summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorRay Kinsella <mdr@ashroe.eu>2021-11-15 06:48:50 +0000
committerDamjan Marion <dmarion@me.com>2021-11-15 12:41:14 +0000
commite75084025ac967c45af3d4f485a0399b0718c8ec (patch)
treeeb2cc49b57e3434aeefc455dba74916dac3d3e45
parent50bd1d3e256154212198a31932896b07af5b129f (diff)
perfmon: rename bundle to memory stalls
Rename the memory bandwidth bundle to memory stalls, to differentiate it from the bundle that measures memory controller bandwidth boundedness. Type: refactor Signed-off-by: Ray Kinsella <mdr@ashroe.eu> Change-Id: I828c73b6f769046e1ab592712bdf81ceefcd7911
Diffstat
-rw-r--r--src/plugins/perfmon/CMakeLists.txt2
-rw-r--r--src/plugins/perfmon/intel/bundle/memory_stalls.c (renamed from src/plugins/perfmon/intel/bundle/membw_bound.c)16
2 files changed, 9 insertions, 9 deletions
diff --git a/src/plugins/perfmon/CMakeLists.txt b/src/plugins/perfmon/CMakeLists.txt
index 73598cf11a7..44be59ef539 100644
--- a/src/plugins/perfmon/CMakeLists.txt
+++ b/src/plugins/perfmon/CMakeLists.txt
@@ -23,7 +23,7 @@ add_vpp_plugin(perfmon
perfmon.c
intel/core.c
intel/uncore.c
- intel/bundle/membw_bound.c
+ intel/bundle/memory_stalls.c
intel/bundle/inst_and_clock.c
intel/bundle/load_blocks.c
intel/bundle/mem_bw.c
diff --git a/src/plugins/perfmon/intel/bundle/membw_bound.c b/src/plugins/perfmon/intel/bundle/memory_stalls.c
index 2e4b4aa57bb..3de3a615732 100644
--- a/src/plugins/perfmon/intel/bundle/membw_bound.c
+++ b/src/plugins/perfmon/intel/bundle/memory_stalls.c
@@ -17,7 +17,7 @@
#include <perfmon/intel/core.h>
static u8 *
-format_intel_membw_bound (u8 *s, va_list *args)
+format_intel_memory_stalls (u8 *s, va_list *args)
{
perfmon_node_stats_t *ss = va_arg (*args, perfmon_node_stats_t *);
int row = va_arg (*args, int);
@@ -33,13 +33,13 @@ format_intel_membw_bound (u8 *s, va_list *args)
return s;
}
-static perfmon_cpu_supports_t membw_bound_cpu_supports[] = {
+static perfmon_cpu_supports_t memory_stalls_cpu_supports[] = {
{ clib_cpu_supports_avx512_bitalg, PERFMON_BUNDLE_TYPE_NODE },
};
-PERFMON_REGISTER_BUNDLE (intel_core_membw_bound) = {
- .name = "membw-bound",
- .description = "memory bandwidth boundedness",
+PERFMON_REGISTER_BUNDLE (intel_core_memory_stalls) = {
+ .name = "memory-stalls",
+ .description = "cycles not retiring instructions due to memory stalls",
.source = "intel-core",
.events[0] = INTEL_CORE_E_CPU_CLK_UNHALTED_THREAD_P, /* FIXED */
.events[1] = INTEL_CORE_E_CYCLE_ACTIVITY_CYCLES_NO_EXECUTE, /*CMask: 0xFF*/
@@ -49,9 +49,9 @@ PERFMON_REGISTER_BUNDLE (intel_core_membw_bound) = {
.events[5] = INTEL_CORE_E_CYCLE_ACTIVITY_STALLS_L3_MISS, /*CMask: 0xF*/
.events[6] = INTEL_CORE_E_SQ_MISC_SQ_FULL, /*CMask: 0xF*/
.n_events = 7,
- .format_fn = format_intel_membw_bound,
- .cpu_supports = membw_bound_cpu_supports,
- .n_cpu_supports = ARRAY_LEN (membw_bound_cpu_supports),
+ .format_fn = format_intel_memory_stalls,
+ .cpu_supports = memory_stalls_cpu_supports,
+ .n_cpu_supports = ARRAY_LEN (memory_stalls_cpu_supports),
.column_headers = PERFMON_STRINGS ("Cycles/Packet", "Cycles Stall/Packet",
"Mem Stall/Packet",
"L1D Miss Stall/Packet", "FB Full/Packet",
id='n268' href='#n268'>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 
/*
 *------------------------------------------------------------------
 * svm_queue.c - unidirectional shared-memory queues
 *
 * Copyright (c) 2009 Cisco and/or its affiliates.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *------------------------------------------------------------------
 */


#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <pthread.h>
#include <vppinfra/mem.h>
#include <vppinfra/format.h>
#include <vppinfra/cache.h>
#include <svm/queue.h>
#include <vppinfra/time.h>
#include <vppinfra/lock.h>

svm_queue_t *
svm_queue_init (void *base, int nels, int elsize)
{
  svm_queue_t *q;
  pthread_mutexattr_t attr;
  pthread_condattr_t cattr;

  q = (svm_queue_t *) base;
  memset (q, 0, sizeof (*q));

  q->elsize = elsize;
  q->maxsize = nels;
  q->producer_evtfd = -1;
  q->consumer_evtfd = -1;

  memset (&attr, 0, sizeof (attr));
  memset (&cattr, 0, sizeof (cattr));

  if (pthread_mutexattr_init (&attr))
    clib_unix_warning ("mutexattr_init");
  if (pthread_mutexattr_setpshared (&attr, PTHREAD_PROCESS_SHARED))
    clib_unix_warning ("pthread_mutexattr_setpshared");
  if (pthread_mutex_init (&q->mutex, &attr))
    clib_unix_warning ("mutex_init");
  if (pthread_mutexattr_destroy (&attr))
    clib_unix_warning ("mutexattr_destroy");
  if (pthread_condattr_init (&cattr))
    clib_unix_warning ("condattr_init");
  /* prints funny-looking messages in the Linux target */
  if (pthread_condattr_setpshared (&cattr, PTHREAD_PROCESS_SHARED))
    clib_unix_warning ("condattr_setpshared");
  if (pthread_cond_init (&q->condvar, &cattr))
    clib_unix_warning ("cond_init1");
  if (pthread_condattr_destroy (&cattr))
    clib_unix_warning ("cond_init2");

  return (q);
}

svm_queue_t *
svm_queue_alloc_and_init (int nels, int elsize, int consumer_pid)
{
  svm_queue_t *q;

  q = clib_mem_alloc_aligned (sizeof (svm_queue_t)
			      + nels * elsize, CLIB_CACHE_LINE_BYTES);
  memset (q, 0, sizeof (*q));
  q = svm_queue_init (q, nels, elsize);
  q->consumer_pid = consumer_pid;

  return q;
}

/*
 * svm_queue_free
 */
void
svm_queue_free (svm_queue_t * q)
{
  (void) pthread_mutex_destroy (&q->mutex);
  (void) pthread_cond_destroy (&q->condvar);
  clib_mem_free (q);
}

void
svm_queue_lock (svm_queue_t * q)
{
  pthread_mutex_lock (&q->mutex);
}

void
svm_queue_unlock (svm_queue_t * q)
{
  pthread_mutex_unlock (&q->mutex);
}

int
svm_queue_is_full (svm_queue_t * q)
{
  return q->cursize == q->maxsize;
}

static inline void
svm_queue_send_signal (svm_queue_t * q, u8 is_prod)
{
  if (q->producer_evtfd == -1)
    {
      (void) pthread_cond_broadcast (&q->condvar);
    }
  else
    {
      int __clib_unused rv, fd;
      u64 data = 1;
      ASSERT (q->consumer_evtfd != -1);
      fd = is_prod ? q->producer_evtfd : q->consumer_evtfd;
      rv = write (fd, &data, sizeof (data));
    }
}

static inline void
svm_queue_wait_inline (svm_queue_t * q)
{
  if (q->producer_evtfd == -1)
    {
      pthread_cond_wait (&q->condvar, &q->mutex);
    }
  else
    {
      /* Fake a wait for event. We could use epoll but that would mean
       * using yet another fd. Should do for now */
      u32 cursize = q->cursize;
      pthread_mutex_unlock (&q->mutex);
      while (q->cursize == cursize)
	CLIB_PAUSE ();
      pthread_mutex_lock (&q->mutex);
    }
}

void
svm_queue_wait (svm_queue_t * q)
{
  svm_queue_wait_inline (q);
}

static inline int
svm_queue_timedwait_inline (svm_queue_t * q, double timeout)
{
  struct timespec ts;
  ts.tv_sec = unix_time_now () + (u32) timeout;
  ts.tv_nsec = (timeout - (u32) timeout) * 1e9;

  if (q->producer_evtfd == -1)
    {
      return pthread_cond_timedwait (&q->condvar, &q->mutex, &ts);
    }
  else
    {
      double max_time = unix_time_now () + timeout;
      u32 cursize = q->cursize;
      int rv;

      pthread_mutex_unlock (&q->mutex);
      while (q->cursize == cursize && unix_time_now () < max_time)
	CLIB_PAUSE ();
      rv = unix_time_now () < max_time ? 0 : ETIMEDOUT;
      pthread_mutex_lock (&q->mutex);
      return rv;
    }
}

int
svm_queue_timedwait (svm_queue_t * q, double timeout)
{
  return svm_queue_timedwait_inline (q, timeout);
}

/*
 * svm_queue_add_nolock
 */
int
svm_queue_add_nolock (svm_queue_t * q, u8 * elem)
{
  i8 *tailp;
  int need_broadcast = 0;

  if (PREDICT_FALSE (q->cursize == q->maxsize))
    {
      while (q->cursize == q->maxsize)
	svm_queue_wait_inline (q);
    }

  tailp = (i8 *) (&q->data[0] + q->elsize * q->tail);
  clib_memcpy (tailp, elem, q->elsize);

  q->tail++;
  q->cursize++;

  need_broadcast = (q->cursize == 1);

  if (q->tail == q->maxsize)
    q->tail = 0;

  if (need_broadcast)
    svm_queue_send_signal (q, 1);
  return 0;
}

void
svm_queue_add_raw (svm_queue_t * q, u8 * elem)
{
  i8 *tailp;

  tailp = (i8 *) (&q->data[0] + q->elsize * q->tail);
  clib_memcpy (tailp, elem, q->elsize);

  q->tail = (q->tail + 1) % q->maxsize;
  q->cursize++;

  if (q->cursize == 1)
    svm_queue_send_signal (q, 1);
}


/*
 * svm_queue_add
 */
int
svm_queue_add (svm_queue_t * q, u8 * elem, int nowait)
{
  i8 *tailp;
  int need_broadcast = 0;

  if (nowait)
    {
      /* zero on success */
      if (pthread_mutex_trylock (&q->mutex))
	{
	  return (-1);
	}
    }
  else
    pthread_mutex_lock (&q->mutex);

  if (PREDICT_FALSE (q->cursize == q->maxsize))
    {
      if (nowait)
	{
	  pthread_mutex_unlock (&q->mutex);
	  return (-2);
	}
      while (q->cursize == q->maxsize)
	svm_queue_wait_inline (q);
    }

  tailp = (i8 *) (&q->data[0] + q->elsize * q->tail);
  clib_memcpy (tailp, elem, q->elsize);

  q->tail++;
  q->cursize++;

  need_broadcast = (q->cursize == 1);

  if (q->tail == q->maxsize)
    q->tail = 0;

  if (need_broadcast)
    svm_queue_send_signal (q, 1);

  pthread_mutex_unlock (&q->mutex);

  return 0;
}

/*
 * svm_queue_add2
 */
int
svm_queue_add2 (svm_queue_t * q, u8 * elem, u8 * elem2, int nowait)
{
  i8 *tailp;
  int need_broadcast = 0;

  if (nowait)
    {
      /* zero on success */
      if (pthread_mutex_trylock (&q->mutex))
	{
	  return (-1);
	}
    }
  else
    pthread_mutex_lock (&q->mutex);

  if (PREDICT_FALSE (q->cursize + 1 == q->maxsize))
    {
      if (nowait)
	{
	  pthread_mutex_unlock (&q->mutex);
	  return (-2);
	}
      while (q->cursize + 1 == q->maxsize)
	svm_queue_wait_inline (q);
    }

  tailp = (i8 *) (&q->data[0] + q->elsize * q->tail);
  clib_memcpy (tailp, elem, q->elsize);

  q->tail++;
  q->cursize++;

  if (q->tail == q->maxsize)
    q->tail = 0;

  need_broadcast = (q->cursize == 1);

  tailp = (i8 *) (&q->data[0] + q->elsize * q->tail);
  clib_memcpy (tailp, elem2, q->elsize);

  q->tail++;
  q->cursize++;

  if (q->tail == q->maxsize)
    q->tail = 0;

  if (need_broadcast)
    svm_queue_send_signal (q, 1);

  pthread_mutex_unlock (&q->mutex);

  return 0;
}

/*
 * svm_queue_sub
 */
int
svm_queue_sub (svm_queue_t * q, u8 * elem, svm_q_conditional_wait_t cond,
	       u32 time)
{
  i8 *headp;
  int need_broadcast = 0;
  int rc = 0;

  if (cond == SVM_Q_NOWAIT)
    {
      /* zero on success */
      if (pthread_mutex_trylock (&q->mutex))
	{
	  return (-1);
	}
    }
  else
    pthread_mutex_lock (&q->mutex);

  if (PREDICT_FALSE (q->cursize == 0))
    {
      if (cond == SVM_Q_NOWAIT)
	{
	  pthread_mutex_unlock (&q->mutex);
	  return (-2);
	}
      else if (cond == SVM_Q_TIMEDWAIT)
	{
	  while (q->cursize == 0 && rc == 0)
	    rc = svm_queue_timedwait_inline (q, time);

	  if (rc == ETIMEDOUT)
	    {
	      pthread_mutex_unlock (&q->mutex);
	      return ETIMEDOUT;
	    }
	}
      else
	{
	  while (q->cursize == 0)
	    svm_queue_wait_inline (q);
	}
    }

  headp = (i8 *) (&q->data[0] + q->elsize * q->head);
  clib_memcpy (elem, headp, q->elsize);

  q->head++;
  /* $$$$ JFC shouldn't this be == 0? */
  if (q->cursize == q->maxsize)
    need_broadcast = 1;

  q->cursize--;

  if (q->head == q->maxsize)
    q->head = 0;

  if (need_broadcast)
    svm_queue_send_signal (q, 0);

  pthread_mutex_unlock (&q->mutex);

  return 0;
}

int
svm_queue_sub2 (svm_queue_t * q, u8 * elem)
{
  int need_broadcast;
  i8 *headp;

  pthread_mutex_lock (&q->mutex);
  if (q->cursize == 0)
    {
      pthread_mutex_unlock (&q->mutex);
      return -1;
    }

  headp = (i8 *) (&q->data[0] + q->elsize * q->head);
  clib_memcpy (elem, headp, q->elsize);

  q->head++;
  need_broadcast = (q->cursize == q->maxsize / 2);
  q->cursize--;

  if (PREDICT_FALSE (q->head == q->maxsize))
    q->head = 0;
  pthread_mutex_unlock (&q->mutex);

  if (need_broadcast)
    svm_queue_send_signal (q, 0);

  return 0;
}

int
svm_queue_sub_raw (svm_queue_t * q, u8 * elem)
{
  i8 *headp;

  if (PREDICT_FALSE (q->cursize == 0))
    {
      while (q->cursize == 0)
	;
    }

  headp = (i8 *) (&q->data[0] + q->elsize * q->head);
  clib_memcpy (elem, headp, q->elsize);

  q->head = (q->head + 1) % q->maxsize;
  q->cursize--;

  return 0;
}

void
svm_queue_set_producer_event_fd (svm_queue_t * q, int fd)
{
  q->producer_evtfd = fd;
}

void
svm_queue_set_consumer_event_fd (svm_queue_t * q, int fd)
{
  q->consumer_evtfd = fd;
}

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */