stat-shadow.c 15.2 KB
Newer Older
1 2 3 4
#include <stdio.h>
#include "evsel.h"
#include "stat.h"
#include "color.h"
5
#include "pmu.h"
6 7 8 9 10 11 12 13 14 15 16 17

enum {
	CTX_BIT_USER	= 1 << 0,
	CTX_BIT_KERNEL	= 1 << 1,
	CTX_BIT_HV	= 1 << 2,
	CTX_BIT_HOST	= 1 << 3,
	CTX_BIT_IDLE	= 1 << 4,
	CTX_BIT_MAX	= 1 << 5,
};

#define NUM_CTX CTX_BIT_MAX

18 19 20 21 22 23 24
/*
 * AGGR_GLOBAL: Use CPU 0
 * AGGR_SOCKET: Use first CPU of socket
 * AGGR_CORE: Use first CPU of core
 * AGGR_NONE: Use matching CPU
 * AGGR_THREAD: Not supported?
 */
25 26 27 28 29 30 31 32 33 34 35 36 37 38
static struct stats runtime_nsecs_stats[MAX_NR_CPUS];
static struct stats runtime_cycles_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_stalled_cycles_front_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_stalled_cycles_back_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_branches_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_cacherefs_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_l1_dcache_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_l1_icache_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_ll_cache_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_itlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_dtlb_cache_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_cycles_in_tx_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_transaction_stats[NUM_CTX][MAX_NR_CPUS];
static struct stats runtime_elision_stats[NUM_CTX][MAX_NR_CPUS];
39
static bool have_frontend_stalled;
40 41 42

struct stats walltime_nsecs_stats;

43 44 45 46 47
void perf_stat__init_shadow_stats(void)
{
	have_frontend_stalled = pmu_have_event("cpu", "stalled-cycles-frontend");
}

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
static int evsel_context(struct perf_evsel *evsel)
{
	int ctx = 0;

	if (evsel->attr.exclude_kernel)
		ctx |= CTX_BIT_KERNEL;
	if (evsel->attr.exclude_user)
		ctx |= CTX_BIT_USER;
	if (evsel->attr.exclude_hv)
		ctx |= CTX_BIT_HV;
	if (evsel->attr.exclude_host)
		ctx |= CTX_BIT_HOST;
	if (evsel->attr.exclude_idle)
		ctx |= CTX_BIT_IDLE;

	return ctx;
}

void perf_stat__reset_shadow_stats(void)
{
	memset(runtime_nsecs_stats, 0, sizeof(runtime_nsecs_stats));
	memset(runtime_cycles_stats, 0, sizeof(runtime_cycles_stats));
	memset(runtime_stalled_cycles_front_stats, 0, sizeof(runtime_stalled_cycles_front_stats));
	memset(runtime_stalled_cycles_back_stats, 0, sizeof(runtime_stalled_cycles_back_stats));
	memset(runtime_branches_stats, 0, sizeof(runtime_branches_stats));
	memset(runtime_cacherefs_stats, 0, sizeof(runtime_cacherefs_stats));
	memset(runtime_l1_dcache_stats, 0, sizeof(runtime_l1_dcache_stats));
	memset(runtime_l1_icache_stats, 0, sizeof(runtime_l1_icache_stats));
	memset(runtime_ll_cache_stats, 0, sizeof(runtime_ll_cache_stats));
	memset(runtime_itlb_cache_stats, 0, sizeof(runtime_itlb_cache_stats));
	memset(runtime_dtlb_cache_stats, 0, sizeof(runtime_dtlb_cache_stats));
	memset(runtime_cycles_in_tx_stats, 0,
			sizeof(runtime_cycles_in_tx_stats));
	memset(runtime_transaction_stats, 0,
		sizeof(runtime_transaction_stats));
	memset(runtime_elision_stats, 0, sizeof(runtime_elision_stats));
	memset(&walltime_nsecs_stats, 0, sizeof(walltime_nsecs_stats));
}

/*
 * Update various tracking values we maintain to print
 * more semantic information such as miss/hit ratios,
 * instruction rates, etc:
 */
void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
				    int cpu)
{
	int ctx = evsel_context(counter);

97 98
	if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK) ||
	    perf_evsel__match(counter, SOFTWARE, SW_CPU_CLOCK))
99 100 101 102
		update_stats(&runtime_nsecs_stats[cpu], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
103
		update_stats(&runtime_cycles_in_tx_stats[ctx][cpu], count[0]);
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
	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
	else if (perf_stat_evsel__is(counter, ELISION_START))
		update_stats(&runtime_elision_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
		update_stats(&runtime_stalled_cycles_front_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
		update_stats(&runtime_stalled_cycles_back_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
		update_stats(&runtime_branches_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
		update_stats(&runtime_cacherefs_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
		update_stats(&runtime_l1_dcache_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1I))
		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_LL))
		update_stats(&runtime_ll_cache_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_DTLB))
		update_stats(&runtime_dtlb_cache_stats[ctx][cpu], count[0]);
	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_ITLB))
		update_stats(&runtime_itlb_cache_stats[ctx][cpu], count[0]);
}

/* used for get_ratio_color() */
enum grc_type {
	GRC_STALLED_CYCLES_FE,
	GRC_STALLED_CYCLES_BE,
	GRC_CACHE_MISSES,
	GRC_MAX_NR
};

static const char *get_ratio_color(enum grc_type type, double ratio)
{
	static const double grc_table[GRC_MAX_NR][3] = {
		[GRC_STALLED_CYCLES_FE] = { 50.0, 30.0, 10.0 },
		[GRC_STALLED_CYCLES_BE] = { 75.0, 50.0, 20.0 },
		[GRC_CACHE_MISSES] 	= { 20.0, 10.0, 5.0 },
	};
	const char *color = PERF_COLOR_NORMAL;

	if (ratio > grc_table[type][0])
		color = PERF_COLOR_RED;
	else if (ratio > grc_table[type][1])
		color = PERF_COLOR_MAGENTA;
	else if (ratio > grc_table[type][2])
		color = PERF_COLOR_YELLOW;

	return color;
}

155
static void print_stalled_cycles_frontend(int cpu,
156
					  struct perf_evsel *evsel, double avg,
157
					  struct perf_stat_output_ctx *out)
158 159 160 161 162 163 164 165 166 167 168 169
{
	double total, ratio = 0.0;
	const char *color;
	int ctx = evsel_context(evsel);

	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = get_ratio_color(GRC_STALLED_CYCLES_FE, ratio);

170 171 172 173 174
	if (ratio)
		out->print_metric(out->ctx, color, "%7.2f%%", "frontend cycles idle",
				  ratio);
	else
		out->print_metric(out->ctx, NULL, NULL, "frontend cycles idle", 0);
175 176
}

177
static void print_stalled_cycles_backend(int cpu,
178
					 struct perf_evsel *evsel, double avg,
179
					 struct perf_stat_output_ctx *out)
180 181 182 183 184 185 186 187 188 189 190 191
{
	double total, ratio = 0.0;
	const char *color;
	int ctx = evsel_context(evsel);

	total = avg_stats(&runtime_cycles_stats[ctx][cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = get_ratio_color(GRC_STALLED_CYCLES_BE, ratio);

192
	out->print_metric(out->ctx, color, "%7.2f%%", "backend cycles idle", ratio);
193 194
}

195
static void print_branch_misses(int cpu,
196
				struct perf_evsel *evsel,
197 198
				double avg,
				struct perf_stat_output_ctx *out)
199 200 201 202 203 204 205 206 207 208 209 210
{
	double total, ratio = 0.0;
	const char *color;
	int ctx = evsel_context(evsel);

	total = avg_stats(&runtime_branches_stats[ctx][cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = get_ratio_color(GRC_CACHE_MISSES, ratio);

211
	out->print_metric(out->ctx, color, "%7.2f%%", "of all branches", ratio);
212 213
}

214
static void print_l1_dcache_misses(int cpu,
215
				   struct perf_evsel *evsel,
216 217
				   double avg,
				   struct perf_stat_output_ctx *out)
218 219 220 221 222 223 224 225 226 227 228 229
{
	double total, ratio = 0.0;
	const char *color;
	int ctx = evsel_context(evsel);

	total = avg_stats(&runtime_l1_dcache_stats[ctx][cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = get_ratio_color(GRC_CACHE_MISSES, ratio);

230
	out->print_metric(out->ctx, color, "%7.2f%%", "of all L1-dcache hits", ratio);
231 232
}

233
static void print_l1_icache_misses(int cpu,
234
				   struct perf_evsel *evsel,
235 236
				   double avg,
				   struct perf_stat_output_ctx *out)
237 238 239 240 241 242 243 244 245 246 247
{
	double total, ratio = 0.0;
	const char *color;
	int ctx = evsel_context(evsel);

	total = avg_stats(&runtime_l1_icache_stats[ctx][cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
248
	out->print_metric(out->ctx, color, "%7.2f%%", "of all L1-icache hits", ratio);
249 250
}

251
static void print_dtlb_cache_misses(int cpu,
252
				    struct perf_evsel *evsel,
253 254
				    double avg,
				    struct perf_stat_output_ctx *out)
255 256 257 258 259 260 261 262 263 264 265
{
	double total, ratio = 0.0;
	const char *color;
	int ctx = evsel_context(evsel);

	total = avg_stats(&runtime_dtlb_cache_stats[ctx][cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
266
	out->print_metric(out->ctx, color, "%7.2f%%", "of all dTLB cache hits", ratio);
267 268
}

269
static void print_itlb_cache_misses(int cpu,
270
				    struct perf_evsel *evsel,
271 272
				    double avg,
				    struct perf_stat_output_ctx *out)
273 274 275 276 277 278 279 280 281 282 283
{
	double total, ratio = 0.0;
	const char *color;
	int ctx = evsel_context(evsel);

	total = avg_stats(&runtime_itlb_cache_stats[ctx][cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
284
	out->print_metric(out->ctx, color, "%7.2f%%", "of all iTLB cache hits", ratio);
285 286
}

287
static void print_ll_cache_misses(int cpu,
288
				  struct perf_evsel *evsel,
289 290
				  double avg,
				  struct perf_stat_output_ctx *out)
291 292 293 294 295 296 297 298 299 300 301
{
	double total, ratio = 0.0;
	const char *color;
	int ctx = evsel_context(evsel);

	total = avg_stats(&runtime_ll_cache_stats[ctx][cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = get_ratio_color(GRC_CACHE_MISSES, ratio);
302
	out->print_metric(out->ctx, color, "%7.2f%%", "of all LL-cache hits", ratio);
303 304
}

305 306 307
void perf_stat__print_shadow_stats(struct perf_evsel *evsel,
				   double avg, int cpu,
				   struct perf_stat_output_ctx *out)
308
{
309 310
	void *ctxp = out->ctx;
	print_metric_t print_metric = out->print_metric;
311 312 313 314 315 316 317
	double total, ratio = 0.0, total2;
	int ctx = evsel_context(evsel);

	if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
		if (total) {
			ratio = avg / total;
318 319
			print_metric(ctxp, NULL, "%7.2f ",
					"insn per cycle", ratio);
320
		} else {
321
			print_metric(ctxp, NULL, NULL, "insn per cycle", 0);
322 323 324 325 326
		}
		total = avg_stats(&runtime_stalled_cycles_front_stats[ctx][cpu]);
		total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[ctx][cpu]));

		if (total && avg) {
327
			out->new_line(ctxp);
328
			ratio = total / avg;
329 330 331
			print_metric(ctxp, NULL, "%7.2f ",
					"stalled cycles per insn",
					ratio);
332
		} else if (have_frontend_stalled) {
333 334
			print_metric(ctxp, NULL, NULL,
				     "stalled cycles per insn", 0);
335
		}
336 337 338 339 340
	} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES)) {
		if (runtime_branches_stats[ctx][cpu].n != 0)
			print_branch_misses(cpu, evsel, avg, out);
		else
			print_metric(ctxp, NULL, NULL, "of all branches", 0);
341 342 343 344
	} else if (
		evsel->attr.type == PERF_TYPE_HW_CACHE &&
		evsel->attr.config ==  ( PERF_COUNT_HW_CACHE_L1D |
					((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
345 346 347 348 349
					 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
		if (runtime_l1_dcache_stats[ctx][cpu].n != 0)
			print_l1_dcache_misses(cpu, evsel, avg, out);
		else
			print_metric(ctxp, NULL, NULL, "of all L1-dcache hits", 0);
350 351 352 353
	} else if (
		evsel->attr.type == PERF_TYPE_HW_CACHE &&
		evsel->attr.config ==  ( PERF_COUNT_HW_CACHE_L1I |
					((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
354 355 356 357 358
					 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
		if (runtime_l1_icache_stats[ctx][cpu].n != 0)
			print_l1_icache_misses(cpu, evsel, avg, out);
		else
			print_metric(ctxp, NULL, NULL, "of all L1-icache hits", 0);
359 360 361 362
	} else if (
		evsel->attr.type == PERF_TYPE_HW_CACHE &&
		evsel->attr.config ==  ( PERF_COUNT_HW_CACHE_DTLB |
					((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
363 364 365 366 367
					 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
		if (runtime_dtlb_cache_stats[ctx][cpu].n != 0)
			print_dtlb_cache_misses(cpu, evsel, avg, out);
		else
			print_metric(ctxp, NULL, NULL, "of all dTLB cache hits", 0);
368 369 370 371
	} else if (
		evsel->attr.type == PERF_TYPE_HW_CACHE &&
		evsel->attr.config ==  ( PERF_COUNT_HW_CACHE_ITLB |
					((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
372 373 374 375 376
					 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
		if (runtime_itlb_cache_stats[ctx][cpu].n != 0)
			print_itlb_cache_misses(cpu, evsel, avg, out);
		else
			print_metric(ctxp, NULL, NULL, "of all iTLB cache hits", 0);
377 378 379 380
	} else if (
		evsel->attr.type == PERF_TYPE_HW_CACHE &&
		evsel->attr.config ==  ( PERF_COUNT_HW_CACHE_LL |
					((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
381 382 383 384 385 386
					 ((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16))) {
		if (runtime_ll_cache_stats[ctx][cpu].n != 0)
			print_ll_cache_misses(cpu, evsel, avg, out);
		else
			print_metric(ctxp, NULL, NULL, "of all LL-cache hits", 0);
	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES)) {
387 388 389 390 391
		total = avg_stats(&runtime_cacherefs_stats[ctx][cpu]);

		if (total)
			ratio = avg * 100 / total;

392 393 394 395 396
		if (runtime_cacherefs_stats[ctx][cpu].n != 0)
			print_metric(ctxp, NULL, "%8.3f %%",
				     "of all cache refs", ratio);
		else
			print_metric(ctxp, NULL, NULL, "of all cache refs", 0);
397
	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
398
		print_stalled_cycles_frontend(cpu, evsel, avg, out);
399
	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
400
		print_stalled_cycles_backend(cpu, evsel, avg, out);
401 402 403 404 405
	} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
		total = avg_stats(&runtime_nsecs_stats[cpu]);

		if (total) {
			ratio = avg / total;
406
			print_metric(ctxp, NULL, "%8.3f", "GHz", ratio);
407
		} else {
408
			print_metric(ctxp, NULL, NULL, "Ghz", 0);
409 410 411 412
		}
	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX)) {
		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
		if (total)
413 414 415 416 417 418
			print_metric(ctxp, NULL,
					"%7.2f%%", "transactional cycles",
					100.0 * (avg / total));
		else
			print_metric(ctxp, NULL, NULL, "transactional cycles",
				     0);
419 420 421 422 423 424
	} else if (perf_stat_evsel__is(evsel, CYCLES_IN_TX_CP)) {
		total = avg_stats(&runtime_cycles_stats[ctx][cpu]);
		total2 = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
		if (total2 < avg)
			total2 = avg;
		if (total)
425
			print_metric(ctxp, NULL, "%7.2f%%", "aborted cycles",
426
				100.0 * ((total2-avg) / total));
427 428 429
		else
			print_metric(ctxp, NULL, NULL, "aborted cycles", 0);
	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START)) {
430 431
		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);

432
		if (avg)
433 434
			ratio = total / avg;

435 436 437 438 439 440 441
		if (runtime_cycles_in_tx_stats[ctx][cpu].n != 0)
			print_metric(ctxp, NULL, "%8.0f",
				     "cycles / transaction", ratio);
		else
			print_metric(ctxp, NULL, NULL, "cycles / transaction",
				     0);
	} else if (perf_stat_evsel__is(evsel, ELISION_START)) {
442 443
		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);

444
		if (avg)
445 446
			ratio = total / avg;

447
		print_metric(ctxp, NULL, "%8.0f", "cycles / elision", ratio);
448 449
	} else if (perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK) ||
		   perf_evsel__match(evsel, SOFTWARE, SW_CPU_CLOCK)) {
450
		if ((ratio = avg_stats(&walltime_nsecs_stats)) != 0)
451 452
			print_metric(ctxp, NULL, "%8.3f", "CPUs utilized",
				     avg / ratio);
453
		else
454
			print_metric(ctxp, NULL, NULL, "CPUs utilized", 0);
455 456
	} else if (runtime_nsecs_stats[cpu].n != 0) {
		char unit = 'M';
457
		char unit_buf[10];
458 459 460 461 462 463 464 465 466

		total = avg_stats(&runtime_nsecs_stats[cpu]);

		if (total)
			ratio = 1000.0 * avg / total;
		if (ratio < 0.001) {
			ratio *= 1000;
			unit = 'K';
		}
467 468
		snprintf(unit_buf, sizeof(unit_buf), "%c/sec", unit);
		print_metric(ctxp, NULL, "%8.3f", unit_buf, ratio);
469
	} else {
470
		print_metric(ctxp, NULL, NULL, NULL, 0);
471 472
	}
}