LLVM OpenMP* Runtime Library
kmp_dispatch.cpp
1 /*
2  * kmp_dispatch.cpp: dynamic scheduling - iteration initialization and dispatch.
3  */
4 
5 //===----------------------------------------------------------------------===//
6 //
7 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
8 // See https://llvm.org/LICENSE.txt for license information.
9 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
10 //
11 //===----------------------------------------------------------------------===//
12 
13 /* Dynamic scheduling initialization and dispatch.
14  *
15  * NOTE: __kmp_nth is a constant inside of any dispatch loop, however
16  * it may change values between parallel regions. __kmp_max_nth
17  * is the largest value __kmp_nth may take, 1 is the smallest.
18  */
19 
20 #include "kmp.h"
21 #include "kmp_error.h"
22 #include "kmp_i18n.h"
23 #include "kmp_itt.h"
24 #include "kmp_stats.h"
25 #include "kmp_str.h"
26 #if KMP_USE_X87CONTROL
27 #include <float.h>
28 #endif
29 #include "kmp_lock.h"
30 #include "kmp_dispatch.h"
31 #if KMP_USE_HIER_SCHED
32 #include "kmp_dispatch_hier.h"
33 #endif
34 
35 #if OMPT_SUPPORT
36 #include "ompt-specific.h"
37 #endif
38 
39 /* ------------------------------------------------------------------------ */
40 /* ------------------------------------------------------------------------ */
41 
42 void __kmp_dispatch_deo_error(int *gtid_ref, int *cid_ref, ident_t *loc_ref) {
43  kmp_info_t *th;
44 
45  KMP_DEBUG_ASSERT(gtid_ref);
46 
47  if (__kmp_env_consistency_check) {
48  th = __kmp_threads[*gtid_ref];
49  if (th->th.th_root->r.r_active &&
50  (th->th.th_dispatch->th_dispatch_pr_current->pushed_ws != ct_none)) {
51 #if KMP_USE_DYNAMIC_LOCK
52  __kmp_push_sync(*gtid_ref, ct_ordered_in_pdo, loc_ref, NULL, 0);
53 #else
54  __kmp_push_sync(*gtid_ref, ct_ordered_in_pdo, loc_ref, NULL);
55 #endif
56  }
57  }
58 }
59 
60 void __kmp_dispatch_dxo_error(int *gtid_ref, int *cid_ref, ident_t *loc_ref) {
61  kmp_info_t *th;
62 
63  if (__kmp_env_consistency_check) {
64  th = __kmp_threads[*gtid_ref];
65  if (th->th.th_dispatch->th_dispatch_pr_current->pushed_ws != ct_none) {
66  __kmp_pop_sync(*gtid_ref, ct_ordered_in_pdo, loc_ref);
67  }
68  }
69 }
70 
71 // Returns either SCHEDULE_MONOTONIC or SCHEDULE_NONMONOTONIC
72 static inline int __kmp_get_monotonicity(enum sched_type schedule,
73  bool use_hier = false) {
74  // Pick up the nonmonotonic/monotonic bits from the scheduling type
75  int monotonicity;
76  // default to monotonic
77  monotonicity = SCHEDULE_MONOTONIC;
78  if (SCHEDULE_HAS_NONMONOTONIC(schedule))
79  monotonicity = SCHEDULE_NONMONOTONIC;
80  else if (SCHEDULE_HAS_MONOTONIC(schedule))
81  monotonicity = SCHEDULE_MONOTONIC;
82  return monotonicity;
83 }
84 
85 // Initialize a dispatch_private_info_template<T> buffer for a particular
86 // type of schedule,chunk. The loop description is found in lb (lower bound),
87 // ub (upper bound), and st (stride). nproc is the number of threads relevant
88 // to the scheduling (often the number of threads in a team, but not always if
89 // hierarchical scheduling is used). tid is the id of the thread calling
90 // the function within the group of nproc threads. It will have a value
91 // between 0 and nproc - 1. This is often just the thread id within a team, but
92 // is not necessarily the case when using hierarchical scheduling.
93 // loc is the source file location of the corresponding loop
94 // gtid is the global thread id
95 template <typename T>
96 void __kmp_dispatch_init_algorithm(ident_t *loc, int gtid,
97  dispatch_private_info_template<T> *pr,
98  enum sched_type schedule, T lb, T ub,
99  typename traits_t<T>::signed_t st,
100 #if USE_ITT_BUILD
101  kmp_uint64 *cur_chunk,
102 #endif
103  typename traits_t<T>::signed_t chunk,
104  T nproc, T tid) {
105  typedef typename traits_t<T>::unsigned_t UT;
106  typedef typename traits_t<T>::floating_t DBL;
107 
108  int active;
109  T tc;
110  kmp_info_t *th;
111  kmp_team_t *team;
112  int monotonicity;
113  bool use_hier;
114 
115 #ifdef KMP_DEBUG
116  typedef typename traits_t<T>::signed_t ST;
117  {
118  char *buff;
119  // create format specifiers before the debug output
120  buff = __kmp_str_format("__kmp_dispatch_init_algorithm: T#%%d called "
121  "pr:%%p lb:%%%s ub:%%%s st:%%%s "
122  "schedule:%%d chunk:%%%s nproc:%%%s tid:%%%s\n",
123  traits_t<T>::spec, traits_t<T>::spec,
124  traits_t<ST>::spec, traits_t<ST>::spec,
125  traits_t<T>::spec, traits_t<T>::spec);
126  KD_TRACE(10, (buff, gtid, pr, lb, ub, st, schedule, chunk, nproc, tid));
127  __kmp_str_free(&buff);
128  }
129 #endif
130  /* setup data */
131  th = __kmp_threads[gtid];
132  team = th->th.th_team;
133  active = !team->t.t_serialized;
134 
135 #if USE_ITT_BUILD
136  int itt_need_metadata_reporting =
137  __itt_metadata_add_ptr && __kmp_forkjoin_frames_mode == 3 &&
138  KMP_MASTER_GTID(gtid) && th->th.th_teams_microtask == NULL &&
139  team->t.t_active_level == 1;
140 #endif
141 
142 #if KMP_USE_HIER_SCHED
143  use_hier = pr->flags.use_hier;
144 #else
145  use_hier = false;
146 #endif
147 
148  /* Pick up the nonmonotonic/monotonic bits from the scheduling type */
149  monotonicity = __kmp_get_monotonicity(schedule, use_hier);
150  schedule = SCHEDULE_WITHOUT_MODIFIERS(schedule);
151 
152  /* Pick up the nomerge/ordered bits from the scheduling type */
153  if ((schedule >= kmp_nm_lower) && (schedule < kmp_nm_upper)) {
154  pr->flags.nomerge = TRUE;
155  schedule =
156  (enum sched_type)(((int)schedule) - (kmp_nm_lower - kmp_sch_lower));
157  } else {
158  pr->flags.nomerge = FALSE;
159  }
160  pr->type_size = traits_t<T>::type_size; // remember the size of variables
161  if (kmp_ord_lower & schedule) {
162  pr->flags.ordered = TRUE;
163  schedule =
164  (enum sched_type)(((int)schedule) - (kmp_ord_lower - kmp_sch_lower));
165  } else {
166  pr->flags.ordered = FALSE;
167  }
168  // Ordered overrides nonmonotonic
169  if (pr->flags.ordered) {
170  monotonicity = SCHEDULE_MONOTONIC;
171  }
172 
173  if (schedule == kmp_sch_static) {
174  schedule = __kmp_static;
175  } else {
176  if (schedule == kmp_sch_runtime) {
177  // Use the scheduling specified by OMP_SCHEDULE (or __kmp_sch_default if
178  // not specified)
179  schedule = team->t.t_sched.r_sched_type;
180  monotonicity = __kmp_get_monotonicity(schedule, use_hier);
181  schedule = SCHEDULE_WITHOUT_MODIFIERS(schedule);
182  // Detail the schedule if needed (global controls are differentiated
183  // appropriately)
184  if (schedule == kmp_sch_guided_chunked) {
185  schedule = __kmp_guided;
186  } else if (schedule == kmp_sch_static) {
187  schedule = __kmp_static;
188  }
189  // Use the chunk size specified by OMP_SCHEDULE (or default if not
190  // specified)
191  chunk = team->t.t_sched.chunk;
192 #if USE_ITT_BUILD
193  if (cur_chunk)
194  *cur_chunk = chunk;
195 #endif
196 #ifdef KMP_DEBUG
197  {
198  char *buff;
199  // create format specifiers before the debug output
200  buff = __kmp_str_format("__kmp_dispatch_init_algorithm: T#%%d new: "
201  "schedule:%%d chunk:%%%s\n",
202  traits_t<ST>::spec);
203  KD_TRACE(10, (buff, gtid, schedule, chunk));
204  __kmp_str_free(&buff);
205  }
206 #endif
207  } else {
208  if (schedule == kmp_sch_guided_chunked) {
209  schedule = __kmp_guided;
210  }
211  if (chunk <= 0) {
212  chunk = KMP_DEFAULT_CHUNK;
213  }
214  }
215 
216  if (schedule == kmp_sch_auto) {
217  // mapping and differentiation: in the __kmp_do_serial_initialize()
218  schedule = __kmp_auto;
219 #ifdef KMP_DEBUG
220  {
221  char *buff;
222  // create format specifiers before the debug output
223  buff = __kmp_str_format(
224  "__kmp_dispatch_init_algorithm: kmp_sch_auto: T#%%d new: "
225  "schedule:%%d chunk:%%%s\n",
226  traits_t<ST>::spec);
227  KD_TRACE(10, (buff, gtid, schedule, chunk));
228  __kmp_str_free(&buff);
229  }
230 #endif
231  }
232 #if KMP_STATIC_STEAL_ENABLED
233  // map nonmonotonic:dynamic to static steal
234  if (schedule == kmp_sch_dynamic_chunked) {
235  if (monotonicity == SCHEDULE_NONMONOTONIC)
236  schedule = kmp_sch_static_steal;
237  }
238 #endif
239  /* guided analytical not safe for too many threads */
240  if (schedule == kmp_sch_guided_analytical_chunked && nproc > 1 << 20) {
241  schedule = kmp_sch_guided_iterative_chunked;
242  KMP_WARNING(DispatchManyThreads);
243  }
244  if (schedule == kmp_sch_runtime_simd) {
245  // compiler provides simd_width in the chunk parameter
246  schedule = team->t.t_sched.r_sched_type;
247  monotonicity = __kmp_get_monotonicity(schedule, use_hier);
248  schedule = SCHEDULE_WITHOUT_MODIFIERS(schedule);
249  // Detail the schedule if needed (global controls are differentiated
250  // appropriately)
251  if (schedule == kmp_sch_static || schedule == kmp_sch_auto ||
252  schedule == __kmp_static) {
253  schedule = kmp_sch_static_balanced_chunked;
254  } else {
255  if (schedule == kmp_sch_guided_chunked || schedule == __kmp_guided) {
256  schedule = kmp_sch_guided_simd;
257  }
258  chunk = team->t.t_sched.chunk * chunk;
259  }
260 #if USE_ITT_BUILD
261  if (cur_chunk)
262  *cur_chunk = chunk;
263 #endif
264 #ifdef KMP_DEBUG
265  {
266  char *buff;
267  // create format specifiers before the debug output
268  buff = __kmp_str_format(
269  "__kmp_dispatch_init_algorithm: T#%%d new: schedule:%%d"
270  " chunk:%%%s\n",
271  traits_t<ST>::spec);
272  KD_TRACE(10, (buff, gtid, schedule, chunk));
273  __kmp_str_free(&buff);
274  }
275 #endif
276  }
277  pr->u.p.parm1 = chunk;
278  }
279  KMP_ASSERT2((kmp_sch_lower < schedule && schedule < kmp_sch_upper),
280  "unknown scheduling type");
281 
282  pr->u.p.count = 0;
283 
284  if (__kmp_env_consistency_check) {
285  if (st == 0) {
286  __kmp_error_construct(kmp_i18n_msg_CnsLoopIncrZeroProhibited,
287  (pr->flags.ordered ? ct_pdo_ordered : ct_pdo), loc);
288  }
289  }
290  // compute trip count
291  if (st == 1) { // most common case
292  if (ub >= lb) {
293  tc = ub - lb + 1;
294  } else { // ub < lb
295  tc = 0; // zero-trip
296  }
297  } else if (st < 0) {
298  if (lb >= ub) {
299  // AC: cast to unsigned is needed for loops like (i=2B; i>-2B; i-=1B),
300  // where the division needs to be unsigned regardless of the result type
301  tc = (UT)(lb - ub) / (-st) + 1;
302  } else { // lb < ub
303  tc = 0; // zero-trip
304  }
305  } else { // st > 0
306  if (ub >= lb) {
307  // AC: cast to unsigned is needed for loops like (i=-2B; i<2B; i+=1B),
308  // where the division needs to be unsigned regardless of the result type
309  tc = (UT)(ub - lb) / st + 1;
310  } else { // ub < lb
311  tc = 0; // zero-trip
312  }
313  }
314 
315 #if KMP_STATS_ENABLED
316  if (KMP_MASTER_GTID(gtid)) {
317  KMP_COUNT_VALUE(OMP_loop_dynamic_total_iterations, tc);
318  }
319 #endif
320 
321  pr->u.p.lb = lb;
322  pr->u.p.ub = ub;
323  pr->u.p.st = st;
324  pr->u.p.tc = tc;
325 
326 #if KMP_OS_WINDOWS
327  pr->u.p.last_upper = ub + st;
328 #endif /* KMP_OS_WINDOWS */
329 
330  /* NOTE: only the active parallel region(s) has active ordered sections */
331 
332  if (active) {
333  if (pr->flags.ordered) {
334  pr->ordered_bumped = 0;
335  pr->u.p.ordered_lower = 1;
336  pr->u.p.ordered_upper = 0;
337  }
338  }
339 
340  switch (schedule) {
341 #if (KMP_STATIC_STEAL_ENABLED)
342  case kmp_sch_static_steal: {
343  T ntc, init;
344 
345  KD_TRACE(100,
346  ("__kmp_dispatch_init_algorithm: T#%d kmp_sch_static_steal case\n",
347  gtid));
348 
349  ntc = (tc % chunk ? 1 : 0) + tc / chunk;
350  if (nproc > 1 && ntc >= nproc) {
351  KMP_COUNT_BLOCK(OMP_LOOP_STATIC_STEAL);
352  T id = tid;
353  T small_chunk, extras;
354 
355  small_chunk = ntc / nproc;
356  extras = ntc % nproc;
357 
358  init = id * small_chunk + (id < extras ? id : extras);
359  pr->u.p.count = init;
360  pr->u.p.ub = init + small_chunk + (id < extras ? 1 : 0);
361 
362  pr->u.p.parm2 = lb;
363  // parm3 is the number of times to attempt stealing which is
364  // proportional to the number of chunks per thread up until
365  // the maximum value of nproc.
366  pr->u.p.parm3 = KMP_MIN(small_chunk + extras, nproc);
367  pr->u.p.parm4 = (id + 1) % nproc; // remember neighbour tid
368  pr->u.p.st = st;
369  if (traits_t<T>::type_size > 4) {
370  // AC: TODO: check if 16-byte CAS available and use it to
371  // improve performance (probably wait for explicit request
372  // before spending time on this).
373  // For now use dynamically allocated per-thread lock,
374  // free memory in __kmp_dispatch_next when status==0.
375  KMP_DEBUG_ASSERT(pr->u.p.th_steal_lock == NULL);
376  pr->u.p.th_steal_lock =
377  (kmp_lock_t *)__kmp_allocate(sizeof(kmp_lock_t));
378  __kmp_init_lock(pr->u.p.th_steal_lock);
379  }
380  break;
381  } else {
382  /* too few chunks: switching to kmp_sch_dynamic_chunked */
383  schedule = kmp_sch_dynamic_chunked;
384  KD_TRACE(100, ("__kmp_dispatch_init_algorithm: T#%d switching to "
385  "kmp_sch_dynamic_chunked\n",
386  gtid));
387  if (pr->u.p.parm1 <= 0)
388  pr->u.p.parm1 = KMP_DEFAULT_CHUNK;
389  break;
390  } // if
391  } // case
392 #endif
393  case kmp_sch_static_balanced: {
394  T init, limit;
395 
396  KD_TRACE(
397  100,
398  ("__kmp_dispatch_init_algorithm: T#%d kmp_sch_static_balanced case\n",
399  gtid));
400 
401  if (nproc > 1) {
402  T id = tid;
403 
404  if (tc < nproc) {
405  if (id < tc) {
406  init = id;
407  limit = id;
408  pr->u.p.parm1 = (id == tc - 1); /* parm1 stores *plastiter */
409  } else {
410  pr->u.p.count = 1; /* means no more chunks to execute */
411  pr->u.p.parm1 = FALSE;
412  break;
413  }
414  } else {
415  T small_chunk = tc / nproc;
416  T extras = tc % nproc;
417  init = id * small_chunk + (id < extras ? id : extras);
418  limit = init + small_chunk - (id < extras ? 0 : 1);
419  pr->u.p.parm1 = (id == nproc - 1);
420  }
421  } else {
422  if (tc > 0) {
423  init = 0;
424  limit = tc - 1;
425  pr->u.p.parm1 = TRUE;
426  } else {
427  // zero trip count
428  pr->u.p.count = 1; /* means no more chunks to execute */
429  pr->u.p.parm1 = FALSE;
430  break;
431  }
432  }
433 #if USE_ITT_BUILD
434  // Calculate chunk for metadata report
435  if (itt_need_metadata_reporting)
436  if (cur_chunk)
437  *cur_chunk = limit - init + 1;
438 #endif
439  if (st == 1) {
440  pr->u.p.lb = lb + init;
441  pr->u.p.ub = lb + limit;
442  } else {
443  // calculated upper bound, "ub" is user-defined upper bound
444  T ub_tmp = lb + limit * st;
445  pr->u.p.lb = lb + init * st;
446  // adjust upper bound to "ub" if needed, so that MS lastprivate will match
447  // it exactly
448  if (st > 0) {
449  pr->u.p.ub = (ub_tmp + st > ub ? ub : ub_tmp);
450  } else {
451  pr->u.p.ub = (ub_tmp + st < ub ? ub : ub_tmp);
452  }
453  }
454  if (pr->flags.ordered) {
455  pr->u.p.ordered_lower = init;
456  pr->u.p.ordered_upper = limit;
457  }
458  break;
459  } // case
460  case kmp_sch_static_balanced_chunked: {
461  // similar to balanced, but chunk adjusted to multiple of simd width
462  T nth = nproc;
463  KD_TRACE(100, ("__kmp_dispatch_init_algorithm: T#%d runtime(simd:static)"
464  " -> falling-through to static_greedy\n",
465  gtid));
466  schedule = kmp_sch_static_greedy;
467  if (nth > 1)
468  pr->u.p.parm1 = ((tc + nth - 1) / nth + chunk - 1) & ~(chunk - 1);
469  else
470  pr->u.p.parm1 = tc;
471  break;
472  } // case
473  case kmp_sch_guided_simd:
474  case kmp_sch_guided_iterative_chunked: {
475  KD_TRACE(
476  100,
477  ("__kmp_dispatch_init_algorithm: T#%d kmp_sch_guided_iterative_chunked"
478  " case\n",
479  gtid));
480 
481  if (nproc > 1) {
482  if ((2L * chunk + 1) * nproc >= tc) {
483  /* chunk size too large, switch to dynamic */
484  schedule = kmp_sch_dynamic_chunked;
485  } else {
486  // when remaining iters become less than parm2 - switch to dynamic
487  pr->u.p.parm2 = guided_int_param * nproc * (chunk + 1);
488  *(double *)&pr->u.p.parm3 =
489  guided_flt_param / nproc; // may occupy parm3 and parm4
490  }
491  } else {
492  KD_TRACE(100, ("__kmp_dispatch_init_algorithm: T#%d falling-through to "
493  "kmp_sch_static_greedy\n",
494  gtid));
495  schedule = kmp_sch_static_greedy;
496  /* team->t.t_nproc == 1: fall-through to kmp_sch_static_greedy */
497  KD_TRACE(
498  100,
499  ("__kmp_dispatch_init_algorithm: T#%d kmp_sch_static_greedy case\n",
500  gtid));
501  pr->u.p.parm1 = tc;
502  } // if
503  } // case
504  break;
505  case kmp_sch_guided_analytical_chunked: {
506  KD_TRACE(100, ("__kmp_dispatch_init_algorithm: T#%d "
507  "kmp_sch_guided_analytical_chunked case\n",
508  gtid));
509 
510  if (nproc > 1) {
511  if ((2L * chunk + 1) * nproc >= tc) {
512  /* chunk size too large, switch to dynamic */
513  schedule = kmp_sch_dynamic_chunked;
514  } else {
515  /* commonly used term: (2 nproc - 1)/(2 nproc) */
516  DBL x;
517 
518 #if KMP_USE_X87CONTROL
519  /* Linux* OS already has 64-bit computation by default for long double,
520  and on Windows* OS on Intel(R) 64, /Qlong_double doesn't work. On
521  Windows* OS on IA-32 architecture, we need to set precision to 64-bit
522  instead of the default 53-bit. Even though long double doesn't work
523  on Windows* OS on Intel(R) 64, the resulting lack of precision is not
524  expected to impact the correctness of the algorithm, but this has not
525  been mathematically proven. */
526  // save original FPCW and set precision to 64-bit, as
527  // Windows* OS on IA-32 architecture defaults to 53-bit
528  unsigned int oldFpcw = _control87(0, 0);
529  _control87(_PC_64, _MCW_PC); // 0,0x30000
530 #endif
531  /* value used for comparison in solver for cross-over point */
532  long double target = ((long double)chunk * 2 + 1) * nproc / tc;
533 
534  /* crossover point--chunk indexes equal to or greater than
535  this point switch to dynamic-style scheduling */
536  UT cross;
537 
538  /* commonly used term: (2 nproc - 1)/(2 nproc) */
539  x = (long double)1.0 - (long double)0.5 / nproc;
540 
541 #ifdef KMP_DEBUG
542  { // test natural alignment
543  struct _test_a {
544  char a;
545  union {
546  char b;
547  DBL d;
548  };
549  } t;
550  ptrdiff_t natural_alignment =
551  (ptrdiff_t)&t.b - (ptrdiff_t)&t - (ptrdiff_t)1;
552  //__kmp_warn( " %llx %llx %lld", (long long)&t.d, (long long)&t, (long
553  // long)natural_alignment );
554  KMP_DEBUG_ASSERT(
555  (((ptrdiff_t)&pr->u.p.parm3) & (natural_alignment)) == 0);
556  }
557 #endif // KMP_DEBUG
558 
559  /* save the term in thread private dispatch structure */
560  *(DBL *)&pr->u.p.parm3 = x;
561 
562  /* solve for the crossover point to the nearest integer i for which C_i
563  <= chunk */
564  {
565  UT left, right, mid;
566  long double p;
567 
568  /* estimate initial upper and lower bound */
569 
570  /* doesn't matter what value right is as long as it is positive, but
571  it affects performance of the solver */
572  right = 229;
573  p = __kmp_pow<UT>(x, right);
574  if (p > target) {
575  do {
576  p *= p;
577  right <<= 1;
578  } while (p > target && right < (1 << 27));
579  /* lower bound is previous (failed) estimate of upper bound */
580  left = right >> 1;
581  } else {
582  left = 0;
583  }
584 
585  /* bisection root-finding method */
586  while (left + 1 < right) {
587  mid = (left + right) / 2;
588  if (__kmp_pow<UT>(x, mid) > target) {
589  left = mid;
590  } else {
591  right = mid;
592  }
593  } // while
594  cross = right;
595  }
596  /* assert sanity of computed crossover point */
597  KMP_ASSERT(cross && __kmp_pow<UT>(x, cross - 1) > target &&
598  __kmp_pow<UT>(x, cross) <= target);
599 
600  /* save the crossover point in thread private dispatch structure */
601  pr->u.p.parm2 = cross;
602 
603 // C75803
604 #if ((KMP_OS_LINUX || KMP_OS_WINDOWS) && KMP_ARCH_X86) && (!defined(KMP_I8))
605 #define GUIDED_ANALYTICAL_WORKAROUND (*(DBL *)&pr->u.p.parm3)
606 #else
607 #define GUIDED_ANALYTICAL_WORKAROUND (x)
608 #endif
609  /* dynamic-style scheduling offset */
610  pr->u.p.count = tc - __kmp_dispatch_guided_remaining(
611  tc, GUIDED_ANALYTICAL_WORKAROUND, cross) -
612  cross * chunk;
613 #if KMP_USE_X87CONTROL
614  // restore FPCW
615  _control87(oldFpcw, _MCW_PC);
616 #endif
617  } // if
618  } else {
619  KD_TRACE(100, ("__kmp_dispatch_init_algorithm: T#%d falling-through to "
620  "kmp_sch_static_greedy\n",
621  gtid));
622  schedule = kmp_sch_static_greedy;
623  /* team->t.t_nproc == 1: fall-through to kmp_sch_static_greedy */
624  pr->u.p.parm1 = tc;
625  } // if
626  } // case
627  break;
628  case kmp_sch_static_greedy:
629  KD_TRACE(
630  100,
631  ("__kmp_dispatch_init_algorithm: T#%d kmp_sch_static_greedy case\n",
632  gtid));
633  pr->u.p.parm1 = (nproc > 1) ? (tc + nproc - 1) / nproc : tc;
634  break;
635  case kmp_sch_static_chunked:
636  case kmp_sch_dynamic_chunked:
637  if (pr->u.p.parm1 <= 0) {
638  pr->u.p.parm1 = KMP_DEFAULT_CHUNK;
639  }
640  KD_TRACE(100, ("__kmp_dispatch_init_algorithm: T#%d "
641  "kmp_sch_static_chunked/kmp_sch_dynamic_chunked cases\n",
642  gtid));
643  break;
644  case kmp_sch_trapezoidal: {
645  /* TSS: trapezoid self-scheduling, minimum chunk_size = parm1 */
646 
647  T parm1, parm2, parm3, parm4;
648  KD_TRACE(100,
649  ("__kmp_dispatch_init_algorithm: T#%d kmp_sch_trapezoidal case\n",
650  gtid));
651 
652  parm1 = chunk;
653 
654  /* F : size of the first cycle */
655  parm2 = (tc / (2 * nproc));
656 
657  if (parm2 < 1) {
658  parm2 = 1;
659  }
660 
661  /* L : size of the last cycle. Make sure the last cycle is not larger
662  than the first cycle. */
663  if (parm1 < 1) {
664  parm1 = 1;
665  } else if (parm1 > parm2) {
666  parm1 = parm2;
667  }
668 
669  /* N : number of cycles */
670  parm3 = (parm2 + parm1);
671  parm3 = (2 * tc + parm3 - 1) / parm3;
672 
673  if (parm3 < 2) {
674  parm3 = 2;
675  }
676 
677  /* sigma : decreasing incr of the trapezoid */
678  parm4 = (parm3 - 1);
679  parm4 = (parm2 - parm1) / parm4;
680 
681  // pointless check, because parm4 >= 0 always
682  // if ( parm4 < 0 ) {
683  // parm4 = 0;
684  //}
685 
686  pr->u.p.parm1 = parm1;
687  pr->u.p.parm2 = parm2;
688  pr->u.p.parm3 = parm3;
689  pr->u.p.parm4 = parm4;
690  } // case
691  break;
692 
693  default: {
694  __kmp_fatal(KMP_MSG(UnknownSchedTypeDetected), // Primary message
695  KMP_HNT(GetNewerLibrary), // Hint
696  __kmp_msg_null // Variadic argument list terminator
697  );
698  } break;
699  } // switch
700  pr->schedule = schedule;
701 }
702 
703 #if KMP_USE_HIER_SCHED
704 template <typename T>
705 inline void __kmp_dispatch_init_hier_runtime(ident_t *loc, T lb, T ub,
706  typename traits_t<T>::signed_t st);
707 template <>
708 inline void
709 __kmp_dispatch_init_hier_runtime<kmp_int32>(ident_t *loc, kmp_int32 lb,
710  kmp_int32 ub, kmp_int32 st) {
711  __kmp_dispatch_init_hierarchy<kmp_int32>(
712  loc, __kmp_hier_scheds.size, __kmp_hier_scheds.layers,
713  __kmp_hier_scheds.scheds, __kmp_hier_scheds.small_chunks, lb, ub, st);
714 }
715 template <>
716 inline void
717 __kmp_dispatch_init_hier_runtime<kmp_uint32>(ident_t *loc, kmp_uint32 lb,
718  kmp_uint32 ub, kmp_int32 st) {
719  __kmp_dispatch_init_hierarchy<kmp_uint32>(
720  loc, __kmp_hier_scheds.size, __kmp_hier_scheds.layers,
721  __kmp_hier_scheds.scheds, __kmp_hier_scheds.small_chunks, lb, ub, st);
722 }
723 template <>
724 inline void
725 __kmp_dispatch_init_hier_runtime<kmp_int64>(ident_t *loc, kmp_int64 lb,
726  kmp_int64 ub, kmp_int64 st) {
727  __kmp_dispatch_init_hierarchy<kmp_int64>(
728  loc, __kmp_hier_scheds.size, __kmp_hier_scheds.layers,
729  __kmp_hier_scheds.scheds, __kmp_hier_scheds.large_chunks, lb, ub, st);
730 }
731 template <>
732 inline void
733 __kmp_dispatch_init_hier_runtime<kmp_uint64>(ident_t *loc, kmp_uint64 lb,
734  kmp_uint64 ub, kmp_int64 st) {
735  __kmp_dispatch_init_hierarchy<kmp_uint64>(
736  loc, __kmp_hier_scheds.size, __kmp_hier_scheds.layers,
737  __kmp_hier_scheds.scheds, __kmp_hier_scheds.large_chunks, lb, ub, st);
738 }
739 
740 // free all the hierarchy scheduling memory associated with the team
741 void __kmp_dispatch_free_hierarchies(kmp_team_t *team) {
742  int num_disp_buff = team->t.t_max_nproc > 1 ? __kmp_dispatch_num_buffers : 2;
743  for (int i = 0; i < num_disp_buff; ++i) {
744  // type does not matter here so use kmp_int32
745  auto sh =
746  reinterpret_cast<dispatch_shared_info_template<kmp_int32> volatile *>(
747  &team->t.t_disp_buffer[i]);
748  if (sh->hier) {
749  sh->hier->deallocate();
750  __kmp_free(sh->hier);
751  }
752  }
753 }
754 #endif
755 
756 // UT - unsigned flavor of T, ST - signed flavor of T,
757 // DBL - double if sizeof(T)==4, or long double if sizeof(T)==8
758 template <typename T>
759 static void
760 __kmp_dispatch_init(ident_t *loc, int gtid, enum sched_type schedule, T lb,
761  T ub, typename traits_t<T>::signed_t st,
762  typename traits_t<T>::signed_t chunk, int push_ws) {
763  typedef typename traits_t<T>::unsigned_t UT;
764 
765  int active;
766  kmp_info_t *th;
767  kmp_team_t *team;
768  kmp_uint32 my_buffer_index;
769  dispatch_private_info_template<T> *pr;
770  dispatch_shared_info_template<T> volatile *sh;
771 
772  KMP_BUILD_ASSERT(sizeof(dispatch_private_info_template<T>) ==
773  sizeof(dispatch_private_info));
774  KMP_BUILD_ASSERT(sizeof(dispatch_shared_info_template<UT>) ==
775  sizeof(dispatch_shared_info));
776 
777  if (!TCR_4(__kmp_init_parallel))
778  __kmp_parallel_initialize();
779 
780  __kmp_resume_if_soft_paused();
781 
782 #if INCLUDE_SSC_MARKS
783  SSC_MARK_DISPATCH_INIT();
784 #endif
785 #ifdef KMP_DEBUG
786  typedef typename traits_t<T>::signed_t ST;
787  {
788  char *buff;
789  // create format specifiers before the debug output
790  buff = __kmp_str_format("__kmp_dispatch_init: T#%%d called: schedule:%%d "
791  "chunk:%%%s lb:%%%s ub:%%%s st:%%%s\n",
792  traits_t<ST>::spec, traits_t<T>::spec,
793  traits_t<T>::spec, traits_t<ST>::spec);
794  KD_TRACE(10, (buff, gtid, schedule, chunk, lb, ub, st));
795  __kmp_str_free(&buff);
796  }
797 #endif
798  /* setup data */
799  th = __kmp_threads[gtid];
800  team = th->th.th_team;
801  active = !team->t.t_serialized;
802  th->th.th_ident = loc;
803 
804  // Any half-decent optimizer will remove this test when the blocks are empty
805  // since the macros expand to nothing
806  // when statistics are disabled.
807  if (schedule == __kmp_static) {
808  KMP_COUNT_BLOCK(OMP_LOOP_STATIC);
809  } else {
810  KMP_COUNT_BLOCK(OMP_LOOP_DYNAMIC);
811  }
812 
813 #if KMP_USE_HIER_SCHED
814  // Initialize the scheduling hierarchy if requested in OMP_SCHEDULE envirable
815  // Hierarchical scheduling does not work with ordered, so if ordered is
816  // detected, then revert back to threaded scheduling.
817  bool ordered;
818  enum sched_type my_sched = schedule;
819  my_buffer_index = th->th.th_dispatch->th_disp_index;
820  pr = reinterpret_cast<dispatch_private_info_template<T> *>(
821  &th->th.th_dispatch
822  ->th_disp_buffer[my_buffer_index % __kmp_dispatch_num_buffers]);
823  my_sched = SCHEDULE_WITHOUT_MODIFIERS(my_sched);
824  if ((my_sched >= kmp_nm_lower) && (my_sched < kmp_nm_upper))
825  my_sched =
826  (enum sched_type)(((int)my_sched) - (kmp_nm_lower - kmp_sch_lower));
827  ordered = (kmp_ord_lower & my_sched);
828  if (pr->flags.use_hier) {
829  if (ordered) {
830  KD_TRACE(100, ("__kmp_dispatch_init: T#%d ordered loop detected. "
831  "Disabling hierarchical scheduling.\n",
832  gtid));
833  pr->flags.use_hier = FALSE;
834  }
835  }
836  if (schedule == kmp_sch_runtime && __kmp_hier_scheds.size > 0) {
837  // Don't use hierarchical for ordered parallel loops and don't
838  // use the runtime hierarchy if one was specified in the program
839  if (!ordered && !pr->flags.use_hier)
840  __kmp_dispatch_init_hier_runtime<T>(loc, lb, ub, st);
841  }
842 #endif // KMP_USE_HIER_SCHED
843 
844 #if USE_ITT_BUILD
845  kmp_uint64 cur_chunk = chunk;
846  int itt_need_metadata_reporting =
847  __itt_metadata_add_ptr && __kmp_forkjoin_frames_mode == 3 &&
848  KMP_MASTER_GTID(gtid) && th->th.th_teams_microtask == NULL &&
849  team->t.t_active_level == 1;
850 #endif
851  if (!active) {
852  pr = reinterpret_cast<dispatch_private_info_template<T> *>(
853  th->th.th_dispatch->th_disp_buffer); /* top of the stack */
854  } else {
855  KMP_DEBUG_ASSERT(th->th.th_dispatch ==
856  &th->th.th_team->t.t_dispatch[th->th.th_info.ds.ds_tid]);
857 
858  my_buffer_index = th->th.th_dispatch->th_disp_index++;
859 
860  /* What happens when number of threads changes, need to resize buffer? */
861  pr = reinterpret_cast<dispatch_private_info_template<T> *>(
862  &th->th.th_dispatch
863  ->th_disp_buffer[my_buffer_index % __kmp_dispatch_num_buffers]);
864  sh = reinterpret_cast<dispatch_shared_info_template<T> volatile *>(
865  &team->t.t_disp_buffer[my_buffer_index % __kmp_dispatch_num_buffers]);
866  KD_TRACE(10, ("__kmp_dispatch_init: T#%d my_buffer_index:%d\n", gtid,
867  my_buffer_index));
868  }
869 
870  __kmp_dispatch_init_algorithm(loc, gtid, pr, schedule, lb, ub, st,
871 #if USE_ITT_BUILD
872  &cur_chunk,
873 #endif
874  chunk, (T)th->th.th_team_nproc,
875  (T)th->th.th_info.ds.ds_tid);
876  if (active) {
877  if (pr->flags.ordered == 0) {
878  th->th.th_dispatch->th_deo_fcn = __kmp_dispatch_deo_error;
879  th->th.th_dispatch->th_dxo_fcn = __kmp_dispatch_dxo_error;
880  } else {
881  th->th.th_dispatch->th_deo_fcn = __kmp_dispatch_deo<UT>;
882  th->th.th_dispatch->th_dxo_fcn = __kmp_dispatch_dxo<UT>;
883  }
884  }
885 
886  if (active) {
887  /* The name of this buffer should be my_buffer_index when it's free to use
888  * it */
889 
890  KD_TRACE(100, ("__kmp_dispatch_init: T#%d before wait: my_buffer_index:%d "
891  "sh->buffer_index:%d\n",
892  gtid, my_buffer_index, sh->buffer_index));
893  __kmp_wait<kmp_uint32>(&sh->buffer_index, my_buffer_index,
894  __kmp_eq<kmp_uint32> USE_ITT_BUILD_ARG(NULL));
895  // Note: KMP_WAIT() cannot be used there: buffer index and
896  // my_buffer_index are *always* 32-bit integers.
897  KMP_MB(); /* is this necessary? */
898  KD_TRACE(100, ("__kmp_dispatch_init: T#%d after wait: my_buffer_index:%d "
899  "sh->buffer_index:%d\n",
900  gtid, my_buffer_index, sh->buffer_index));
901 
902  th->th.th_dispatch->th_dispatch_pr_current = (dispatch_private_info_t *)pr;
903  th->th.th_dispatch->th_dispatch_sh_current =
904  CCAST(dispatch_shared_info_t *, (volatile dispatch_shared_info_t *)sh);
905 #if USE_ITT_BUILD
906  if (pr->flags.ordered) {
907  __kmp_itt_ordered_init(gtid);
908  }
909  // Report loop metadata
910  if (itt_need_metadata_reporting) {
911  // Only report metadata by master of active team at level 1
912  kmp_uint64 schedtype = 0;
913  switch (schedule) {
914  case kmp_sch_static_chunked:
915  case kmp_sch_static_balanced: // Chunk is calculated in the switch above
916  break;
917  case kmp_sch_static_greedy:
918  cur_chunk = pr->u.p.parm1;
919  break;
920  case kmp_sch_dynamic_chunked:
921  schedtype = 1;
922  break;
923  case kmp_sch_guided_iterative_chunked:
924  case kmp_sch_guided_analytical_chunked:
925  case kmp_sch_guided_simd:
926  schedtype = 2;
927  break;
928  default:
929  // Should we put this case under "static"?
930  // case kmp_sch_static_steal:
931  schedtype = 3;
932  break;
933  }
934  __kmp_itt_metadata_loop(loc, schedtype, pr->u.p.tc, cur_chunk);
935  }
936 #if KMP_USE_HIER_SCHED
937  if (pr->flags.use_hier) {
938  pr->u.p.count = 0;
939  pr->u.p.ub = pr->u.p.lb = pr->u.p.st = pr->u.p.tc = 0;
940  }
941 #endif // KMP_USER_HIER_SCHED
942 #endif /* USE_ITT_BUILD */
943  }
944 
945 #ifdef KMP_DEBUG
946  {
947  char *buff;
948  // create format specifiers before the debug output
949  buff = __kmp_str_format(
950  "__kmp_dispatch_init: T#%%d returning: schedule:%%d ordered:%%%s "
951  "lb:%%%s ub:%%%s"
952  " st:%%%s tc:%%%s count:%%%s\n\tordered_lower:%%%s ordered_upper:%%%s"
953  " parm1:%%%s parm2:%%%s parm3:%%%s parm4:%%%s\n",
954  traits_t<UT>::spec, traits_t<T>::spec, traits_t<T>::spec,
955  traits_t<ST>::spec, traits_t<UT>::spec, traits_t<UT>::spec,
956  traits_t<UT>::spec, traits_t<UT>::spec, traits_t<T>::spec,
957  traits_t<T>::spec, traits_t<T>::spec, traits_t<T>::spec);
958  KD_TRACE(10, (buff, gtid, pr->schedule, pr->flags.ordered, pr->u.p.lb,
959  pr->u.p.ub, pr->u.p.st, pr->u.p.tc, pr->u.p.count,
960  pr->u.p.ordered_lower, pr->u.p.ordered_upper, pr->u.p.parm1,
961  pr->u.p.parm2, pr->u.p.parm3, pr->u.p.parm4));
962  __kmp_str_free(&buff);
963  }
964 #endif
965 #if (KMP_STATIC_STEAL_ENABLED)
966  // It cannot be guaranteed that after execution of a loop with some other
967  // schedule kind all the parm3 variables will contain the same value. Even if
968  // all parm3 will be the same, it still exists a bad case like using 0 and 1
969  // rather than program life-time increment. So the dedicated variable is
970  // required. The 'static_steal_counter' is used.
971  if (pr->schedule == kmp_sch_static_steal) {
972  // Other threads will inspect this variable when searching for a victim.
973  // This is a flag showing that other threads may steal from this thread
974  // since then.
975  volatile T *p = &pr->u.p.static_steal_counter;
976  *p = *p + 1;
977  }
978 #endif // ( KMP_STATIC_STEAL_ENABLED )
979 
980 #if OMPT_SUPPORT && OMPT_OPTIONAL
981  if (ompt_enabled.ompt_callback_work) {
982  ompt_team_info_t *team_info = __ompt_get_teaminfo(0, NULL);
983  ompt_task_info_t *task_info = __ompt_get_task_info_object(0);
984  ompt_callbacks.ompt_callback(ompt_callback_work)(
985  ompt_work_loop, ompt_scope_begin, &(team_info->parallel_data),
986  &(task_info->task_data), pr->u.p.tc, OMPT_LOAD_RETURN_ADDRESS(gtid));
987  }
988 #endif
989  KMP_PUSH_PARTITIONED_TIMER(OMP_loop_dynamic);
990 }
991 
992 /* For ordered loops, either __kmp_dispatch_finish() should be called after
993  * every iteration, or __kmp_dispatch_finish_chunk() should be called after
994  * every chunk of iterations. If the ordered section(s) were not executed
995  * for this iteration (or every iteration in this chunk), we need to set the
996  * ordered iteration counters so that the next thread can proceed. */
997 template <typename UT>
998 static void __kmp_dispatch_finish(int gtid, ident_t *loc) {
999  typedef typename traits_t<UT>::signed_t ST;
1000  kmp_info_t *th = __kmp_threads[gtid];
1001 
1002  KD_TRACE(100, ("__kmp_dispatch_finish: T#%d called\n", gtid));
1003  if (!th->th.th_team->t.t_serialized) {
1004 
1005  dispatch_private_info_template<UT> *pr =
1006  reinterpret_cast<dispatch_private_info_template<UT> *>(
1007  th->th.th_dispatch->th_dispatch_pr_current);
1008  dispatch_shared_info_template<UT> volatile *sh =
1009  reinterpret_cast<dispatch_shared_info_template<UT> volatile *>(
1010  th->th.th_dispatch->th_dispatch_sh_current);
1011  KMP_DEBUG_ASSERT(pr);
1012  KMP_DEBUG_ASSERT(sh);
1013  KMP_DEBUG_ASSERT(th->th.th_dispatch ==
1014  &th->th.th_team->t.t_dispatch[th->th.th_info.ds.ds_tid]);
1015 
1016  if (pr->ordered_bumped) {
1017  KD_TRACE(
1018  1000,
1019  ("__kmp_dispatch_finish: T#%d resetting ordered_bumped to zero\n",
1020  gtid));
1021  pr->ordered_bumped = 0;
1022  } else {
1023  UT lower = pr->u.p.ordered_lower;
1024 
1025 #ifdef KMP_DEBUG
1026  {
1027  char *buff;
1028  // create format specifiers before the debug output
1029  buff = __kmp_str_format("__kmp_dispatch_finish: T#%%d before wait: "
1030  "ordered_iteration:%%%s lower:%%%s\n",
1031  traits_t<UT>::spec, traits_t<UT>::spec);
1032  KD_TRACE(1000, (buff, gtid, sh->u.s.ordered_iteration, lower));
1033  __kmp_str_free(&buff);
1034  }
1035 #endif
1036 
1037  __kmp_wait<UT>(&sh->u.s.ordered_iteration, lower,
1038  __kmp_ge<UT> USE_ITT_BUILD_ARG(NULL));
1039  KMP_MB(); /* is this necessary? */
1040 #ifdef KMP_DEBUG
1041  {
1042  char *buff;
1043  // create format specifiers before the debug output
1044  buff = __kmp_str_format("__kmp_dispatch_finish: T#%%d after wait: "
1045  "ordered_iteration:%%%s lower:%%%s\n",
1046  traits_t<UT>::spec, traits_t<UT>::spec);
1047  KD_TRACE(1000, (buff, gtid, sh->u.s.ordered_iteration, lower));
1048  __kmp_str_free(&buff);
1049  }
1050 #endif
1051 
1052  test_then_inc<ST>((volatile ST *)&sh->u.s.ordered_iteration);
1053  } // if
1054  } // if
1055  KD_TRACE(100, ("__kmp_dispatch_finish: T#%d returned\n", gtid));
1056 }
1057 
1058 #ifdef KMP_GOMP_COMPAT
1059 
1060 template <typename UT>
1061 static void __kmp_dispatch_finish_chunk(int gtid, ident_t *loc) {
1062  typedef typename traits_t<UT>::signed_t ST;
1063  kmp_info_t *th = __kmp_threads[gtid];
1064 
1065  KD_TRACE(100, ("__kmp_dispatch_finish_chunk: T#%d called\n", gtid));
1066  if (!th->th.th_team->t.t_serialized) {
1067  // int cid;
1068  dispatch_private_info_template<UT> *pr =
1069  reinterpret_cast<dispatch_private_info_template<UT> *>(
1070  th->th.th_dispatch->th_dispatch_pr_current);
1071  dispatch_shared_info_template<UT> volatile *sh =
1072  reinterpret_cast<dispatch_shared_info_template<UT> volatile *>(
1073  th->th.th_dispatch->th_dispatch_sh_current);
1074  KMP_DEBUG_ASSERT(pr);
1075  KMP_DEBUG_ASSERT(sh);
1076  KMP_DEBUG_ASSERT(th->th.th_dispatch ==
1077  &th->th.th_team->t.t_dispatch[th->th.th_info.ds.ds_tid]);
1078 
1079  // for (cid = 0; cid < KMP_MAX_ORDERED; ++cid) {
1080  UT lower = pr->u.p.ordered_lower;
1081  UT upper = pr->u.p.ordered_upper;
1082  UT inc = upper - lower + 1;
1083 
1084  if (pr->ordered_bumped == inc) {
1085  KD_TRACE(
1086  1000,
1087  ("__kmp_dispatch_finish: T#%d resetting ordered_bumped to zero\n",
1088  gtid));
1089  pr->ordered_bumped = 0;
1090  } else {
1091  inc -= pr->ordered_bumped;
1092 
1093 #ifdef KMP_DEBUG
1094  {
1095  char *buff;
1096  // create format specifiers before the debug output
1097  buff = __kmp_str_format(
1098  "__kmp_dispatch_finish_chunk: T#%%d before wait: "
1099  "ordered_iteration:%%%s lower:%%%s upper:%%%s\n",
1100  traits_t<UT>::spec, traits_t<UT>::spec, traits_t<UT>::spec);
1101  KD_TRACE(1000, (buff, gtid, sh->u.s.ordered_iteration, lower, upper));
1102  __kmp_str_free(&buff);
1103  }
1104 #endif
1105 
1106  __kmp_wait<UT>(&sh->u.s.ordered_iteration, lower,
1107  __kmp_ge<UT> USE_ITT_BUILD_ARG(NULL));
1108 
1109  KMP_MB(); /* is this necessary? */
1110  KD_TRACE(1000, ("__kmp_dispatch_finish_chunk: T#%d resetting "
1111  "ordered_bumped to zero\n",
1112  gtid));
1113  pr->ordered_bumped = 0;
1115 #ifdef KMP_DEBUG
1116  {
1117  char *buff;
1118  // create format specifiers before the debug output
1119  buff = __kmp_str_format(
1120  "__kmp_dispatch_finish_chunk: T#%%d after wait: "
1121  "ordered_iteration:%%%s inc:%%%s lower:%%%s upper:%%%s\n",
1122  traits_t<UT>::spec, traits_t<UT>::spec, traits_t<UT>::spec,
1123  traits_t<UT>::spec);
1124  KD_TRACE(1000,
1125  (buff, gtid, sh->u.s.ordered_iteration, inc, lower, upper));
1126  __kmp_str_free(&buff);
1127  }
1128 #endif
1129 
1130  test_then_add<ST>((volatile ST *)&sh->u.s.ordered_iteration, inc);
1131  }
1132  // }
1133  }
1134  KD_TRACE(100, ("__kmp_dispatch_finish_chunk: T#%d returned\n", gtid));
1135 }
1136 
1137 #endif /* KMP_GOMP_COMPAT */
1138 
1139 template <typename T>
1140 int __kmp_dispatch_next_algorithm(int gtid,
1141  dispatch_private_info_template<T> *pr,
1142  dispatch_shared_info_template<T> volatile *sh,
1143  kmp_int32 *p_last, T *p_lb, T *p_ub,
1144  typename traits_t<T>::signed_t *p_st, T nproc,
1145  T tid) {
1146  typedef typename traits_t<T>::unsigned_t UT;
1147  typedef typename traits_t<T>::signed_t ST;
1148  typedef typename traits_t<T>::floating_t DBL;
1149  int status = 0;
1150  kmp_int32 last = 0;
1151  T start;
1152  ST incr;
1153  UT limit, trip, init;
1154  kmp_info_t *th = __kmp_threads[gtid];
1155  kmp_team_t *team = th->th.th_team;
1156 
1157  KMP_DEBUG_ASSERT(th->th.th_dispatch ==
1158  &th->th.th_team->t.t_dispatch[th->th.th_info.ds.ds_tid]);
1159  KMP_DEBUG_ASSERT(pr);
1160  KMP_DEBUG_ASSERT(sh);
1161  KMP_DEBUG_ASSERT(tid >= 0 && tid < nproc);
1162 #ifdef KMP_DEBUG
1163  {
1164  char *buff;
1165  // create format specifiers before the debug output
1166  buff =
1167  __kmp_str_format("__kmp_dispatch_next_algorithm: T#%%d called pr:%%p "
1168  "sh:%%p nproc:%%%s tid:%%%s\n",
1169  traits_t<T>::spec, traits_t<T>::spec);
1170  KD_TRACE(10, (buff, gtid, pr, sh, nproc, tid));
1171  __kmp_str_free(&buff);
1172  }
1173 #endif
1174 
1175  // zero trip count
1176  if (pr->u.p.tc == 0) {
1177  KD_TRACE(10,
1178  ("__kmp_dispatch_next_algorithm: T#%d early exit trip count is "
1179  "zero status:%d\n",
1180  gtid, status));
1181  return 0;
1182  }
1183 
1184  switch (pr->schedule) {
1185 #if (KMP_STATIC_STEAL_ENABLED)
1186  case kmp_sch_static_steal: {
1187  T chunk = pr->u.p.parm1;
1188 
1189  KD_TRACE(100,
1190  ("__kmp_dispatch_next_algorithm: T#%d kmp_sch_static_steal case\n",
1191  gtid));
1192 
1193  trip = pr->u.p.tc - 1;
1194 
1195  if (traits_t<T>::type_size > 4) {
1196  // use lock for 8-byte and CAS for 4-byte induction
1197  // variable. TODO (optional): check and use 16-byte CAS
1198  kmp_lock_t *lck = pr->u.p.th_steal_lock;
1199  KMP_DEBUG_ASSERT(lck != NULL);
1200  if (pr->u.p.count < (UT)pr->u.p.ub) {
1201  __kmp_acquire_lock(lck, gtid);
1202  // try to get own chunk of iterations
1203  init = (pr->u.p.count)++;
1204  status = (init < (UT)pr->u.p.ub);
1205  __kmp_release_lock(lck, gtid);
1206  } else {
1207  status = 0; // no own chunks
1208  }
1209  if (!status) { // try to steal
1210  kmp_info_t **other_threads = team->t.t_threads;
1211  int while_limit = pr->u.p.parm3;
1212  int while_index = 0;
1213  T id = pr->u.p.static_steal_counter; // loop id
1214  int idx = (th->th.th_dispatch->th_disp_index - 1) %
1215  __kmp_dispatch_num_buffers; // current loop index
1216  // note: victim thread can potentially execute another loop
1217  // TODO: algorithm of searching for a victim
1218  // should be cleaned up and measured
1219  while ((!status) && (while_limit != ++while_index)) {
1220  dispatch_private_info_template<T> *victim;
1221  T remaining;
1222  T victimIdx = pr->u.p.parm4;
1223  T oldVictimIdx = victimIdx ? victimIdx - 1 : nproc - 1;
1224  victim = reinterpret_cast<dispatch_private_info_template<T> *>(
1225  &other_threads[victimIdx]->th.th_dispatch->th_disp_buffer[idx]);
1226  KMP_DEBUG_ASSERT(victim);
1227  while ((victim == pr || id != victim->u.p.static_steal_counter) &&
1228  oldVictimIdx != victimIdx) {
1229  victimIdx = (victimIdx + 1) % nproc;
1230  victim = reinterpret_cast<dispatch_private_info_template<T> *>(
1231  &other_threads[victimIdx]->th.th_dispatch->th_disp_buffer[idx]);
1232  KMP_DEBUG_ASSERT(victim);
1233  }
1234  if (victim == pr || id != victim->u.p.static_steal_counter) {
1235  continue; // try once more (nproc attempts in total)
1236  // no victim is ready yet to participate in stealing
1237  // because no victim passed kmp_init_dispatch yet
1238  }
1239  if (victim->u.p.count + 2 > (UT)victim->u.p.ub) {
1240  pr->u.p.parm4 = (victimIdx + 1) % nproc; // shift start tid
1241  continue; // not enough chunks to steal, goto next victim
1242  }
1243 
1244  lck = victim->u.p.th_steal_lock;
1245  KMP_ASSERT(lck != NULL);
1246  __kmp_acquire_lock(lck, gtid);
1247  limit = victim->u.p.ub; // keep initial ub
1248  if (victim->u.p.count >= limit ||
1249  (remaining = limit - victim->u.p.count) < 2) {
1250  __kmp_release_lock(lck, gtid);
1251  pr->u.p.parm4 = (victimIdx + 1) % nproc; // next victim
1252  continue; // not enough chunks to steal
1253  }
1254  // stealing succeeded, reduce victim's ub by 1/4 of undone chunks or
1255  // by 1
1256  if (remaining > 3) {
1257  // steal 1/4 of remaining
1258  KMP_COUNT_DEVELOPER_VALUE(FOR_static_steal_stolen, remaining >> 2);
1259  init = (victim->u.p.ub -= (remaining >> 2));
1260  } else {
1261  // steal 1 chunk of 2 or 3 remaining
1262  KMP_COUNT_DEVELOPER_VALUE(FOR_static_steal_stolen, 1);
1263  init = (victim->u.p.ub -= 1);
1264  }
1265  __kmp_release_lock(lck, gtid);
1266 
1267  KMP_DEBUG_ASSERT(init + 1 <= limit);
1268  pr->u.p.parm4 = victimIdx; // remember victim to steal from
1269  status = 1;
1270  while_index = 0;
1271  // now update own count and ub with stolen range but init chunk
1272  __kmp_acquire_lock(pr->u.p.th_steal_lock, gtid);
1273  pr->u.p.count = init + 1;
1274  pr->u.p.ub = limit;
1275  __kmp_release_lock(pr->u.p.th_steal_lock, gtid);
1276  } // while (search for victim)
1277  } // if (try to find victim and steal)
1278  } else {
1279  // 4-byte induction variable, use 8-byte CAS for pair (count, ub)
1280  typedef union {
1281  struct {
1282  UT count;
1283  T ub;
1284  } p;
1285  kmp_int64 b;
1286  } union_i4;
1287  // All operations on 'count' or 'ub' must be combined atomically
1288  // together.
1289  {
1290  union_i4 vold, vnew;
1291  vold.b = *(volatile kmp_int64 *)(&pr->u.p.count);
1292  vnew = vold;
1293  vnew.p.count++;
1294  while (!KMP_COMPARE_AND_STORE_ACQ64(
1295  (volatile kmp_int64 *)&pr->u.p.count,
1296  *VOLATILE_CAST(kmp_int64 *) & vold.b,
1297  *VOLATILE_CAST(kmp_int64 *) & vnew.b)) {
1298  KMP_CPU_PAUSE();
1299  vold.b = *(volatile kmp_int64 *)(&pr->u.p.count);
1300  vnew = vold;
1301  vnew.p.count++;
1302  }
1303  vnew = vold;
1304  init = vnew.p.count;
1305  status = (init < (UT)vnew.p.ub);
1306  }
1307 
1308  if (!status) {
1309  kmp_info_t **other_threads = team->t.t_threads;
1310  int while_limit = pr->u.p.parm3;
1311  int while_index = 0;
1312  T id = pr->u.p.static_steal_counter; // loop id
1313  int idx = (th->th.th_dispatch->th_disp_index - 1) %
1314  __kmp_dispatch_num_buffers; // current loop index
1315  // note: victim thread can potentially execute another loop
1316  // TODO: algorithm of searching for a victim
1317  // should be cleaned up and measured
1318  while ((!status) && (while_limit != ++while_index)) {
1319  dispatch_private_info_template<T> *victim;
1320  union_i4 vold, vnew;
1321  kmp_int32 remaining;
1322  T victimIdx = pr->u.p.parm4;
1323  T oldVictimIdx = victimIdx ? victimIdx - 1 : nproc - 1;
1324  victim = reinterpret_cast<dispatch_private_info_template<T> *>(
1325  &other_threads[victimIdx]->th.th_dispatch->th_disp_buffer[idx]);
1326  KMP_DEBUG_ASSERT(victim);
1327  while ((victim == pr || id != victim->u.p.static_steal_counter) &&
1328  oldVictimIdx != victimIdx) {
1329  victimIdx = (victimIdx + 1) % nproc;
1330  victim = reinterpret_cast<dispatch_private_info_template<T> *>(
1331  &other_threads[victimIdx]->th.th_dispatch->th_disp_buffer[idx]);
1332  KMP_DEBUG_ASSERT(victim);
1333  }
1334  if (victim == pr || id != victim->u.p.static_steal_counter) {
1335  continue; // try once more (nproc attempts in total)
1336  // no victim is ready yet to participate in stealing
1337  // because no victim passed kmp_init_dispatch yet
1338  }
1339  pr->u.p.parm4 = victimIdx; // new victim found
1340  while (1) { // CAS loop if victim has enough chunks to steal
1341  vold.b = *(volatile kmp_int64 *)(&victim->u.p.count);
1342  vnew = vold;
1343 
1344  KMP_DEBUG_ASSERT((vnew.p.ub - 1) * (UT)chunk <= trip);
1345  if (vnew.p.count >= (UT)vnew.p.ub ||
1346  (remaining = vnew.p.ub - vnew.p.count) < 2) {
1347  pr->u.p.parm4 = (victimIdx + 1) % nproc; // shift start victim id
1348  break; // not enough chunks to steal, goto next victim
1349  }
1350  if (remaining > 3) {
1351  vnew.p.ub -= (remaining >> 2); // try to steal 1/4 of remaining
1352  } else {
1353  vnew.p.ub -= 1; // steal 1 chunk of 2 or 3 remaining
1354  }
1355  KMP_DEBUG_ASSERT((vnew.p.ub - 1) * (UT)chunk <= trip);
1356  // TODO: Should this be acquire or release?
1357  if (KMP_COMPARE_AND_STORE_ACQ64(
1358  (volatile kmp_int64 *)&victim->u.p.count,
1359  *VOLATILE_CAST(kmp_int64 *) & vold.b,
1360  *VOLATILE_CAST(kmp_int64 *) & vnew.b)) {
1361  // stealing succeeded
1362  KMP_COUNT_DEVELOPER_VALUE(FOR_static_steal_stolen,
1363  vold.p.ub - vnew.p.ub);
1364  status = 1;
1365  while_index = 0;
1366  // now update own count and ub
1367  init = vnew.p.ub;
1368  vold.p.count = init + 1;
1369 #if KMP_ARCH_X86
1370  KMP_XCHG_FIXED64((volatile kmp_int64 *)(&pr->u.p.count), vold.b);
1371 #else
1372  *(volatile kmp_int64 *)(&pr->u.p.count) = vold.b;
1373 #endif
1374  break;
1375  } // if (check CAS result)
1376  KMP_CPU_PAUSE(); // CAS failed, repeatedly attempt
1377  } // while (try to steal from particular victim)
1378  } // while (search for victim)
1379  } // if (try to find victim and steal)
1380  } // if (4-byte induction variable)
1381  if (!status) {
1382  *p_lb = 0;
1383  *p_ub = 0;
1384  if (p_st != NULL)
1385  *p_st = 0;
1386  } else {
1387  start = pr->u.p.parm2;
1388  init *= chunk;
1389  limit = chunk + init - 1;
1390  incr = pr->u.p.st;
1391  KMP_COUNT_DEVELOPER_VALUE(FOR_static_steal_chunks, 1);
1392 
1393  KMP_DEBUG_ASSERT(init <= trip);
1394  if ((last = (limit >= trip)) != 0)
1395  limit = trip;
1396  if (p_st != NULL)
1397  *p_st = incr;
1398 
1399  if (incr == 1) {
1400  *p_lb = start + init;
1401  *p_ub = start + limit;
1402  } else {
1403  *p_lb = start + init * incr;
1404  *p_ub = start + limit * incr;
1405  }
1406 
1407  if (pr->flags.ordered) {
1408  pr->u.p.ordered_lower = init;
1409  pr->u.p.ordered_upper = limit;
1410  } // if
1411  } // if
1412  break;
1413  } // case
1414 #endif // ( KMP_STATIC_STEAL_ENABLED )
1415  case kmp_sch_static_balanced: {
1416  KD_TRACE(
1417  10,
1418  ("__kmp_dispatch_next_algorithm: T#%d kmp_sch_static_balanced case\n",
1419  gtid));
1420  /* check if thread has any iteration to do */
1421  if ((status = !pr->u.p.count) != 0) {
1422  pr->u.p.count = 1;
1423  *p_lb = pr->u.p.lb;
1424  *p_ub = pr->u.p.ub;
1425  last = pr->u.p.parm1;
1426  if (p_st != NULL)
1427  *p_st = pr->u.p.st;
1428  } else { /* no iterations to do */
1429  pr->u.p.lb = pr->u.p.ub + pr->u.p.st;
1430  }
1431  } // case
1432  break;
1433  case kmp_sch_static_greedy: /* original code for kmp_sch_static_greedy was
1434  merged here */
1435  case kmp_sch_static_chunked: {
1436  T parm1;
1437 
1438  KD_TRACE(100, ("__kmp_dispatch_next_algorithm: T#%d "
1439  "kmp_sch_static_[affinity|chunked] case\n",
1440  gtid));
1441  parm1 = pr->u.p.parm1;
1442 
1443  trip = pr->u.p.tc - 1;
1444  init = parm1 * (pr->u.p.count + tid);
1445 
1446  if ((status = (init <= trip)) != 0) {
1447  start = pr->u.p.lb;
1448  incr = pr->u.p.st;
1449  limit = parm1 + init - 1;
1450 
1451  if ((last = (limit >= trip)) != 0)
1452  limit = trip;
1453 
1454  if (p_st != NULL)
1455  *p_st = incr;
1456 
1457  pr->u.p.count += nproc;
1458 
1459  if (incr == 1) {
1460  *p_lb = start + init;
1461  *p_ub = start + limit;
1462  } else {
1463  *p_lb = start + init * incr;
1464  *p_ub = start + limit * incr;
1465  }
1466 
1467  if (pr->flags.ordered) {
1468  pr->u.p.ordered_lower = init;
1469  pr->u.p.ordered_upper = limit;
1470  } // if
1471  } // if
1472  } // case
1473  break;
1474 
1475  case kmp_sch_dynamic_chunked: {
1476  T chunk = pr->u.p.parm1;
1477 
1478  KD_TRACE(
1479  100,
1480  ("__kmp_dispatch_next_algorithm: T#%d kmp_sch_dynamic_chunked case\n",
1481  gtid));
1482 
1483  init = chunk * test_then_inc_acq<ST>((volatile ST *)&sh->u.s.iteration);
1484  trip = pr->u.p.tc - 1;
1485 
1486  if ((status = (init <= trip)) == 0) {
1487  *p_lb = 0;
1488  *p_ub = 0;
1489  if (p_st != NULL)
1490  *p_st = 0;
1491  } else {
1492  start = pr->u.p.lb;
1493  limit = chunk + init - 1;
1494  incr = pr->u.p.st;
1495 
1496  if ((last = (limit >= trip)) != 0)
1497  limit = trip;
1498 
1499  if (p_st != NULL)
1500  *p_st = incr;
1501 
1502  if (incr == 1) {
1503  *p_lb = start + init;
1504  *p_ub = start + limit;
1505  } else {
1506  *p_lb = start + init * incr;
1507  *p_ub = start + limit * incr;
1508  }
1509 
1510  if (pr->flags.ordered) {
1511  pr->u.p.ordered_lower = init;
1512  pr->u.p.ordered_upper = limit;
1513  } // if
1514  } // if
1515  } // case
1516  break;
1517 
1518  case kmp_sch_guided_iterative_chunked: {
1519  T chunkspec = pr->u.p.parm1;
1520  KD_TRACE(100, ("__kmp_dispatch_next_algorithm: T#%d kmp_sch_guided_chunked "
1521  "iterative case\n",
1522  gtid));
1523  trip = pr->u.p.tc;
1524  // Start atomic part of calculations
1525  while (1) {
1526  ST remaining; // signed, because can be < 0
1527  init = sh->u.s.iteration; // shared value
1528  remaining = trip - init;
1529  if (remaining <= 0) { // AC: need to compare with 0 first
1530  // nothing to do, don't try atomic op
1531  status = 0;
1532  break;
1533  }
1534  if ((T)remaining <
1535  pr->u.p.parm2) { // compare with K*nproc*(chunk+1), K=2 by default
1536  // use dynamic-style schedule
1537  // atomically increment iterations, get old value
1538  init = test_then_add<ST>(RCAST(volatile ST *, &sh->u.s.iteration),
1539  (ST)chunkspec);
1540  remaining = trip - init;
1541  if (remaining <= 0) {
1542  status = 0; // all iterations got by other threads
1543  } else {
1544  // got some iterations to work on
1545  status = 1;
1546  if ((T)remaining > chunkspec) {
1547  limit = init + chunkspec - 1;
1548  } else {
1549  last = 1; // the last chunk
1550  limit = init + remaining - 1;
1551  } // if
1552  } // if
1553  break;
1554  } // if
1555  limit = init +
1556  (UT)(remaining * *(double *)&pr->u.p.parm3); // divide by K*nproc
1557  if (compare_and_swap<ST>(RCAST(volatile ST *, &sh->u.s.iteration),
1558  (ST)init, (ST)limit)) {
1559  // CAS was successful, chunk obtained
1560  status = 1;
1561  --limit;
1562  break;
1563  } // if
1564  } // while
1565  if (status != 0) {
1566  start = pr->u.p.lb;
1567  incr = pr->u.p.st;
1568  if (p_st != NULL)
1569  *p_st = incr;
1570  *p_lb = start + init * incr;
1571  *p_ub = start + limit * incr;
1572  if (pr->flags.ordered) {
1573  pr->u.p.ordered_lower = init;
1574  pr->u.p.ordered_upper = limit;
1575  } // if
1576  } else {
1577  *p_lb = 0;
1578  *p_ub = 0;
1579  if (p_st != NULL)
1580  *p_st = 0;
1581  } // if
1582  } // case
1583  break;
1584 
1585  case kmp_sch_guided_simd: {
1586  // same as iterative but curr-chunk adjusted to be multiple of given
1587  // chunk
1588  T chunk = pr->u.p.parm1;
1589  KD_TRACE(100,
1590  ("__kmp_dispatch_next_algorithm: T#%d kmp_sch_guided_simd case\n",
1591  gtid));
1592  trip = pr->u.p.tc;
1593  // Start atomic part of calculations
1594  while (1) {
1595  ST remaining; // signed, because can be < 0
1596  init = sh->u.s.iteration; // shared value
1597  remaining = trip - init;
1598  if (remaining <= 0) { // AC: need to compare with 0 first
1599  status = 0; // nothing to do, don't try atomic op
1600  break;
1601  }
1602  KMP_DEBUG_ASSERT(init % chunk == 0);
1603  // compare with K*nproc*(chunk+1), K=2 by default
1604  if ((T)remaining < pr->u.p.parm2) {
1605  // use dynamic-style schedule
1606  // atomically increment iterations, get old value
1607  init = test_then_add<ST>(RCAST(volatile ST *, &sh->u.s.iteration),
1608  (ST)chunk);
1609  remaining = trip - init;
1610  if (remaining <= 0) {
1611  status = 0; // all iterations got by other threads
1612  } else {
1613  // got some iterations to work on
1614  status = 1;
1615  if ((T)remaining > chunk) {
1616  limit = init + chunk - 1;
1617  } else {
1618  last = 1; // the last chunk
1619  limit = init + remaining - 1;
1620  } // if
1621  } // if
1622  break;
1623  } // if
1624  // divide by K*nproc
1625  UT span = remaining * (*(double *)&pr->u.p.parm3);
1626  UT rem = span % chunk;
1627  if (rem) // adjust so that span%chunk == 0
1628  span += chunk - rem;
1629  limit = init + span;
1630  if (compare_and_swap<ST>(RCAST(volatile ST *, &sh->u.s.iteration),
1631  (ST)init, (ST)limit)) {
1632  // CAS was successful, chunk obtained
1633  status = 1;
1634  --limit;
1635  break;
1636  } // if
1637  } // while
1638  if (status != 0) {
1639  start = pr->u.p.lb;
1640  incr = pr->u.p.st;
1641  if (p_st != NULL)
1642  *p_st = incr;
1643  *p_lb = start + init * incr;
1644  *p_ub = start + limit * incr;
1645  if (pr->flags.ordered) {
1646  pr->u.p.ordered_lower = init;
1647  pr->u.p.ordered_upper = limit;
1648  } // if
1649  } else {
1650  *p_lb = 0;
1651  *p_ub = 0;
1652  if (p_st != NULL)
1653  *p_st = 0;
1654  } // if
1655  } // case
1656  break;
1657 
1658  case kmp_sch_guided_analytical_chunked: {
1659  T chunkspec = pr->u.p.parm1;
1660  UT chunkIdx;
1661 #if KMP_USE_X87CONTROL
1662  /* for storing original FPCW value for Windows* OS on
1663  IA-32 architecture 8-byte version */
1664  unsigned int oldFpcw;
1665  unsigned int fpcwSet = 0;
1666 #endif
1667  KD_TRACE(100, ("__kmp_dispatch_next_algorithm: T#%d "
1668  "kmp_sch_guided_analytical_chunked case\n",
1669  gtid));
1670 
1671  trip = pr->u.p.tc;
1672 
1673  KMP_DEBUG_ASSERT(nproc > 1);
1674  KMP_DEBUG_ASSERT((2UL * chunkspec + 1) * (UT)nproc < trip);
1675 
1676  while (1) { /* this while loop is a safeguard against unexpected zero
1677  chunk sizes */
1678  chunkIdx = test_then_inc_acq<ST>((volatile ST *)&sh->u.s.iteration);
1679  if (chunkIdx >= (UT)pr->u.p.parm2) {
1680  --trip;
1681  /* use dynamic-style scheduling */
1682  init = chunkIdx * chunkspec + pr->u.p.count;
1683  /* need to verify init > 0 in case of overflow in the above
1684  * calculation */
1685  if ((status = (init > 0 && init <= trip)) != 0) {
1686  limit = init + chunkspec - 1;
1687 
1688  if ((last = (limit >= trip)) != 0)
1689  limit = trip;
1690  }
1691  break;
1692  } else {
1693 /* use exponential-style scheduling */
1694 /* The following check is to workaround the lack of long double precision on
1695  Windows* OS.
1696  This check works around the possible effect that init != 0 for chunkIdx == 0.
1697  */
1698 #if KMP_USE_X87CONTROL
1699  /* If we haven't already done so, save original
1700  FPCW and set precision to 64-bit, as Windows* OS
1701  on IA-32 architecture defaults to 53-bit */
1702  if (!fpcwSet) {
1703  oldFpcw = _control87(0, 0);
1704  _control87(_PC_64, _MCW_PC);
1705  fpcwSet = 0x30000;
1706  }
1707 #endif
1708  if (chunkIdx) {
1709  init = __kmp_dispatch_guided_remaining<T>(
1710  trip, *(DBL *)&pr->u.p.parm3, chunkIdx);
1711  KMP_DEBUG_ASSERT(init);
1712  init = trip - init;
1713  } else
1714  init = 0;
1715  limit = trip - __kmp_dispatch_guided_remaining<T>(
1716  trip, *(DBL *)&pr->u.p.parm3, chunkIdx + 1);
1717  KMP_ASSERT(init <= limit);
1718  if (init < limit) {
1719  KMP_DEBUG_ASSERT(limit <= trip);
1720  --limit;
1721  status = 1;
1722  break;
1723  } // if
1724  } // if
1725  } // while (1)
1726 #if KMP_USE_X87CONTROL
1727  /* restore FPCW if necessary
1728  AC: check fpcwSet flag first because oldFpcw can be uninitialized here
1729  */
1730  if (fpcwSet && (oldFpcw & fpcwSet))
1731  _control87(oldFpcw, _MCW_PC);
1732 #endif
1733  if (status != 0) {
1734  start = pr->u.p.lb;
1735  incr = pr->u.p.st;
1736  if (p_st != NULL)
1737  *p_st = incr;
1738  *p_lb = start + init * incr;
1739  *p_ub = start + limit * incr;
1740  if (pr->flags.ordered) {
1741  pr->u.p.ordered_lower = init;
1742  pr->u.p.ordered_upper = limit;
1743  }
1744  } else {
1745  *p_lb = 0;
1746  *p_ub = 0;
1747  if (p_st != NULL)
1748  *p_st = 0;
1749  }
1750  } // case
1751  break;
1752 
1753  case kmp_sch_trapezoidal: {
1754  UT index;
1755  T parm2 = pr->u.p.parm2;
1756  T parm3 = pr->u.p.parm3;
1757  T parm4 = pr->u.p.parm4;
1758  KD_TRACE(100,
1759  ("__kmp_dispatch_next_algorithm: T#%d kmp_sch_trapezoidal case\n",
1760  gtid));
1761 
1762  index = test_then_inc<ST>((volatile ST *)&sh->u.s.iteration);
1763 
1764  init = (index * ((2 * parm2) - (index - 1) * parm4)) / 2;
1765  trip = pr->u.p.tc - 1;
1766 
1767  if ((status = ((T)index < parm3 && init <= trip)) == 0) {
1768  *p_lb = 0;
1769  *p_ub = 0;
1770  if (p_st != NULL)
1771  *p_st = 0;
1772  } else {
1773  start = pr->u.p.lb;
1774  limit = ((index + 1) * (2 * parm2 - index * parm4)) / 2 - 1;
1775  incr = pr->u.p.st;
1776 
1777  if ((last = (limit >= trip)) != 0)
1778  limit = trip;
1779 
1780  if (p_st != NULL)
1781  *p_st = incr;
1782 
1783  if (incr == 1) {
1784  *p_lb = start + init;
1785  *p_ub = start + limit;
1786  } else {
1787  *p_lb = start + init * incr;
1788  *p_ub = start + limit * incr;
1789  }
1790 
1791  if (pr->flags.ordered) {
1792  pr->u.p.ordered_lower = init;
1793  pr->u.p.ordered_upper = limit;
1794  } // if
1795  } // if
1796  } // case
1797  break;
1798  default: {
1799  status = 0; // to avoid complaints on uninitialized variable use
1800  __kmp_fatal(KMP_MSG(UnknownSchedTypeDetected), // Primary message
1801  KMP_HNT(GetNewerLibrary), // Hint
1802  __kmp_msg_null // Variadic argument list terminator
1803  );
1804  } break;
1805  } // switch
1806  if (p_last)
1807  *p_last = last;
1808 #ifdef KMP_DEBUG
1809  if (pr->flags.ordered) {
1810  char *buff;
1811  // create format specifiers before the debug output
1812  buff = __kmp_str_format("__kmp_dispatch_next_algorithm: T#%%d "
1813  "ordered_lower:%%%s ordered_upper:%%%s\n",
1814  traits_t<UT>::spec, traits_t<UT>::spec);
1815  KD_TRACE(1000, (buff, gtid, pr->u.p.ordered_lower, pr->u.p.ordered_upper));
1816  __kmp_str_free(&buff);
1817  }
1818  {
1819  char *buff;
1820  // create format specifiers before the debug output
1821  buff = __kmp_str_format(
1822  "__kmp_dispatch_next_algorithm: T#%%d exit status:%%d p_last:%%d "
1823  "p_lb:%%%s p_ub:%%%s p_st:%%%s\n",
1824  traits_t<T>::spec, traits_t<T>::spec, traits_t<ST>::spec);
1825  KD_TRACE(10, (buff, gtid, status, *p_last, *p_lb, *p_ub, *p_st));
1826  __kmp_str_free(&buff);
1827  }
1828 #endif
1829  return status;
1830 }
1831 
1832 /* Define a macro for exiting __kmp_dispatch_next(). If status is 0 (no more
1833  work), then tell OMPT the loop is over. In some cases kmp_dispatch_fini()
1834  is not called. */
1835 #if OMPT_SUPPORT && OMPT_OPTIONAL
1836 #define OMPT_LOOP_END \
1837  if (status == 0) { \
1838  if (ompt_enabled.ompt_callback_work) { \
1839  ompt_team_info_t *team_info = __ompt_get_teaminfo(0, NULL); \
1840  ompt_task_info_t *task_info = __ompt_get_task_info_object(0); \
1841  ompt_callbacks.ompt_callback(ompt_callback_work)( \
1842  ompt_work_loop, ompt_scope_end, &(team_info->parallel_data), \
1843  &(task_info->task_data), 0, codeptr); \
1844  } \
1845  }
1846 // TODO: implement count
1847 #else
1848 #define OMPT_LOOP_END // no-op
1849 #endif
1850 
1851 #if KMP_STATS_ENABLED
1852 #define KMP_STATS_LOOP_END \
1853  { \
1854  kmp_int64 u, l, t, i; \
1855  l = (kmp_int64)(*p_lb); \
1856  u = (kmp_int64)(*p_ub); \
1857  i = (kmp_int64)(pr->u.p.st); \
1858  if (status == 0) { \
1859  t = 0; \
1860  KMP_POP_PARTITIONED_TIMER(); \
1861  } else if (i == 1) { \
1862  if (u >= l) \
1863  t = u - l + 1; \
1864  else \
1865  t = 0; \
1866  } else if (i < 0) { \
1867  if (l >= u) \
1868  t = (l - u) / (-i) + 1; \
1869  else \
1870  t = 0; \
1871  } else { \
1872  if (u >= l) \
1873  t = (u - l) / i + 1; \
1874  else \
1875  t = 0; \
1876  } \
1877  KMP_COUNT_VALUE(OMP_loop_dynamic_iterations, t); \
1878  }
1879 #else
1880 #define KMP_STATS_LOOP_END /* Nothing */
1881 #endif
1882 
1883 template <typename T>
1884 static int __kmp_dispatch_next(ident_t *loc, int gtid, kmp_int32 *p_last,
1885  T *p_lb, T *p_ub,
1886  typename traits_t<T>::signed_t *p_st
1887 #if OMPT_SUPPORT && OMPT_OPTIONAL
1888  ,
1889  void *codeptr
1890 #endif
1891  ) {
1892 
1893  typedef typename traits_t<T>::unsigned_t UT;
1894  typedef typename traits_t<T>::signed_t ST;
1895  // This is potentially slightly misleading, schedule(runtime) will appear here
1896  // even if the actual runtime schedule is static. (Which points out a
1897  // disadvantage of schedule(runtime): even when static scheduling is used it
1898  // costs more than a compile time choice to use static scheduling would.)
1899  KMP_TIME_PARTITIONED_BLOCK(OMP_loop_dynamic_scheduling);
1900 
1901  int status;
1902  dispatch_private_info_template<T> *pr;
1903  kmp_info_t *th = __kmp_threads[gtid];
1904  kmp_team_t *team = th->th.th_team;
1905 
1906  KMP_DEBUG_ASSERT(p_lb && p_ub && p_st); // AC: these cannot be NULL
1907  KD_TRACE(
1908  1000,
1909  ("__kmp_dispatch_next: T#%d called p_lb:%p p_ub:%p p_st:%p p_last: %p\n",
1910  gtid, p_lb, p_ub, p_st, p_last));
1911 
1912  if (team->t.t_serialized) {
1913  /* NOTE: serialize this dispatch because we are not at the active level */
1914  pr = reinterpret_cast<dispatch_private_info_template<T> *>(
1915  th->th.th_dispatch->th_disp_buffer); /* top of the stack */
1916  KMP_DEBUG_ASSERT(pr);
1917 
1918  if ((status = (pr->u.p.tc != 0)) == 0) {
1919  *p_lb = 0;
1920  *p_ub = 0;
1921  // if ( p_last != NULL )
1922  // *p_last = 0;
1923  if (p_st != NULL)
1924  *p_st = 0;
1925  if (__kmp_env_consistency_check) {
1926  if (pr->pushed_ws != ct_none) {
1927  pr->pushed_ws = __kmp_pop_workshare(gtid, pr->pushed_ws, loc);
1928  }
1929  }
1930  } else if (pr->flags.nomerge) {
1931  kmp_int32 last;
1932  T start;
1933  UT limit, trip, init;
1934  ST incr;
1935  T chunk = pr->u.p.parm1;
1936 
1937  KD_TRACE(100, ("__kmp_dispatch_next: T#%d kmp_sch_dynamic_chunked case\n",
1938  gtid));
1939 
1940  init = chunk * pr->u.p.count++;
1941  trip = pr->u.p.tc - 1;
1942 
1943  if ((status = (init <= trip)) == 0) {
1944  *p_lb = 0;
1945  *p_ub = 0;
1946  // if ( p_last != NULL )
1947  // *p_last = 0;
1948  if (p_st != NULL)
1949  *p_st = 0;
1950  if (__kmp_env_consistency_check) {
1951  if (pr->pushed_ws != ct_none) {
1952  pr->pushed_ws = __kmp_pop_workshare(gtid, pr->pushed_ws, loc);
1953  }
1954  }
1955  } else {
1956  start = pr->u.p.lb;
1957  limit = chunk + init - 1;
1958  incr = pr->u.p.st;
1959 
1960  if ((last = (limit >= trip)) != 0) {
1961  limit = trip;
1962 #if KMP_OS_WINDOWS
1963  pr->u.p.last_upper = pr->u.p.ub;
1964 #endif /* KMP_OS_WINDOWS */
1965  }
1966  if (p_last != NULL)
1967  *p_last = last;
1968  if (p_st != NULL)
1969  *p_st = incr;
1970  if (incr == 1) {
1971  *p_lb = start + init;
1972  *p_ub = start + limit;
1973  } else {
1974  *p_lb = start + init * incr;
1975  *p_ub = start + limit * incr;
1976  }
1977 
1978  if (pr->flags.ordered) {
1979  pr->u.p.ordered_lower = init;
1980  pr->u.p.ordered_upper = limit;
1981 #ifdef KMP_DEBUG
1982  {
1983  char *buff;
1984  // create format specifiers before the debug output
1985  buff = __kmp_str_format("__kmp_dispatch_next: T#%%d "
1986  "ordered_lower:%%%s ordered_upper:%%%s\n",
1987  traits_t<UT>::spec, traits_t<UT>::spec);
1988  KD_TRACE(1000, (buff, gtid, pr->u.p.ordered_lower,
1989  pr->u.p.ordered_upper));
1990  __kmp_str_free(&buff);
1991  }
1992 #endif
1993  } // if
1994  } // if
1995  } else {
1996  pr->u.p.tc = 0;
1997  *p_lb = pr->u.p.lb;
1998  *p_ub = pr->u.p.ub;
1999 #if KMP_OS_WINDOWS
2000  pr->u.p.last_upper = *p_ub;
2001 #endif /* KMP_OS_WINDOWS */
2002  if (p_last != NULL)
2003  *p_last = TRUE;
2004  if (p_st != NULL)
2005  *p_st = pr->u.p.st;
2006  } // if
2007 #ifdef KMP_DEBUG
2008  {
2009  char *buff;
2010  // create format specifiers before the debug output
2011  buff = __kmp_str_format(
2012  "__kmp_dispatch_next: T#%%d serialized case: p_lb:%%%s "
2013  "p_ub:%%%s p_st:%%%s p_last:%%p %%d returning:%%d\n",
2014  traits_t<T>::spec, traits_t<T>::spec, traits_t<ST>::spec);
2015  KD_TRACE(10, (buff, gtid, *p_lb, *p_ub, *p_st, p_last, *p_last, status));
2016  __kmp_str_free(&buff);
2017  }
2018 #endif
2019 #if INCLUDE_SSC_MARKS
2020  SSC_MARK_DISPATCH_NEXT();
2021 #endif
2022  OMPT_LOOP_END;
2023  KMP_STATS_LOOP_END;
2024  return status;
2025  } else {
2026  kmp_int32 last = 0;
2027  dispatch_shared_info_template<T> volatile *sh;
2028 
2029  KMP_DEBUG_ASSERT(th->th.th_dispatch ==
2030  &th->th.th_team->t.t_dispatch[th->th.th_info.ds.ds_tid]);
2031 
2032  pr = reinterpret_cast<dispatch_private_info_template<T> *>(
2033  th->th.th_dispatch->th_dispatch_pr_current);
2034  KMP_DEBUG_ASSERT(pr);
2035  sh = reinterpret_cast<dispatch_shared_info_template<T> volatile *>(
2036  th->th.th_dispatch->th_dispatch_sh_current);
2037  KMP_DEBUG_ASSERT(sh);
2038 
2039 #if KMP_USE_HIER_SCHED
2040  if (pr->flags.use_hier)
2041  status = sh->hier->next(loc, gtid, pr, &last, p_lb, p_ub, p_st);
2042  else
2043 #endif // KMP_USE_HIER_SCHED
2044  status = __kmp_dispatch_next_algorithm<T>(gtid, pr, sh, &last, p_lb, p_ub,
2045  p_st, th->th.th_team_nproc,
2046  th->th.th_info.ds.ds_tid);
2047  // status == 0: no more iterations to execute
2048  if (status == 0) {
2049  UT num_done;
2050 
2051  num_done = test_then_inc<ST>((volatile ST *)&sh->u.s.num_done);
2052 #ifdef KMP_DEBUG
2053  {
2054  char *buff;
2055  // create format specifiers before the debug output
2056  buff = __kmp_str_format(
2057  "__kmp_dispatch_next: T#%%d increment num_done:%%%s\n",
2058  traits_t<UT>::spec);
2059  KD_TRACE(10, (buff, gtid, sh->u.s.num_done));
2060  __kmp_str_free(&buff);
2061  }
2062 #endif
2063 
2064 #if KMP_USE_HIER_SCHED
2065  pr->flags.use_hier = FALSE;
2066 #endif
2067  if ((ST)num_done == th->th.th_team_nproc - 1) {
2068 #if (KMP_STATIC_STEAL_ENABLED)
2069  if (pr->schedule == kmp_sch_static_steal &&
2070  traits_t<T>::type_size > 4) {
2071  int i;
2072  int idx = (th->th.th_dispatch->th_disp_index - 1) %
2073  __kmp_dispatch_num_buffers; // current loop index
2074  kmp_info_t **other_threads = team->t.t_threads;
2075  // loop complete, safe to destroy locks used for stealing
2076  for (i = 0; i < th->th.th_team_nproc; ++i) {
2077  dispatch_private_info_template<T> *buf =
2078  reinterpret_cast<dispatch_private_info_template<T> *>(
2079  &other_threads[i]->th.th_dispatch->th_disp_buffer[idx]);
2080  kmp_lock_t *lck = buf->u.p.th_steal_lock;
2081  KMP_ASSERT(lck != NULL);
2082  __kmp_destroy_lock(lck);
2083  __kmp_free(lck);
2084  buf->u.p.th_steal_lock = NULL;
2085  }
2086  }
2087 #endif
2088  /* NOTE: release this buffer to be reused */
2089 
2090  KMP_MB(); /* Flush all pending memory write invalidates. */
2091 
2092  sh->u.s.num_done = 0;
2093  sh->u.s.iteration = 0;
2094 
2095  /* TODO replace with general release procedure? */
2096  if (pr->flags.ordered) {
2097  sh->u.s.ordered_iteration = 0;
2098  }
2099 
2100  KMP_MB(); /* Flush all pending memory write invalidates. */
2101 
2102  sh->buffer_index += __kmp_dispatch_num_buffers;
2103  KD_TRACE(100, ("__kmp_dispatch_next: T#%d change buffer_index:%d\n",
2104  gtid, sh->buffer_index));
2105 
2106  KMP_MB(); /* Flush all pending memory write invalidates. */
2107 
2108  } // if
2109  if (__kmp_env_consistency_check) {
2110  if (pr->pushed_ws != ct_none) {
2111  pr->pushed_ws = __kmp_pop_workshare(gtid, pr->pushed_ws, loc);
2112  }
2113  }
2114 
2115  th->th.th_dispatch->th_deo_fcn = NULL;
2116  th->th.th_dispatch->th_dxo_fcn = NULL;
2117  th->th.th_dispatch->th_dispatch_sh_current = NULL;
2118  th->th.th_dispatch->th_dispatch_pr_current = NULL;
2119  } // if (status == 0)
2120 #if KMP_OS_WINDOWS
2121  else if (last) {
2122  pr->u.p.last_upper = pr->u.p.ub;
2123  }
2124 #endif /* KMP_OS_WINDOWS */
2125  if (p_last != NULL && status != 0)
2126  *p_last = last;
2127  } // if
2128 
2129 #ifdef KMP_DEBUG
2130  {
2131  char *buff;
2132  // create format specifiers before the debug output
2133  buff = __kmp_str_format(
2134  "__kmp_dispatch_next: T#%%d normal case: "
2135  "p_lb:%%%s p_ub:%%%s p_st:%%%s p_last:%%p (%%d) returning:%%d\n",
2136  traits_t<T>::spec, traits_t<T>::spec, traits_t<ST>::spec);
2137  KD_TRACE(10, (buff, gtid, *p_lb, *p_ub, p_st ? *p_st : 0, p_last,
2138  (p_last ? *p_last : 0), status));
2139  __kmp_str_free(&buff);
2140  }
2141 #endif
2142 #if INCLUDE_SSC_MARKS
2143  SSC_MARK_DISPATCH_NEXT();
2144 #endif
2145  OMPT_LOOP_END;
2146  KMP_STATS_LOOP_END;
2147  return status;
2148 }
2149 
2150 template <typename T>
2151 static void __kmp_dist_get_bounds(ident_t *loc, kmp_int32 gtid,
2152  kmp_int32 *plastiter, T *plower, T *pupper,
2153  typename traits_t<T>::signed_t incr) {
2154  typedef typename traits_t<T>::unsigned_t UT;
2155  kmp_uint32 team_id;
2156  kmp_uint32 nteams;
2157  UT trip_count;
2158  kmp_team_t *team;
2159  kmp_info_t *th;
2160 
2161  KMP_DEBUG_ASSERT(plastiter && plower && pupper);
2162  KE_TRACE(10, ("__kmpc_dist_get_bounds called (%d)\n", gtid));
2163 #ifdef KMP_DEBUG
2164  typedef typename traits_t<T>::signed_t ST;
2165  {
2166  char *buff;
2167  // create format specifiers before the debug output
2168  buff = __kmp_str_format("__kmpc_dist_get_bounds: T#%%d liter=%%d "
2169  "iter=(%%%s, %%%s, %%%s) signed?<%s>\n",
2170  traits_t<T>::spec, traits_t<T>::spec,
2171  traits_t<ST>::spec, traits_t<T>::spec);
2172  KD_TRACE(100, (buff, gtid, *plastiter, *plower, *pupper, incr));
2173  __kmp_str_free(&buff);
2174  }
2175 #endif
2176 
2177  if (__kmp_env_consistency_check) {
2178  if (incr == 0) {
2179  __kmp_error_construct(kmp_i18n_msg_CnsLoopIncrZeroProhibited, ct_pdo,
2180  loc);
2181  }
2182  if (incr > 0 ? (*pupper < *plower) : (*plower < *pupper)) {
2183  // The loop is illegal.
2184  // Some zero-trip loops maintained by compiler, e.g.:
2185  // for(i=10;i<0;++i) // lower >= upper - run-time check
2186  // for(i=0;i>10;--i) // lower <= upper - run-time check
2187  // for(i=0;i>10;++i) // incr > 0 - compile-time check
2188  // for(i=10;i<0;--i) // incr < 0 - compile-time check
2189  // Compiler does not check the following illegal loops:
2190  // for(i=0;i<10;i+=incr) // where incr<0
2191  // for(i=10;i>0;i-=incr) // where incr<0
2192  __kmp_error_construct(kmp_i18n_msg_CnsLoopIncrIllegal, ct_pdo, loc);
2193  }
2194  }
2195  th = __kmp_threads[gtid];
2196  team = th->th.th_team;
2197  KMP_DEBUG_ASSERT(th->th.th_teams_microtask); // we are in the teams construct
2198  nteams = th->th.th_teams_size.nteams;
2199  team_id = team->t.t_master_tid;
2200  KMP_DEBUG_ASSERT(nteams == (kmp_uint32)team->t.t_parent->t.t_nproc);
2201 
2202  // compute global trip count
2203  if (incr == 1) {
2204  trip_count = *pupper - *plower + 1;
2205  } else if (incr == -1) {
2206  trip_count = *plower - *pupper + 1;
2207  } else if (incr > 0) {
2208  // upper-lower can exceed the limit of signed type
2209  trip_count = (UT)(*pupper - *plower) / incr + 1;
2210  } else {
2211  trip_count = (UT)(*plower - *pupper) / (-incr) + 1;
2212  }
2213 
2214  if (trip_count <= nteams) {
2215  KMP_DEBUG_ASSERT(
2216  __kmp_static == kmp_sch_static_greedy ||
2217  __kmp_static ==
2218  kmp_sch_static_balanced); // Unknown static scheduling type.
2219  // only some teams get single iteration, others get nothing
2220  if (team_id < trip_count) {
2221  *pupper = *plower = *plower + team_id * incr;
2222  } else {
2223  *plower = *pupper + incr; // zero-trip loop
2224  }
2225  if (plastiter != NULL)
2226  *plastiter = (team_id == trip_count - 1);
2227  } else {
2228  if (__kmp_static == kmp_sch_static_balanced) {
2229  UT chunk = trip_count / nteams;
2230  UT extras = trip_count % nteams;
2231  *plower +=
2232  incr * (team_id * chunk + (team_id < extras ? team_id : extras));
2233  *pupper = *plower + chunk * incr - (team_id < extras ? 0 : incr);
2234  if (plastiter != NULL)
2235  *plastiter = (team_id == nteams - 1);
2236  } else {
2237  T chunk_inc_count =
2238  (trip_count / nteams + ((trip_count % nteams) ? 1 : 0)) * incr;
2239  T upper = *pupper;
2240  KMP_DEBUG_ASSERT(__kmp_static == kmp_sch_static_greedy);
2241  // Unknown static scheduling type.
2242  *plower += team_id * chunk_inc_count;
2243  *pupper = *plower + chunk_inc_count - incr;
2244  // Check/correct bounds if needed
2245  if (incr > 0) {
2246  if (*pupper < *plower)
2247  *pupper = traits_t<T>::max_value;
2248  if (plastiter != NULL)
2249  *plastiter = *plower <= upper && *pupper > upper - incr;
2250  if (*pupper > upper)
2251  *pupper = upper; // tracker C73258
2252  } else {
2253  if (*pupper > *plower)
2254  *pupper = traits_t<T>::min_value;
2255  if (plastiter != NULL)
2256  *plastiter = *plower >= upper && *pupper < upper - incr;
2257  if (*pupper < upper)
2258  *pupper = upper; // tracker C73258
2259  }
2260  }
2261  }
2262 }
2263 
2264 //-----------------------------------------------------------------------------
2265 // Dispatch routines
2266 // Transfer call to template< type T >
2267 // __kmp_dispatch_init( ident_t *loc, int gtid, enum sched_type schedule,
2268 // T lb, T ub, ST st, ST chunk )
2269 extern "C" {
2270 
2287 void __kmpc_dispatch_init_4(ident_t *loc, kmp_int32 gtid,
2288  enum sched_type schedule, kmp_int32 lb,
2289  kmp_int32 ub, kmp_int32 st, kmp_int32 chunk) {
2290  KMP_DEBUG_ASSERT(__kmp_init_serial);
2291 #if OMPT_SUPPORT && OMPT_OPTIONAL
2292  OMPT_STORE_RETURN_ADDRESS(gtid);
2293 #endif
2294  __kmp_dispatch_init<kmp_int32>(loc, gtid, schedule, lb, ub, st, chunk, true);
2295 }
2299 void __kmpc_dispatch_init_4u(ident_t *loc, kmp_int32 gtid,
2300  enum sched_type schedule, kmp_uint32 lb,
2301  kmp_uint32 ub, kmp_int32 st, kmp_int32 chunk) {
2302  KMP_DEBUG_ASSERT(__kmp_init_serial);
2303 #if OMPT_SUPPORT && OMPT_OPTIONAL
2304  OMPT_STORE_RETURN_ADDRESS(gtid);
2305 #endif
2306  __kmp_dispatch_init<kmp_uint32>(loc, gtid, schedule, lb, ub, st, chunk, true);
2307 }
2308 
2312 void __kmpc_dispatch_init_8(ident_t *loc, kmp_int32 gtid,
2313  enum sched_type schedule, kmp_int64 lb,
2314  kmp_int64 ub, kmp_int64 st, kmp_int64 chunk) {
2315  KMP_DEBUG_ASSERT(__kmp_init_serial);
2316 #if OMPT_SUPPORT && OMPT_OPTIONAL
2317  OMPT_STORE_RETURN_ADDRESS(gtid);
2318 #endif
2319  __kmp_dispatch_init<kmp_int64>(loc, gtid, schedule, lb, ub, st, chunk, true);
2320 }
2321 
2325 void __kmpc_dispatch_init_8u(ident_t *loc, kmp_int32 gtid,
2326  enum sched_type schedule, kmp_uint64 lb,
2327  kmp_uint64 ub, kmp_int64 st, kmp_int64 chunk) {
2328  KMP_DEBUG_ASSERT(__kmp_init_serial);
2329 #if OMPT_SUPPORT && OMPT_OPTIONAL
2330  OMPT_STORE_RETURN_ADDRESS(gtid);
2331 #endif
2332  __kmp_dispatch_init<kmp_uint64>(loc, gtid, schedule, lb, ub, st, chunk, true);
2333 }
2334 
2344 void __kmpc_dist_dispatch_init_4(ident_t *loc, kmp_int32 gtid,
2345  enum sched_type schedule, kmp_int32 *p_last,
2346  kmp_int32 lb, kmp_int32 ub, kmp_int32 st,
2347  kmp_int32 chunk) {
2348  KMP_DEBUG_ASSERT(__kmp_init_serial);
2349 #if OMPT_SUPPORT && OMPT_OPTIONAL
2350  OMPT_STORE_RETURN_ADDRESS(gtid);
2351 #endif
2352  __kmp_dist_get_bounds<kmp_int32>(loc, gtid, p_last, &lb, &ub, st);
2353  __kmp_dispatch_init<kmp_int32>(loc, gtid, schedule, lb, ub, st, chunk, true);
2354 }
2355 
2356 void __kmpc_dist_dispatch_init_4u(ident_t *loc, kmp_int32 gtid,
2357  enum sched_type schedule, kmp_int32 *p_last,
2358  kmp_uint32 lb, kmp_uint32 ub, kmp_int32 st,
2359  kmp_int32 chunk) {
2360  KMP_DEBUG_ASSERT(__kmp_init_serial);
2361 #if OMPT_SUPPORT && OMPT_OPTIONAL
2362  OMPT_STORE_RETURN_ADDRESS(gtid);
2363 #endif
2364  __kmp_dist_get_bounds<kmp_uint32>(loc, gtid, p_last, &lb, &ub, st);
2365  __kmp_dispatch_init<kmp_uint32>(loc, gtid, schedule, lb, ub, st, chunk, true);
2366 }
2367 
2368 void __kmpc_dist_dispatch_init_8(ident_t *loc, kmp_int32 gtid,
2369  enum sched_type schedule, kmp_int32 *p_last,
2370  kmp_int64 lb, kmp_int64 ub, kmp_int64 st,
2371  kmp_int64 chunk) {
2372  KMP_DEBUG_ASSERT(__kmp_init_serial);
2373 #if OMPT_SUPPORT && OMPT_OPTIONAL
2374  OMPT_STORE_RETURN_ADDRESS(gtid);
2375 #endif
2376  __kmp_dist_get_bounds<kmp_int64>(loc, gtid, p_last, &lb, &ub, st);
2377  __kmp_dispatch_init<kmp_int64>(loc, gtid, schedule, lb, ub, st, chunk, true);
2378 }
2379 
2380 void __kmpc_dist_dispatch_init_8u(ident_t *loc, kmp_int32 gtid,
2381  enum sched_type schedule, kmp_int32 *p_last,
2382  kmp_uint64 lb, kmp_uint64 ub, kmp_int64 st,
2383  kmp_int64 chunk) {
2384  KMP_DEBUG_ASSERT(__kmp_init_serial);
2385 #if OMPT_SUPPORT && OMPT_OPTIONAL
2386  OMPT_STORE_RETURN_ADDRESS(gtid);
2387 #endif
2388  __kmp_dist_get_bounds<kmp_uint64>(loc, gtid, p_last, &lb, &ub, st);
2389  __kmp_dispatch_init<kmp_uint64>(loc, gtid, schedule, lb, ub, st, chunk, true);
2390 }
2391 
2405 int __kmpc_dispatch_next_4(ident_t *loc, kmp_int32 gtid, kmp_int32 *p_last,
2406  kmp_int32 *p_lb, kmp_int32 *p_ub, kmp_int32 *p_st) {
2407 #if OMPT_SUPPORT && OMPT_OPTIONAL
2408  OMPT_STORE_RETURN_ADDRESS(gtid);
2409 #endif
2410  return __kmp_dispatch_next<kmp_int32>(loc, gtid, p_last, p_lb, p_ub, p_st
2411 #if OMPT_SUPPORT && OMPT_OPTIONAL
2412  ,
2413  OMPT_LOAD_RETURN_ADDRESS(gtid)
2414 #endif
2415  );
2416 }
2417 
2421 int __kmpc_dispatch_next_4u(ident_t *loc, kmp_int32 gtid, kmp_int32 *p_last,
2422  kmp_uint32 *p_lb, kmp_uint32 *p_ub,
2423  kmp_int32 *p_st) {
2424 #if OMPT_SUPPORT && OMPT_OPTIONAL
2425  OMPT_STORE_RETURN_ADDRESS(gtid);
2426 #endif
2427  return __kmp_dispatch_next<kmp_uint32>(loc, gtid, p_last, p_lb, p_ub, p_st
2428 #if OMPT_SUPPORT && OMPT_OPTIONAL
2429  ,
2430  OMPT_LOAD_RETURN_ADDRESS(gtid)
2431 #endif
2432  );
2433 }
2434 
2438 int __kmpc_dispatch_next_8(ident_t *loc, kmp_int32 gtid, kmp_int32 *p_last,
2439  kmp_int64 *p_lb, kmp_int64 *p_ub, kmp_int64 *p_st) {
2440 #if OMPT_SUPPORT && OMPT_OPTIONAL
2441  OMPT_STORE_RETURN_ADDRESS(gtid);
2442 #endif
2443  return __kmp_dispatch_next<kmp_int64>(loc, gtid, p_last, p_lb, p_ub, p_st
2444 #if OMPT_SUPPORT && OMPT_OPTIONAL
2445  ,
2446  OMPT_LOAD_RETURN_ADDRESS(gtid)
2447 #endif
2448  );
2449 }
2450 
2454 int __kmpc_dispatch_next_8u(ident_t *loc, kmp_int32 gtid, kmp_int32 *p_last,
2455  kmp_uint64 *p_lb, kmp_uint64 *p_ub,
2456  kmp_int64 *p_st) {
2457 #if OMPT_SUPPORT && OMPT_OPTIONAL
2458  OMPT_STORE_RETURN_ADDRESS(gtid);
2459 #endif
2460  return __kmp_dispatch_next<kmp_uint64>(loc, gtid, p_last, p_lb, p_ub, p_st
2461 #if OMPT_SUPPORT && OMPT_OPTIONAL
2462  ,
2463  OMPT_LOAD_RETURN_ADDRESS(gtid)
2464 #endif
2465  );
2466 }
2467 
2474 void __kmpc_dispatch_fini_4(ident_t *loc, kmp_int32 gtid) {
2475  __kmp_dispatch_finish<kmp_uint32>(gtid, loc);
2476 }
2477 
2481 void __kmpc_dispatch_fini_8(ident_t *loc, kmp_int32 gtid) {
2482  __kmp_dispatch_finish<kmp_uint64>(gtid, loc);
2483 }
2484 
2488 void __kmpc_dispatch_fini_4u(ident_t *loc, kmp_int32 gtid) {
2489  __kmp_dispatch_finish<kmp_uint32>(gtid, loc);
2490 }
2491 
2495 void __kmpc_dispatch_fini_8u(ident_t *loc, kmp_int32 gtid) {
2496  __kmp_dispatch_finish<kmp_uint64>(gtid, loc);
2497 }
2500 //-----------------------------------------------------------------------------
2501 // Non-template routines from kmp_dispatch.cpp used in other sources
2502 
2503 kmp_uint32 __kmp_eq_4(kmp_uint32 value, kmp_uint32 checker) {
2504  return value == checker;
2505 }
2506 
2507 kmp_uint32 __kmp_neq_4(kmp_uint32 value, kmp_uint32 checker) {
2508  return value != checker;
2509 }
2510 
2511 kmp_uint32 __kmp_lt_4(kmp_uint32 value, kmp_uint32 checker) {
2512  return value < checker;
2513 }
2514 
2515 kmp_uint32 __kmp_ge_4(kmp_uint32 value, kmp_uint32 checker) {
2516  return value >= checker;
2517 }
2518 
2519 kmp_uint32 __kmp_le_4(kmp_uint32 value, kmp_uint32 checker) {
2520  return value <= checker;
2521 }
2522 
2523 kmp_uint32
2524 __kmp_wait_4(volatile kmp_uint32 *spinner, kmp_uint32 checker,
2525  kmp_uint32 (*pred)(kmp_uint32, kmp_uint32),
2526  void *obj // Higher-level synchronization object, or NULL.
2527  ) {
2528  // note: we may not belong to a team at this point
2529  volatile kmp_uint32 *spin = spinner;
2530  kmp_uint32 check = checker;
2531  kmp_uint32 spins;
2532  kmp_uint32 (*f)(kmp_uint32, kmp_uint32) = pred;
2533  kmp_uint32 r;
2534 
2535  KMP_FSYNC_SPIN_INIT(obj, CCAST(kmp_uint32 *, spin));
2536  KMP_INIT_YIELD(spins);
2537  // main wait spin loop
2538  while (!f(r = TCR_4(*spin), check)) {
2539  KMP_FSYNC_SPIN_PREPARE(obj);
2540  /* GEH - remove this since it was accidentally introduced when kmp_wait was
2541  split. It causes problems with infinite recursion because of exit lock */
2542  /* if ( TCR_4(__kmp_global.g.g_done) && __kmp_global.g.g_abort)
2543  __kmp_abort_thread(); */
2544  KMP_YIELD_OVERSUB_ELSE_SPIN(spins);
2545  }
2546  KMP_FSYNC_SPIN_ACQUIRED(obj);
2547  return r;
2548 }
2549 
2550 void __kmp_wait_4_ptr(void *spinner, kmp_uint32 checker,
2551  kmp_uint32 (*pred)(void *, kmp_uint32),
2552  void *obj // Higher-level synchronization object, or NULL.
2553  ) {
2554  // note: we may not belong to a team at this point
2555  void *spin = spinner;
2556  kmp_uint32 check = checker;
2557  kmp_uint32 spins;
2558  kmp_uint32 (*f)(void *, kmp_uint32) = pred;
2559 
2560  KMP_FSYNC_SPIN_INIT(obj, spin);
2561  KMP_INIT_YIELD(spins);
2562  // main wait spin loop
2563  while (!f(spin, check)) {
2564  KMP_FSYNC_SPIN_PREPARE(obj);
2565  /* if we have waited a bit, or are noversubscribed, yield */
2566  /* pause is in the following code */
2567  KMP_YIELD_OVERSUB_ELSE_SPIN(spins);
2568  }
2569  KMP_FSYNC_SPIN_ACQUIRED(obj);
2570 }
2571 
2572 } // extern "C"
2573 
2574 #ifdef KMP_GOMP_COMPAT
2575 
2576 void __kmp_aux_dispatch_init_4(ident_t *loc, kmp_int32 gtid,
2577  enum sched_type schedule, kmp_int32 lb,
2578  kmp_int32 ub, kmp_int32 st, kmp_int32 chunk,
2579  int push_ws) {
2580  __kmp_dispatch_init<kmp_int32>(loc, gtid, schedule, lb, ub, st, chunk,
2581  push_ws);
2582 }
2583 
2584 void __kmp_aux_dispatch_init_4u(ident_t *loc, kmp_int32 gtid,
2585  enum sched_type schedule, kmp_uint32 lb,
2586  kmp_uint32 ub, kmp_int32 st, kmp_int32 chunk,
2587  int push_ws) {
2588  __kmp_dispatch_init<kmp_uint32>(loc, gtid, schedule, lb, ub, st, chunk,
2589  push_ws);
2590 }
2591 
2592 void __kmp_aux_dispatch_init_8(ident_t *loc, kmp_int32 gtid,
2593  enum sched_type schedule, kmp_int64 lb,
2594  kmp_int64 ub, kmp_int64 st, kmp_int64 chunk,
2595  int push_ws) {
2596  __kmp_dispatch_init<kmp_int64>(loc, gtid, schedule, lb, ub, st, chunk,
2597  push_ws);
2598 }
2599 
2600 void __kmp_aux_dispatch_init_8u(ident_t *loc, kmp_int32 gtid,
2601  enum sched_type schedule, kmp_uint64 lb,
2602  kmp_uint64 ub, kmp_int64 st, kmp_int64 chunk,
2603  int push_ws) {
2604  __kmp_dispatch_init<kmp_uint64>(loc, gtid, schedule, lb, ub, st, chunk,
2605  push_ws);
2606 }
2607 
2608 void __kmp_aux_dispatch_fini_chunk_4(ident_t *loc, kmp_int32 gtid) {
2609  __kmp_dispatch_finish_chunk<kmp_uint32>(gtid, loc);
2610 }
2611 
2612 void __kmp_aux_dispatch_fini_chunk_8(ident_t *loc, kmp_int32 gtid) {
2613  __kmp_dispatch_finish_chunk<kmp_uint64>(gtid, loc);
2614 }
2615 
2616 void __kmp_aux_dispatch_fini_chunk_4u(ident_t *loc, kmp_int32 gtid) {
2617  __kmp_dispatch_finish_chunk<kmp_uint32>(gtid, loc);
2618 }
2619 
2620 void __kmp_aux_dispatch_fini_chunk_8u(ident_t *loc, kmp_int32 gtid) {
2621  __kmp_dispatch_finish_chunk<kmp_uint64>(gtid, loc);
2622 }
2623 
2624 #endif /* KMP_GOMP_COMPAT */
2625 
2626 /* ------------------------------------------------------------------------ */
#define KMP_COUNT_VALUE(name, value)
Adds value to specified timer (name).
Definition: kmp_stats.h:887
#define KMP_COUNT_BLOCK(name)
Increments specified counter (name).
Definition: kmp_stats.h:900
int __kmpc_dispatch_next_4(ident_t *loc, kmp_int32 gtid, kmp_int32 *p_last, kmp_int32 *p_lb, kmp_int32 *p_ub, kmp_int32 *p_st)
sched_type
Definition: kmp.h:336
void __kmpc_dispatch_fini_4(ident_t *loc, kmp_int32 gtid)
void __kmpc_dist_dispatch_init_4(ident_t *loc, kmp_int32 gtid, enum sched_type schedule, kmp_int32 *p_last, kmp_int32 lb, kmp_int32 ub, kmp_int32 st, kmp_int32 chunk)
int __kmpc_dispatch_next_4u(ident_t *loc, kmp_int32 gtid, kmp_int32 *p_last, kmp_uint32 *p_lb, kmp_uint32 *p_ub, kmp_int32 *p_st)
int __kmpc_dispatch_next_8u(ident_t *loc, kmp_int32 gtid, kmp_int32 *p_last, kmp_uint64 *p_lb, kmp_uint64 *p_ub, kmp_int64 *p_st)
void __kmpc_dispatch_fini_8(ident_t *loc, kmp_int32 gtid)
int __kmpc_dispatch_next_8(ident_t *loc, kmp_int32 gtid, kmp_int32 *p_last, kmp_int64 *p_lb, kmp_int64 *p_ub, kmp_int64 *p_st)
void __kmpc_dispatch_fini_8u(ident_t *loc, kmp_int32 gtid)
void __kmpc_dispatch_init_4(ident_t *loc, kmp_int32 gtid, enum sched_type schedule, kmp_int32 lb, kmp_int32 ub, kmp_int32 st, kmp_int32 chunk)
void __kmpc_dispatch_init_4u(ident_t *loc, kmp_int32 gtid, enum sched_type schedule, kmp_uint32 lb, kmp_uint32 ub, kmp_int32 st, kmp_int32 chunk)
void __kmpc_dispatch_init_8u(ident_t *loc, kmp_int32 gtid, enum sched_type schedule, kmp_uint64 lb, kmp_uint64 ub, kmp_int64 st, kmp_int64 chunk)
void __kmpc_dispatch_fini_4u(ident_t *loc, kmp_int32 gtid)
void __kmpc_dispatch_init_8(ident_t *loc, kmp_int32 gtid, enum sched_type schedule, kmp_int64 lb, kmp_int64 ub, kmp_int64 st, kmp_int64 chunk)
@ kmp_sch_runtime_simd
Definition: kmp.h:358
@ kmp_sch_auto
Definition: kmp.h:343
@ kmp_sch_static
Definition: kmp.h:339
@ kmp_sch_guided_simd
Definition: kmp.h:357
@ kmp_sch_guided_chunked
Definition: kmp.h:341
@ kmp_sch_lower
Definition: kmp.h:337
@ kmp_nm_upper
Definition: kmp.h:408
@ kmp_ord_lower
Definition: kmp.h:363
@ kmp_sch_upper
Definition: kmp.h:361
@ kmp_nm_lower
Definition: kmp.h:381
Definition: kmp.h:222