forked from OpenCloudOS/perf-prof
-
Notifications
You must be signed in to change notification settings - Fork 0
/
ldlat-loads.c
503 lines (428 loc) · 13.4 KB
/
ldlat-loads.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <sys/types.h>
#include <linux/zalloc.h>
#include <linux/rblist.h>
#include <monitor.h>
#include <stack_helpers.h>
#include <latency_helpers.h>
#define START_OF_KERNEL 0xffff000000000000UL
static profiler ldlat_loads;
static struct monitor_ctx {
struct latency_dist *lat_dist;
struct callchain_ctx *ccx;
struct env *env;
} ctx;
static int monitor_ctx_init(struct env *env)
{
if (get_cpu_vendor() != X86_VENDOR_INTEL) {
fprintf(stderr, "Only supports Intel platforms\n");
return -1;
}
ctx.lat_dist = latency_dist_new(env->perins, true, 0);
ctx.ccx = callchain_ctx_new(CALLCHAIN_KERNEL | CALLCHAIN_USER, stdout);
callchain_ctx_config(ctx.ccx, 1, 1, 1, 0, 0, '\n', '\n');
ctx.env = env;
return 0;
}
static void monitor_ctx_exit(void)
{
latency_dist_free(ctx.lat_dist);
callchain_ctx_free(ctx.ccx);
}
static int ldlat_loads_init(struct perf_evlist *evlist, struct env *env)
{
struct perf_event_attr attr = {
.type = PERF_TYPE_RAW,
.config = 0x1cd, //MEM_TRANS_RETIRED.* /sys/bus/event_source/devices/cpu/events/mem-loads
.size = sizeof(struct perf_event_attr),
//Every trigger_freq memory load, the PEBS hardware triggers an assist and causes a PEBS record to be written
.sample_period = env->trigger_freq,
.sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID | PERF_SAMPLE_TIME | PERF_SAMPLE_ADDR | PERF_SAMPLE_CPU |
PERF_SAMPLE_WEIGHT | PERF_SAMPLE_DATA_SRC | PERF_SAMPLE_PHYS_ADDR,
.read_format = 0,
.pinned = 1,
.disabled = 1,
.exclude_guest = env->exclude_guest,
.exclude_host = env->exclude_host,
.precise_ip = 3, // enable PEBS
.config1 = env->ldlat <= 0 ? 3 : env->ldlat, // MSR_PEBS_LD_LAT_THRESHOLD MSR
.watermark = 1,
.wakeup_watermark = (ldlat_loads.pages << 12) / 2, // enable large PEBS, PERF_X86_EVENT_LARGE_PEBS
};
struct perf_evsel *evsel;
if (monitor_ctx_init(env) < 0)
return -1;
evsel = perf_evsel__new(&attr);
if (!evsel) {
return -1;
}
perf_evlist__add(evlist, evsel);
return 0;
}
static void ldlat_loads_interval(void);
static void ldlat_loads_exit(struct perf_evlist *evlist)
{
ldlat_loads_interval();
monitor_ctx_exit();
}
// in linux/perf_event.h
// PERF_SAMPLE_IP | PERF_SAMPLE_TID | PERF_SAMPLE_TIME | PERF_SAMPLE_ADDR | PERF_SAMPLE_CPU |
// PERF_SAMPLE_WEIGHT | PERF_SAMPLE_DATA_SRC | PERF_SAMPLE_PHYS_ADDR
struct sample_type_header {
u64 ip;
struct {
__u32 pid;
__u32 tid;
} tid_entry;
u64 time;
u64 addr;
struct {
__u32 cpu;
__u32 reserved;
} cpu_entry;
union perf_sample_weight weight;
u64 data_src;
u64 phys_addr;
};
struct mem_info {
union perf_mem_data_src data_src;
};
static const char * const tlb_access[] = {
"N/A",
"HIT",
"MISS",
"L1",
"L2",
"Walker",
"Fault",
};
static int perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
{
size_t l = 0, i;
u64 m = PERF_MEM_TLB_NA;
u64 hit, miss;
sz -= 1; /* -1 for null termination */
out[0] = '\0';
if (mem_info)
m = mem_info->data_src.mem_dtlb;
hit = m & PERF_MEM_TLB_HIT;
miss = m & PERF_MEM_TLB_MISS;
/* already taken care of */
m &= ~(PERF_MEM_TLB_HIT|PERF_MEM_TLB_MISS);
for (i = 0; m && i < ARRAY_SIZE(tlb_access); i++, m >>= 1) {
if (!(m & 0x1))
continue;
if (l) {
strcat(out, " or ");
l += 4;
}
l += scnprintf(out + l, sz - l, tlb_access[i]);
}
if (*out == '\0')
l += scnprintf(out, sz - l, "N/A");
if (hit)
l += scnprintf(out + l, sz - l, " hit");
if (miss)
l += scnprintf(out + l, sz - l, " miss");
return l;
}
static const char * const mem_lvl[] = {
"N/A",
"HIT",
"MISS",
"L1",
"LFB",
"L2",
"L3",
"Local RAM",
"Remote RAM (1 hop)",
"Remote RAM (2 hops)",
"Remote Cache (1 hop)",
"Remote Cache (2 hops)",
"I/O",
"Uncached",
};
static const char * const mem_lvlnum[] = {
[PERF_MEM_LVLNUM_ANY_CACHE] = "Any cache",
[PERF_MEM_LVLNUM_LFB] = "LFB",
[PERF_MEM_LVLNUM_RAM] = "RAM",
[PERF_MEM_LVLNUM_PMEM] = "PMEM",
[PERF_MEM_LVLNUM_NA] = "N/A",
};
static int perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
{
size_t i, l = 0;
u64 m = PERF_MEM_LVL_NA;
u64 hit, miss;
int printed;
if (mem_info)
m = mem_info->data_src.mem_lvl;
sz -= 1; /* -1 for null termination */
out[0] = '\0';
hit = m & PERF_MEM_LVL_HIT;
miss = m & PERF_MEM_LVL_MISS;
/* already taken care of */
m &= ~(PERF_MEM_LVL_HIT|PERF_MEM_LVL_MISS);
if (mem_info && mem_info->data_src.mem_remote) {
strcat(out, "Remote ");
l += 7;
}
printed = 0;
for (i = 0; m && i < ARRAY_SIZE(mem_lvl); i++, m >>= 1) {
if (!(m & 0x1))
continue;
if (printed++) {
strcat(out, " or ");
l += 4;
}
l += scnprintf(out + l, sz - l, mem_lvl[i]);
}
if (mem_info && mem_info->data_src.mem_lvl_num) {
int lvl = mem_info->data_src.mem_lvl_num;
if (printed++) {
strcat(out, " or ");
l += 4;
}
if (mem_lvlnum[lvl])
l += scnprintf(out + l, sz - l, mem_lvlnum[lvl]);
else
l += scnprintf(out + l, sz - l, "L%d", lvl);
}
if (l == 0)
l += scnprintf(out + l, sz - l, "N/A");
if (hit)
l += scnprintf(out + l, sz - l, " hit");
if (miss)
l += scnprintf(out + l, sz - l, " miss");
return l;
}
static const char * const snoop_access[] = {
"N/A",
"None",
"Hit",
"Miss",
"HitM",
};
static int perf_mem__snp_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
{
size_t i, l = 0;
u64 m = PERF_MEM_SNOOP_NA;
sz -= 1; /* -1 for null termination */
out[0] = '\0';
if (mem_info)
m = mem_info->data_src.mem_snoop;
for (i = 0; m && i < ARRAY_SIZE(snoop_access); i++, m >>= 1) {
if (!(m & 0x1))
continue;
if (l) {
strcat(out, " or ");
l += 4;
}
l += scnprintf(out + l, sz - l, snoop_access[i]);
}
if (mem_info &&
(mem_info->data_src.mem_snoopx & PERF_MEM_SNOOPX_FWD)) {
if (l) {
strcat(out, " or ");
l += 4;
}
l += scnprintf(out + l, sz - l, "Fwd");
}
if (*out == '\0')
l += scnprintf(out, sz - l, "N/A");
return l;
}
static int perf_mem__lck_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
{
u64 mask = PERF_MEM_LOCK_NA;
int l;
if (mem_info)
mask = mem_info->data_src.mem_lock;
if (mask & PERF_MEM_LOCK_NA)
l = scnprintf(out, sz, "N/A");
else if (mask & PERF_MEM_LOCK_LOCKED)
l = scnprintf(out, sz, "Yes");
else
l = scnprintf(out, sz, "No");
return l;
}
static int perf_mem__blk_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
{
size_t l = 0;
u64 mask = PERF_MEM_BLK_NA;
sz -= 1; /* -1 for null termination */
out[0] = '\0';
if (mem_info)
mask = mem_info->data_src.mem_blk;
if (!mask || (mask & PERF_MEM_BLK_NA)) {
l += scnprintf(out + l, sz - l, " N/A");
return l;
}
if (mask & PERF_MEM_BLK_DATA)
l += scnprintf(out + l, sz - l, " Data");
if (mask & PERF_MEM_BLK_ADDR)
l += scnprintf(out + l, sz - l, " Addr");
return l;
}
static int perf_script__meminfo_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
{
int i = 0;
i += perf_mem__lvl_scnprintf(out, sz, mem_info);
if (mem_info->data_src.mem_snoop) {
i += scnprintf(out + i, sz - i, "|SNP ");
i += perf_mem__snp_scnprintf(out + i, sz - i, mem_info);
}
if (mem_info->data_src.mem_dtlb) {
i += scnprintf(out + i, sz - i, "|TLB ");
i += perf_mem__tlb_scnprintf(out + i, sz - i, mem_info);
}
if (mem_info->data_src.mem_lock) {
i += scnprintf(out + i, sz - i, "|LCK ");
i += perf_mem__lck_scnprintf(out + i, sz - i, mem_info);
}
if (mem_info->data_src.mem_blk) {
i += scnprintf(out + i, sz - i, "|BLK ");
i += perf_mem__blk_scnprintf(out + i, sz - i, mem_info);
}
return i;
}
static void ldlat_print_node(void *opaque, struct latency_node *node)
{
int oncpu = (int)(u64)opaque;
struct mem_info mem_info;
char buf[128];
mem_info.data_src.val = node->key;
perf_script__meminfo_scnprintf(buf, sizeof(buf), &mem_info);
if (ctx.env->perins) {
if (oncpu)
printf("[%03d] ", monitor_instance_cpu(node->instance));
else
printf("%-8d ", monitor_instance_thread(node->instance));
}
printf("%-60s %8lu %16lu %12lu %12lu %12lu\n", buf,
node->n, node->sum, node->min, node->sum/node->n, node->max);
}
static void ldlat_loads_interval(void)
{
int i;
int oncpu = monitor_instance_oncpu();
if (latency_dist_empty(ctx.lat_dist))
return ;
print_time(stdout);
printf("\n");
if (ctx.env->perins)
printf(oncpu ? "[CPU] " : "[THREAD] ");
printf("%-60s %8s %16s %12s %12s %12s\n", "Mem Load Latency",
"samples", "total(cycles)", "min(cycles)", "avg(cycles)", "max(cycles)");
if (ctx.env->perins)
printf(oncpu ? "----- " : "-------- ");
for (i=0; i<60; i++) printf("-");
printf(" %8s %16s %12s %12s %12s\n",
"--------", "----------------", "------------", "------------", "------------");
latency_dist_print(ctx.lat_dist, ldlat_print_node, (void *)(u64)oncpu);
return ;
}
static void ldlat_loads_sample(union perf_event *event, int instance)
{
struct sample_type_header *data = (void *)event->sample.array;
struct mem_info mem_info;
char buf[64];
struct {
__u64 nr;
__u64 ips[2];
} callchain;
if (ctx.env->verbose || (ctx.env->greater_than &&
data->weight.full > ctx.env->greater_than)) {
callchain.nr = 2;
callchain.ips[0] = data->ip >= START_OF_KERNEL ? PERF_CONTEXT_KERNEL : PERF_CONTEXT_USER;
callchain.ips[1] = data->ip;
mem_info.data_src.val = data->data_src;
perf_mem__lvl_scnprintf(buf, sizeof(buf), &mem_info);
printf("CPU %3u PID %6u TID %6u DATA ADDR %016lx PHYS %016lx latency %6llu cycles %s RIP ",
data->cpu_entry.cpu, data->tid_entry.pid, data->tid_entry.tid,
data->addr, data->phys_addr, data->weight.full, buf);
print_callchain(ctx.ccx, (struct callchain *)&callchain, data->tid_entry.pid);
}
latency_dist_input(ctx.lat_dist, instance, data->data_src, data->weight.full, ctx.env->greater_than);
}
static const char *ldlat_loads_desc[] = PROFILER_DESC("ldlat-loads",
"[OPTION...] [--ldlat cycles] [-T trigger] [--perins] [--than cycles]",
"Load Latency Performance Monitoring on Intel Platform.", "",
"SYNOPSIS", "",
" Load Latency Performance Monitoring Facility", "",
"EXAMPLES", "",
" "PROGRAME" ldlat-loads -C 0 -i 1000",
" "PROGRAME" ldlat-loads -p 2347 --ldlat 10 --than 100 -i 1000");
static const char *ldlat_loads_argv[] = PROFILER_ARGV("ldlat-loads",
PROFILER_ARGV_OPTION,
"FILTER OPTION:",
"exclude-host", "exclude-guest",
PROFILER_ARGV_PROFILER, "ldlat", "trigger", "perins", "than");
//PEBS
//18.3.4.4.2 Load Latency Performance Monitoring Facility
static profiler ldlat_loads = {
.name = "ldlat-loads",
.desc = ldlat_loads_desc,
.argv = ldlat_loads_argv,
.pages = 32, // 2x PEBS_BUFFER_SIZE, in kernel PEBS_BUFFER_SIZE=(PAGE_SIZE << 4)
.init = ldlat_loads_init,
.deinit = ldlat_loads_exit,
.interval = ldlat_loads_interval,
.sample = ldlat_loads_sample,
};
PROFILER_REGISTER(ldlat_loads);
static profiler ldlat_stores;
static int ldlat_stores_init(struct perf_evlist *evlist, struct env *env)
{
struct perf_event_attr attr = {
.type = PERF_TYPE_RAW,
.config = 0x82d0, //MEM_UOPS_RETIRED.ALL_STORES /sys/bus/event_source/devices/cpu/events/mem-stores
.size = sizeof(struct perf_event_attr),
//Every trigger_freq memory load, the PEBS hardware triggers an assist and causes a PEBS record to be written
.sample_period = env->trigger_freq,
.sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID | PERF_SAMPLE_TIME | PERF_SAMPLE_ADDR | PERF_SAMPLE_CPU |
PERF_SAMPLE_WEIGHT | PERF_SAMPLE_DATA_SRC | PERF_SAMPLE_PHYS_ADDR,
.read_format = 0,
.pinned = 1,
.disabled = 1,
.exclude_guest = env->exclude_guest,
.exclude_host = env->exclude_host,
.precise_ip = 3, // enable PEBS
.watermark = 1,
.wakeup_watermark = (ldlat_stores.pages << 12) / 2, // enable large PEBS, PERF_X86_EVENT_LARGE_PEBS
};
struct perf_evsel *evsel;
if (monitor_ctx_init(env) < 0)
return -1;
evsel = perf_evsel__new(&attr);
if (!evsel) {
return -1;
}
perf_evlist__add(evlist, evsel);
return 0;
}
static const char *ldlat_stores_desc[] = PROFILER_DESC("ldlat-stores",
"[OPTION...] [-T trigger] [--perins]",
"PEBS Data Address Profiling on Intel Platform.", "",
"EXAMPLES", "",
" "PROGRAME" ldlat-stores -C 0 -i 1000",
" "PROGRAME" ldlat-stores -p 2347 -v -i 1000");
static const char *ldlat_stores_argv[] = PROFILER_ARGV("ldlat-stores",
PROFILER_ARGV_OPTION,
"FILTER OPTION:",
"exclude-host", "exclude-guest",
PROFILER_ARGV_PROFILER, "trigger", "perins", "than");
static profiler ldlat_stores = {
.name = "ldlat-stores",
.desc = ldlat_stores_desc,
.argv = ldlat_stores_argv,
.pages = 32, // 2x PEBS_BUFFER_SIZE, in kernel PEBS_BUFFER_SIZE=(PAGE_SIZE << 4)
.init = ldlat_stores_init,
.deinit = ldlat_loads_exit,
.interval = ldlat_loads_interval,
.sample = ldlat_loads_sample,
};
PROFILER_REGISTER(ldlat_stores);