale));
1036 else
1037 print_metric(config, ctxp, NULL, NULL, "CPUs utilized", 0);
1038 } else if (perf_stat_evsel__is(evsel, TOPDOWN_FETCH_BUBBLES)) {
1039 double fe_bound = td_fe_bound(map_idx, st, &rsd);
1040
1041 if (fe_bound > 0.2)
1042 color = PERF_COLOR_RED;
1043 print_metric(config, ctxp, color, "%8.1f%%", "frontend bound",
1044 fe_bound * 100.);
1045 } else if (perf_stat_evsel__is(evsel, TOPDOWN_SLOTS_RETIRED)) {
1046 double retiring = td_retiring(map_idx, st, &rsd);
1047
1048 if (retiring > 0.7)
1049 color = PERF_COLOR_GREEN;
1050 print_metric(config, ctxp, color, "%8.1f%%", "retiring",
1051 retiring * 100.);
1052 } else if (perf_stat_evsel__is(evsel, TOPDOWN_RECOVERY_BUBBLES)) {
1053 double bad_spec = td_bad_spec(map_idx, st, &rsd);
1054
1055 if (bad_spec > 0.1)
1056 color = PERF_COLOR_RED;
1057 print_metric(config, ctxp, color, "%8.1f%%", "bad speculation",
1058 bad_spec * 100.);
1059 } else if (perf_stat_evsel__is(evsel, TOPDOWN_SLOTS_ISSUED)) {
1060 double be_bound = td_be_bound(map_idx, st, &rsd);
1061 const char *name = "backend bound";
1062 static int have_recovery_bubbles = -1;
1063
1064 /* In case the CPU does not support topdown-recovery-bubbles */
1065 if (have_recovery_bubbles < 0)
1066 have_recovery_bubbles = pmu_have_event("cpu",
1067 "topdown-recovery-bubbles");
1068 if (!have_recovery_bubbles)
1069 name = "backend bound/bad spec";
1070
1071 if (be_bound > 0.2)
1072 color = PERF_COLOR_RED;
1073 if (td_total_slots(map_idx, st, &rsd) > 0)
1074 print_metric(config, ctxp, color, "%8.1f%%", name,
1075 be_bound * 100.);
1076 else
1077 print_metric(config, ctxp, NULL, NULL, name, 0);
1078 } else if (perf_stat_evsel__is(evsel, TOPDOWN_RETIRING) &&
1079 full_td(map_idx, st, &rsd)) {
1080 double retiring = td_metric_ratio(map_idx,
1081 STAT_TOPDOWN_RETIRING, st,
1082 &rsd);
1083 if (retiring > 0.7)
1084 color = PERF_COLOR_GREEN;
1085 print_metric(config, ctxp, color, "%8.1f%%", "Retiring",
1086 retiring * 100.);
1087 } else if (perf_stat_evsel__is(evsel, TOPDOWN_FE_BOUND) &&
1088 full_td(map_idx, st, &rsd)) {
1089 double fe_bound = td_metric_ratio(map_idx,
1090 STAT_TOPDOWN_FE_BOUND, st,
1091 &rsd);
1092 if (fe_bound > 0.2)
1093 color = PERF_COLOR_RED;
1094 print_metric(config, ctxp, color, "%8.1f%%", "Frontend Bound",
1095 fe_bound * 100.);
1096 } else if (perf_stat_evsel__is(evsel, TOPDOWN_BE_BOUND) &&
1097 full_td(map_idx, st, &rsd)) {
1098 double be_bound = td_metric_ratio(map_idx,
1099 STAT_TOPDOWN_BE_BOUND, st,
1100 &rsd);
1101 if (be_bound > 0.2)
1102 color = PERF_COLOR_RED;
1103 print_metric(config, ctxp, color, "%8.1f%%", "Backend Bound",
1104 be_bound * 100.);
1105 } else if (perf_stat_evsel__is(evsel, TOPDOWN_BAD_SPEC) &&
1106 full_td(map_idx, st, &rsd)) {
1107 double bad_spec = td_metric_ratio(map_idx,
1108 STAT_TOPDOWN_BAD_SPEC, st,
1109 &rsd);
1110 if (bad_spec > 0.1)
1111 color = PERF_COLOR_RED;
1112 print_metric(config, ctxp, color, "%8.1f%%", "Bad Speculation",
1113 bad_spec * 100.);
1114 } else if (perf_stat_evsel__is(evsel, TOPDOWN_HEAVY_OPS) &&
1115 full_td(map_idx, st, &rsd) && (config->topdown_level > 1)) {
1116 double retiring = td_metric_ratio(map_idx,
1117 STAT_TOPDOWN_RETIRING, st,
1118 &rsd);
1119 double heavy_ops = td_metric_ratio(map_idx,
1120 STAT_TOPDOWN_HEAVY_OPS, st,
1121 &rsd);
1122 double light_ops = retiring - heavy_ops;
1123
1124 if (retiring > 0.7 && heavy_ops > 0.1)
1125 color = PERF_COLOR_GREEN;
1126 print_metric(config, ctxp, color, "%8.1f%%", "Heavy Operations",
1127 heavy_ops * 100.);
1128 if (retiring > 0.7 && light_ops > 0.6)
1129 color = PERF_COLOR_GREEN;
1130 else
1131 color = NULL;
1132 print_metric(config, ctxp, color, "%8.1f%%", "Light Operations",
1133 light_ops * 100.);
1134 } else if (perf_stat_evsel__is(evsel, TOPDOWN_BR_MISPREDICT) &&
1135 full_td(map_idx, st, &rsd) && (config->topdown_level > 1)) {
1136 double bad_spec = td_metric_ratio(map_idx,
1137 STAT_TOPDOWN_BAD_SPEC, st,
1138 &rsd);
1139 double br_mis = td_metric_ratio(map_idx,
1140 STAT_TOPDOWN_BR_MISPREDICT, st,
1141 &rsd);
1142 double m_clears = bad_spec - br_mis;
1143
1144 if (bad_spec > 0.1 && br_mis > 0.05)
1145 color = PERF_COLOR_RED;
1146 print_metric(config, ctxp, color, "%8.1f%%", "Branch Mispredict",
1147 br_mis * 100.);
1148 if (bad_spec > 0.1 && m_clears > 0.05)
1149 color = PERF_COLOR_RED;
1150 else
1151 color = NULL;
1152 print_metric(config, ctxp, color, "%8.1f%%", "Machine Clears",
1153 m_clears * 100.);
1154 } else if (perf_stat_evsel__is(evsel, TOPDOWN_FETCH_LAT) &&
1155 full_td(map_idx, st, &rsd) && (config->topdown_level > 1)) {
1156 double fe_bound = td_metric_ratio(map_idx,
1157 STAT_TOPDOWN_FE_BOUND, st,
1158 &rsd);
1159 double fetch_lat = td_metric_ratio(map_idx,
1160 STAT_TOPDOWN_FETCH_LAT, st,
1161 &rsd);
1162 double fetch_bw = fe_bound - fetch_lat;
1163
1164 if (fe_bound > 0.2 && fetch_lat > 0.15)
1165 color = PERF_COLOR_RED;
1166 print_metric(config, ctxp, color, "%8.1f%%", "Fetch Latency",
1167 fetch_lat * 100.);
1168 if (fe_bound > 0.2 && fetch_bw > 0.1)
1169 color = PERF_COLOR_RED;
1170 else
1171 color = NULL;
1172 print_metric(config, ctxp, color, "%8.1f%%", "Fetch Bandwidth",
1173 fetch_bw * 100.);
1174 } else if (perf_stat_evsel__is(evsel, TOPDOWN_MEM_BOUND) &&
1175 full_td(map_idx, st, &rsd) && (config->topdown_level > 1)) {
1176 double be_bound = td_metric_ratio(map_idx,
1177 STAT_TOPDOWN_BE_BOUND, st,
1178 &rsd);
1179 double mem_bound = td_metric_ratio(map_idx,
1180 STAT_TOPDOWN_MEM_BOUND, st,
1181 &rsd);
1182 double core_bound = be_bound - mem_bound;
1183
1184 if (be_bound > 0.2 && mem_bound > 0.2)
1185 color = PERF_COLOR_RED;
1186 print_metric(config, ctxp, color, "%8.1f%%", "Memory Bound",
1187 mem_bound * 100.);
1188 if (be_bound > 0.2 && core_bound > 0.1)
1189 color = PERF_COLOR_RED;
1190 else
1191 color = NULL;
1192 print_metric(config, ctxp, color, "%8.1f%%", "Core Bound",
1193 core_bound * 100.);
1194 } else if (runtime_stat_n(st, STAT_NSECS, map_idx, &rsd) != 0) {
1195 char unit = ' ';
1196 char unit_buf[10] = "/sec";
1197
1198 total = runtime_stat_avg(st, STAT_NSECS, map_idx, &rsd);
1199 if (total)
1200 ratio = convert_unit_double(1000000000.0 * avg / total, &unit);
1201
1202 if (unit != ' ')
1203 snprintf(unit_buf, sizeof(unit_buf), "%c/sec", unit);
1204 print_metric(config, ctxp, NULL, "%8.3f", unit_buf, ratio);
1205 } else if (perf_stat_evsel__is(evsel, SMI_NUM)) {
1206 print_smi_cost(config, map_idx, out, st, &rsd);
1207 } else {
1208 num = 0;
1209 }
1210
1211 if ((me = metricgroup__lookup(metric_events, evsel, false)) != NULL) {
1212 struct metric_expr *mexp;
1213
1214 list_for_each_entry (mexp, &me->head, nd) {
1215 if (num++ > 0)
1216 out->new_line(config, ctxp);
1217 generic_metric(config, mexp->metric_expr, mexp->metric_events,
1218 mexp->metric_refs, evsel->name, mexp->metric_name,
1219 mexp->metric_unit, mexp->runtime,
1220 map_idx, out, st);
1221 }
1222 }
1223 if (num == 0)
1224 print_metric(config, ctxp, NULL, NULL, NULL, 0);
1225 }
1226