LLVM OpenMP* Runtime Library
Loading...
Searching...
No Matches
kmp.h
1
2/*
3 * kmp.h -- KPTS runtime header file.
4 */
5
6//===----------------------------------------------------------------------===//
7//
8// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
9// See https://llvm.org/LICENSE.txt for license information.
10// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
11//
12//===----------------------------------------------------------------------===//
13
14#ifndef KMP_H
15#define KMP_H
16
17#include "kmp_config.h"
18
19/* #define BUILD_PARALLEL_ORDERED 1 */
20
21/* This fix replaces gettimeofday with clock_gettime for better scalability on
22 the Altix. Requires user code to be linked with -lrt. */
23//#define FIX_SGI_CLOCK
24
25/* Defines for OpenMP 3.0 tasking and auto scheduling */
26
27#ifndef KMP_STATIC_STEAL_ENABLED
28#define KMP_STATIC_STEAL_ENABLED 1
29#endif
30#define KMP_WEIGHTED_ITERATIONS_SUPPORTED \
31 (KMP_AFFINITY_SUPPORTED && KMP_STATIC_STEAL_ENABLED && \
32 (KMP_ARCH_X86 || KMP_ARCH_X86_64))
33
34#define TASK_CURRENT_NOT_QUEUED 0
35#define TASK_CURRENT_QUEUED 1
36
37#ifdef BUILD_TIED_TASK_STACK
38#define TASK_STACK_EMPTY 0 // entries when the stack is empty
39#define TASK_STACK_BLOCK_BITS 5 // Used in TASK_STACK_SIZE and TASK_STACK_MASK
40// Number of entries in each task stack array
41#define TASK_STACK_BLOCK_SIZE (1 << TASK_STACK_BLOCK_BITS)
42// Mask for determining index into stack block
43#define TASK_STACK_INDEX_MASK (TASK_STACK_BLOCK_SIZE - 1)
44#endif // BUILD_TIED_TASK_STACK
45
46#define TASK_NOT_PUSHED 1
47#define TASK_SUCCESSFULLY_PUSHED 0
48#define TASK_TIED 1
49#define TASK_UNTIED 0
50#define TASK_EXPLICIT 1
51#define TASK_IMPLICIT 0
52#define TASK_PROXY 1
53#define TASK_FULL 0
54#define TASK_DETACHABLE 1
55#define TASK_UNDETACHABLE 0
56
57#define KMP_CANCEL_THREADS
58#define KMP_THREAD_ATTR
59
60// Android does not have pthread_cancel. Undefine KMP_CANCEL_THREADS if being
61// built on Android
62#if defined(__ANDROID__)
63#undef KMP_CANCEL_THREADS
64#endif
65
66// Some WASI targets (e.g., wasm32-wasi-threads) do not support thread
67// cancellation.
68#if KMP_OS_WASI
69#undef KMP_CANCEL_THREADS
70#endif
71
72#if !KMP_OS_WASI
73#include <signal.h>
74#endif
75#include <stdarg.h>
76#include <stddef.h>
77#include <stdio.h>
78#include <stdlib.h>
79#include <string.h>
80#include <limits>
81#include <type_traits>
82/* include <ctype.h> don't use; problems with /MD on Windows* OS NT due to bad
83 Microsoft library. Some macros provided below to replace these functions */
84#ifndef __ABSOFT_WIN
85#include <sys/types.h>
86#endif
87#include <limits.h>
88#include <time.h>
89
90#include <errno.h>
91
92#include "kmp_os.h"
93
94#include "kmp_safe_c_api.h"
95
96#if KMP_STATS_ENABLED
97class kmp_stats_list;
98#endif
99
100#if KMP_USE_HIER_SCHED
101// Only include hierarchical scheduling if affinity is supported
102#undef KMP_USE_HIER_SCHED
103#define KMP_USE_HIER_SCHED KMP_AFFINITY_SUPPORTED
104#endif
105
106// OMPD_SKIP_HWLOC used in libompd/omp-icv.cpp to avoid OMPD depending on hwloc
107#if KMP_USE_HWLOC && KMP_AFFINITY_SUPPORTED && !defined(OMPD_SKIP_HWLOC)
108#include "hwloc.h"
109#ifndef HWLOC_OBJ_NUMANODE
110#define HWLOC_OBJ_NUMANODE HWLOC_OBJ_NODE
111#endif
112#ifndef HWLOC_OBJ_PACKAGE
113#define HWLOC_OBJ_PACKAGE HWLOC_OBJ_SOCKET
114#endif
115#endif
116
117#if KMP_ARCH_X86 || KMP_ARCH_X86_64
118#include <xmmintrin.h>
119#endif
120
121// The below has to be defined before including "kmp_barrier.h".
122#define KMP_INTERNAL_MALLOC(sz) malloc(sz)
123#define KMP_INTERNAL_FREE(p) free(p)
124#define KMP_INTERNAL_REALLOC(p, sz) realloc((p), (sz))
125#define KMP_INTERNAL_CALLOC(n, sz) calloc((n), (sz))
126
127#include "kmp_debug.h"
128#include "kmp_lock.h"
129#include "kmp_version.h"
130#include "kmp_barrier.h"
131#if USE_DEBUGGER
132#include "kmp_debugger.h"
133#endif
134#include "kmp_i18n.h"
135
136#define KMP_HANDLE_SIGNALS ((KMP_OS_UNIX && !KMP_OS_WASI) || KMP_OS_WINDOWS)
137
138#include "kmp_wrapper_malloc.h"
139#if KMP_OS_UNIX
140#include <unistd.h>
141#if !defined NSIG && defined _NSIG
142#define NSIG _NSIG
143#endif
144#endif
145
146#if KMP_OS_LINUX
147#pragma weak clock_gettime
148#endif
149
150#if OMPT_SUPPORT
151#include "ompt-internal.h"
152#endif
153
154#if OMPD_SUPPORT
155#include "ompd-specific.h"
156#endif
157
158#ifndef UNLIKELY
159#define UNLIKELY(x) (x)
160#endif
161
162// Affinity format function
163#include "kmp_str.h"
164
165// 0 - no fast memory allocation, alignment: 8-byte on x86, 16-byte on x64.
166// 3 - fast allocation using sync, non-sync free lists of any size, non-self
167// free lists of limited size.
168#ifndef USE_FAST_MEMORY
169#define USE_FAST_MEMORY 3
170#endif
171
172#ifndef KMP_NESTED_HOT_TEAMS
173#define KMP_NESTED_HOT_TEAMS 0
174#define USE_NESTED_HOT_ARG(x)
175#else
176#if KMP_NESTED_HOT_TEAMS
177#define USE_NESTED_HOT_ARG(x) , x
178#else
179#define USE_NESTED_HOT_ARG(x)
180#endif
181#endif
182
183// Assume using BGET compare_exchange instruction instead of lock by default.
184#ifndef USE_CMP_XCHG_FOR_BGET
185#define USE_CMP_XCHG_FOR_BGET 1
186#endif
187
188// Test to see if queuing lock is better than bootstrap lock for bget
189// #ifndef USE_QUEUING_LOCK_FOR_BGET
190// #define USE_QUEUING_LOCK_FOR_BGET
191// #endif
192
193#define KMP_NSEC_PER_SEC 1000000000L
194#define KMP_USEC_PER_SEC 1000000L
195#define KMP_NSEC_PER_USEC 1000L
196
205enum {
210 /* 0x04 is no longer used */
219 KMP_IDENT_BARRIER_IMPL_MASK = 0x01C0,
220 KMP_IDENT_BARRIER_IMPL_FOR = 0x0040,
221 KMP_IDENT_BARRIER_IMPL_SECTIONS = 0x00C0,
222
223 KMP_IDENT_BARRIER_IMPL_SINGLE = 0x0140,
224 KMP_IDENT_BARRIER_IMPL_WORKSHARE = 0x01C0,
225
237 KMP_IDENT_ATOMIC_HINT_UNCONTENDED = 0x010000,
238 KMP_IDENT_ATOMIC_HINT_CONTENDED = 0x020000,
239 KMP_IDENT_ATOMIC_HINT_NONSPECULATIVE = 0x040000,
240 KMP_IDENT_ATOMIC_HINT_SPECULATIVE = 0x080000,
241 KMP_IDENT_OPENMP_SPEC_VERSION_MASK = 0xFF000000
242};
243
247typedef struct ident {
248 kmp_int32 reserved_1;
249 kmp_int32 flags;
251 kmp_int32 reserved_2;
252#if USE_ITT_BUILD
253/* but currently used for storing region-specific ITT */
254/* contextual information. */
255#endif /* USE_ITT_BUILD */
256 kmp_int32 reserved_3;
257 char const *psource;
261 // Returns the OpenMP version in form major*10+minor (e.g., 50 for 5.0)
262 kmp_int32 get_openmp_version() {
263 return (((flags & KMP_IDENT_OPENMP_SPEC_VERSION_MASK) >> 24) & 0xFF);
264 }
270// Some forward declarations.
271typedef union kmp_team kmp_team_t;
272typedef struct kmp_taskdata kmp_taskdata_t;
273typedef union kmp_task_team kmp_task_team_t;
274typedef union kmp_team kmp_team_p;
275typedef union kmp_info kmp_info_p;
276typedef union kmp_root kmp_root_p;
277
278template <bool C = false, bool S = true> class kmp_flag_32;
279template <bool C = false, bool S = true> class kmp_flag_64;
280template <bool C = false, bool S = true> class kmp_atomic_flag_64;
281class kmp_flag_oncore;
282
283#ifdef __cplusplus
284extern "C" {
285#endif
286
287/* ------------------------------------------------------------------------ */
288
289/* Pack two 32-bit signed integers into a 64-bit signed integer */
290/* ToDo: Fix word ordering for big-endian machines. */
291#define KMP_PACK_64(HIGH_32, LOW_32) \
292 ((kmp_int64)((((kmp_uint64)(HIGH_32)) << 32) | (kmp_uint64)(LOW_32)))
293
294// Generic string manipulation macros. Assume that _x is of type char *
295#define SKIP_WS(_x) \
296 { \
297 while (*(_x) == ' ' || *(_x) == '\t') \
298 (_x)++; \
299 }
300#define SKIP_DIGITS(_x) \
301 { \
302 while (*(_x) >= '0' && *(_x) <= '9') \
303 (_x)++; \
304 }
305#define SKIP_TOKEN(_x) \
306 { \
307 while ((*(_x) >= '0' && *(_x) <= '9') || (*(_x) >= 'a' && *(_x) <= 'z') || \
308 (*(_x) >= 'A' && *(_x) <= 'Z') || *(_x) == '_') \
309 (_x)++; \
310 }
311#define SKIP_TO(_x, _c) \
312 { \
313 while (*(_x) != '\0' && *(_x) != (_c)) \
314 (_x)++; \
315 }
316
317/* ------------------------------------------------------------------------ */
318
319#define KMP_MAX(x, y) ((x) > (y) ? (x) : (y))
320#define KMP_MIN(x, y) ((x) < (y) ? (x) : (y))
321
322/* ------------------------------------------------------------------------ */
323/* Enumeration types */
324
325enum kmp_state_timer {
326 ts_stop,
327 ts_start,
328 ts_pause,
329
330 ts_last_state
331};
332
333enum dynamic_mode {
334 dynamic_default,
335#ifdef USE_LOAD_BALANCE
336 dynamic_load_balance,
337#endif /* USE_LOAD_BALANCE */
338 dynamic_random,
339 dynamic_thread_limit,
340 dynamic_max
341};
342
343/* external schedule constants, duplicate enum omp_sched in omp.h in order to
344 * not include it here */
345#ifndef KMP_SCHED_TYPE_DEFINED
346#define KMP_SCHED_TYPE_DEFINED
347typedef enum kmp_sched {
348 kmp_sched_lower = 0, // lower and upper bounds are for routine parameter check
349 // Note: need to adjust __kmp_sch_map global array in case enum is changed
350 kmp_sched_static = 1, // mapped to kmp_sch_static_chunked (33)
351 kmp_sched_dynamic = 2, // mapped to kmp_sch_dynamic_chunked (35)
352 kmp_sched_guided = 3, // mapped to kmp_sch_guided_chunked (36)
353 kmp_sched_auto = 4, // mapped to kmp_sch_auto (38)
354 kmp_sched_upper_std = 5, // upper bound for standard schedules
355 kmp_sched_lower_ext = 100, // lower bound of Intel extension schedules
356 kmp_sched_trapezoidal = 101, // mapped to kmp_sch_trapezoidal (39)
357#if KMP_STATIC_STEAL_ENABLED
358 kmp_sched_static_steal = 102, // mapped to kmp_sch_static_steal (44)
359#endif
360 kmp_sched_upper,
361 kmp_sched_default = kmp_sched_static, // default scheduling
362 kmp_sched_monotonic = 0x80000000
363} kmp_sched_t;
364#endif
365
370enum sched_type : kmp_int32 {
372 kmp_sch_static_chunked = 33,
374 kmp_sch_dynamic_chunked = 35,
376 kmp_sch_runtime = 37,
378 kmp_sch_trapezoidal = 39,
379
380 /* accessible only through KMP_SCHEDULE environment variable */
381 kmp_sch_static_greedy = 40,
382 kmp_sch_static_balanced = 41,
383 /* accessible only through KMP_SCHEDULE environment variable */
384 kmp_sch_guided_iterative_chunked = 42,
385 kmp_sch_guided_analytical_chunked = 43,
386 /* accessible only through KMP_SCHEDULE environment variable */
387 kmp_sch_static_steal = 44,
388
389 /* static with chunk adjustment (e.g., simd) */
390 kmp_sch_static_balanced_chunked = 45,
394 /* accessible only through KMP_SCHEDULE environment variable */
398 kmp_ord_static_chunked = 65,
400 kmp_ord_dynamic_chunked = 67,
401 kmp_ord_guided_chunked = 68,
402 kmp_ord_runtime = 69,
404 kmp_ord_trapezoidal = 71,
407 /* Schedules for Distribute construct */
411 /* For the "nomerge" versions, kmp_dispatch_next*() will always return a
412 single iteration/chunk, even if the loop is serialized. For the schedule
413 types listed above, the entire iteration vector is returned if the loop is
414 serialized. This doesn't work for gcc/gcomp sections. */
417 kmp_nm_static_chunked =
418 (kmp_sch_static_chunked - kmp_sch_lower + kmp_nm_lower),
420 kmp_nm_dynamic_chunked = 163,
422 kmp_nm_runtime = 165,
424 kmp_nm_trapezoidal = 167,
425
426 /* accessible only through KMP_SCHEDULE environment variable */
427 kmp_nm_static_greedy = 168,
428 kmp_nm_static_balanced = 169,
429 /* accessible only through KMP_SCHEDULE environment variable */
430 kmp_nm_guided_iterative_chunked = 170,
431 kmp_nm_guided_analytical_chunked = 171,
432 kmp_nm_static_steal =
433 172, /* accessible only through OMP_SCHEDULE environment variable */
434
435 kmp_nm_ord_static_chunked = 193,
437 kmp_nm_ord_dynamic_chunked = 195,
438 kmp_nm_ord_guided_chunked = 196,
439 kmp_nm_ord_runtime = 197,
441 kmp_nm_ord_trapezoidal = 199,
444 /* Support for OpenMP 4.5 monotonic and nonmonotonic schedule modifiers. Since
445 we need to distinguish the three possible cases (no modifier, monotonic
446 modifier, nonmonotonic modifier), we need separate bits for each modifier.
447 The absence of monotonic does not imply nonmonotonic, especially since 4.5
448 says that the behaviour of the "no modifier" case is implementation defined
449 in 4.5, but will become "nonmonotonic" in 5.0.
450
451 Since we're passing a full 32 bit value, we can use a couple of high bits
452 for these flags; out of paranoia we avoid the sign bit.
453
454 These modifiers can be or-ed into non-static schedules by the compiler to
455 pass the additional information. They will be stripped early in the
456 processing in __kmp_dispatch_init when setting up schedules, so most of the
457 code won't ever see schedules with these bits set. */
459 (1 << 29),
461 (1 << 30),
463#define SCHEDULE_WITHOUT_MODIFIERS(s) \
464 (enum sched_type)( \
466#define SCHEDULE_HAS_MONOTONIC(s) (((s)&kmp_sch_modifier_monotonic) != 0)
467#define SCHEDULE_HAS_NONMONOTONIC(s) (((s)&kmp_sch_modifier_nonmonotonic) != 0)
468#define SCHEDULE_HAS_NO_MODIFIERS(s) \
469 (((s) & (kmp_sch_modifier_nonmonotonic | kmp_sch_modifier_monotonic)) == 0)
470#define SCHEDULE_GET_MODIFIERS(s) \
471 ((enum sched_type)( \
472 (s) & (kmp_sch_modifier_nonmonotonic | kmp_sch_modifier_monotonic)))
473#define SCHEDULE_SET_MODIFIERS(s, m) \
474 (s = (enum sched_type)((kmp_int32)s | (kmp_int32)m))
475#define SCHEDULE_NONMONOTONIC 0
476#define SCHEDULE_MONOTONIC 1
477
480
481// Apply modifiers on internal kind to standard kind
482static inline void
483__kmp_sched_apply_mods_stdkind(kmp_sched_t *kind,
484 enum sched_type internal_kind) {
485 if (SCHEDULE_HAS_MONOTONIC(internal_kind)) {
486 *kind = (kmp_sched_t)((int)*kind | (int)kmp_sched_monotonic);
487 }
488}
489
490// Apply modifiers on standard kind to internal kind
491static inline void
492__kmp_sched_apply_mods_intkind(kmp_sched_t kind,
493 enum sched_type *internal_kind) {
494 if ((int)kind & (int)kmp_sched_monotonic) {
495 *internal_kind = (enum sched_type)((int)*internal_kind |
497 }
498}
499
500// Get standard schedule without modifiers
501static inline kmp_sched_t __kmp_sched_without_mods(kmp_sched_t kind) {
502 return (kmp_sched_t)((int)kind & ~((int)kmp_sched_monotonic));
503}
504
505/* Type to keep runtime schedule set via OMP_SCHEDULE or omp_set_schedule() */
506typedef union kmp_r_sched {
507 struct {
508 enum sched_type r_sched_type;
509 int chunk;
510 };
511 kmp_int64 sched;
512} kmp_r_sched_t;
513
514extern enum sched_type __kmp_sch_map[]; // map OMP 3.0 schedule types with our
515// internal schedule types
516
517enum library_type {
518 library_none,
519 library_serial,
520 library_turnaround,
521 library_throughput
522};
523
524#if KMP_MIC_SUPPORTED
525enum mic_type { non_mic, mic1, mic2, mic3, dummy };
526#endif
527
528// OpenMP 3.1 - Nested num threads array
529typedef struct kmp_nested_nthreads_t {
530 int *nth;
531 int size;
532 int used;
533} kmp_nested_nthreads_t;
534
535extern kmp_nested_nthreads_t __kmp_nested_nth;
536
537/* -- fast reduction stuff ------------------------------------------------ */
538
539#undef KMP_FAST_REDUCTION_BARRIER
540#define KMP_FAST_REDUCTION_BARRIER 1
541
542#undef KMP_FAST_REDUCTION_CORE_DUO
543#if KMP_ARCH_X86 || KMP_ARCH_X86_64
544#define KMP_FAST_REDUCTION_CORE_DUO 1
545#endif
546
547enum _reduction_method {
548 reduction_method_not_defined = 0,
549 critical_reduce_block = (1 << 8),
550 atomic_reduce_block = (2 << 8),
551 tree_reduce_block = (3 << 8),
552 empty_reduce_block = (4 << 8)
553};
554
555// Description of the packed_reduction_method variable:
556// The packed_reduction_method variable consists of two enum types variables
557// that are packed together into 0-th byte and 1-st byte:
558// 0: (packed_reduction_method & 0x000000FF) is a 'enum barrier_type' value of
559// barrier that will be used in fast reduction: bs_plain_barrier or
560// bs_reduction_barrier
561// 1: (packed_reduction_method & 0x0000FF00) is a reduction method that will
562// be used in fast reduction;
563// Reduction method is of 'enum _reduction_method' type and it's defined the way
564// so that the bits of 0-th byte are empty, so no need to execute a shift
565// instruction while packing/unpacking
566
567#if KMP_FAST_REDUCTION_BARRIER
568#define PACK_REDUCTION_METHOD_AND_BARRIER(reduction_method, barrier_type) \
569 ((reduction_method) | (barrier_type))
570
571#define UNPACK_REDUCTION_METHOD(packed_reduction_method) \
572 ((enum _reduction_method)((packed_reduction_method) & (0x0000FF00)))
573
574#define UNPACK_REDUCTION_BARRIER(packed_reduction_method) \
575 ((enum barrier_type)((packed_reduction_method) & (0x000000FF)))
576#else
577#define PACK_REDUCTION_METHOD_AND_BARRIER(reduction_method, barrier_type) \
578 (reduction_method)
579
580#define UNPACK_REDUCTION_METHOD(packed_reduction_method) \
581 (packed_reduction_method)
582
583#define UNPACK_REDUCTION_BARRIER(packed_reduction_method) (bs_plain_barrier)
584#endif
585
586#define TEST_REDUCTION_METHOD(packed_reduction_method, which_reduction_block) \
587 ((UNPACK_REDUCTION_METHOD(packed_reduction_method)) == \
588 (which_reduction_block))
589
590#if KMP_FAST_REDUCTION_BARRIER
591#define TREE_REDUCE_BLOCK_WITH_REDUCTION_BARRIER \
592 (PACK_REDUCTION_METHOD_AND_BARRIER(tree_reduce_block, bs_reduction_barrier))
593
594#define TREE_REDUCE_BLOCK_WITH_PLAIN_BARRIER \
595 (PACK_REDUCTION_METHOD_AND_BARRIER(tree_reduce_block, bs_plain_barrier))
596#endif
597
598typedef int PACKED_REDUCTION_METHOD_T;
599
600/* -- end of fast reduction stuff ----------------------------------------- */
601
602#if KMP_OS_WINDOWS
603#define USE_CBLKDATA
604#if KMP_MSVC_COMPAT
605#pragma warning(push)
606#pragma warning(disable : 271 310)
607#endif
608#include <windows.h>
609#if KMP_MSVC_COMPAT
610#pragma warning(pop)
611#endif
612#endif
613
614#if KMP_OS_UNIX
615#if !KMP_OS_WASI
616#include <dlfcn.h>
617#endif
618#include <pthread.h>
619#endif
620
621enum kmp_hw_t : int {
622 KMP_HW_UNKNOWN = -1,
623 KMP_HW_SOCKET = 0,
624 KMP_HW_PROC_GROUP,
625 KMP_HW_NUMA,
626 KMP_HW_DIE,
627 KMP_HW_LLC,
628 KMP_HW_L3,
629 KMP_HW_TILE,
630 KMP_HW_MODULE,
631 KMP_HW_L2,
632 KMP_HW_L1,
633 KMP_HW_CORE,
634 KMP_HW_THREAD,
635 KMP_HW_LAST
636};
637
638typedef enum kmp_hw_core_type_t {
639 KMP_HW_CORE_TYPE_UNKNOWN = 0x0,
640#if KMP_ARCH_X86 || KMP_ARCH_X86_64
641 KMP_HW_CORE_TYPE_ATOM = 0x20,
642 KMP_HW_CORE_TYPE_CORE = 0x40,
643 KMP_HW_MAX_NUM_CORE_TYPES = 3,
644#else
645 KMP_HW_MAX_NUM_CORE_TYPES = 1,
646#endif
647} kmp_hw_core_type_t;
648
649#define KMP_HW_MAX_NUM_CORE_EFFS 8
650
651#define KMP_DEBUG_ASSERT_VALID_HW_TYPE(type) \
652 KMP_DEBUG_ASSERT(type >= (kmp_hw_t)0 && type < KMP_HW_LAST)
653#define KMP_ASSERT_VALID_HW_TYPE(type) \
654 KMP_ASSERT(type >= (kmp_hw_t)0 && type < KMP_HW_LAST)
655
656#define KMP_FOREACH_HW_TYPE(type) \
657 for (kmp_hw_t type = (kmp_hw_t)0; type < KMP_HW_LAST; \
658 type = (kmp_hw_t)((int)type + 1))
659
660const char *__kmp_hw_get_keyword(kmp_hw_t type, bool plural = false);
661const char *__kmp_hw_get_catalog_string(kmp_hw_t type, bool plural = false);
662const char *__kmp_hw_get_core_type_string(kmp_hw_core_type_t type);
663
664/* Only Linux* OS and Windows* OS support thread affinity. */
665#if KMP_AFFINITY_SUPPORTED
666
667// GROUP_AFFINITY is already defined for _MSC_VER>=1600 (VS2010 and later).
668#if KMP_OS_WINDOWS
669#if _MSC_VER < 1600 && KMP_MSVC_COMPAT
670typedef struct GROUP_AFFINITY {
671 KAFFINITY Mask;
672 WORD Group;
673 WORD Reserved[3];
674} GROUP_AFFINITY;
675#endif /* _MSC_VER < 1600 */
676#if KMP_GROUP_AFFINITY
677extern int __kmp_num_proc_groups;
678#else
679static const int __kmp_num_proc_groups = 1;
680#endif /* KMP_GROUP_AFFINITY */
681typedef DWORD (*kmp_GetActiveProcessorCount_t)(WORD);
682extern kmp_GetActiveProcessorCount_t __kmp_GetActiveProcessorCount;
683
684typedef WORD (*kmp_GetActiveProcessorGroupCount_t)(void);
685extern kmp_GetActiveProcessorGroupCount_t __kmp_GetActiveProcessorGroupCount;
686
687typedef BOOL (*kmp_GetThreadGroupAffinity_t)(HANDLE, GROUP_AFFINITY *);
688extern kmp_GetThreadGroupAffinity_t __kmp_GetThreadGroupAffinity;
689
690typedef BOOL (*kmp_SetThreadGroupAffinity_t)(HANDLE, const GROUP_AFFINITY *,
691 GROUP_AFFINITY *);
692extern kmp_SetThreadGroupAffinity_t __kmp_SetThreadGroupAffinity;
693#endif /* KMP_OS_WINDOWS */
694
695#if KMP_USE_HWLOC && !defined(OMPD_SKIP_HWLOC)
696extern hwloc_topology_t __kmp_hwloc_topology;
697extern int __kmp_hwloc_error;
698#endif
699
700extern size_t __kmp_affin_mask_size;
701#define KMP_AFFINITY_CAPABLE() (__kmp_affin_mask_size > 0)
702#define KMP_AFFINITY_DISABLE() (__kmp_affin_mask_size = 0)
703#define KMP_AFFINITY_ENABLE(mask_size) (__kmp_affin_mask_size = mask_size)
704#define KMP_CPU_SET_ITERATE(i, mask) \
705 for (i = (mask)->begin(); (int)i != (mask)->end(); i = (mask)->next(i))
706#define KMP_CPU_SET(i, mask) (mask)->set(i)
707#define KMP_CPU_ISSET(i, mask) (mask)->is_set(i)
708#define KMP_CPU_CLR(i, mask) (mask)->clear(i)
709#define KMP_CPU_ZERO(mask) (mask)->zero()
710#define KMP_CPU_ISEMPTY(mask) (mask)->empty()
711#define KMP_CPU_COPY(dest, src) (dest)->copy(src)
712#define KMP_CPU_AND(dest, src) (dest)->bitwise_and(src)
713#define KMP_CPU_COMPLEMENT(max_bit_number, mask) (mask)->bitwise_not()
714#define KMP_CPU_UNION(dest, src) (dest)->bitwise_or(src)
715#define KMP_CPU_EQUAL(dest, src) (dest)->is_equal(src)
716#define KMP_CPU_ALLOC(ptr) (ptr = __kmp_affinity_dispatch->allocate_mask())
717#define KMP_CPU_FREE(ptr) __kmp_affinity_dispatch->deallocate_mask(ptr)
718#define KMP_CPU_ALLOC_ON_STACK(ptr) KMP_CPU_ALLOC(ptr)
719#define KMP_CPU_FREE_FROM_STACK(ptr) KMP_CPU_FREE(ptr)
720#define KMP_CPU_INTERNAL_ALLOC(ptr) KMP_CPU_ALLOC(ptr)
721#define KMP_CPU_INTERNAL_FREE(ptr) KMP_CPU_FREE(ptr)
722#define KMP_CPU_INDEX(arr, i) __kmp_affinity_dispatch->index_mask_array(arr, i)
723#define KMP_CPU_ALLOC_ARRAY(arr, n) \
724 (arr = __kmp_affinity_dispatch->allocate_mask_array(n))
725#define KMP_CPU_FREE_ARRAY(arr, n) \
726 __kmp_affinity_dispatch->deallocate_mask_array(arr)
727#define KMP_CPU_INTERNAL_ALLOC_ARRAY(arr, n) KMP_CPU_ALLOC_ARRAY(arr, n)
728#define KMP_CPU_INTERNAL_FREE_ARRAY(arr, n) KMP_CPU_FREE_ARRAY(arr, n)
729#define __kmp_get_system_affinity(mask, abort_bool) \
730 (mask)->get_system_affinity(abort_bool)
731#define __kmp_set_system_affinity(mask, abort_bool) \
732 (mask)->set_system_affinity(abort_bool)
733#define __kmp_get_proc_group(mask) (mask)->get_proc_group()
734
735class KMPAffinity {
736public:
737 class Mask {
738 public:
739 void *operator new(size_t n);
740 void operator delete(void *p);
741 void *operator new[](size_t n);
742 void operator delete[](void *p);
743 virtual ~Mask() {}
744 // Set bit i to 1
745 virtual void set(int i) {}
746 // Return bit i
747 virtual bool is_set(int i) const { return false; }
748 // Set bit i to 0
749 virtual void clear(int i) {}
750 // Zero out entire mask
751 virtual void zero() {}
752 // Check whether mask is empty
753 virtual bool empty() const { return true; }
754 // Copy src into this mask
755 virtual void copy(const Mask *src) {}
756 // this &= rhs
757 virtual void bitwise_and(const Mask *rhs) {}
758 // this |= rhs
759 virtual void bitwise_or(const Mask *rhs) {}
760 // this = ~this
761 virtual void bitwise_not() {}
762 // this == rhs
763 virtual bool is_equal(const Mask *rhs) const { return false; }
764 // API for iterating over an affinity mask
765 // for (int i = mask->begin(); i != mask->end(); i = mask->next(i))
766 virtual int begin() const { return 0; }
767 virtual int end() const { return 0; }
768 virtual int next(int previous) const { return 0; }
769#if KMP_OS_WINDOWS
770 virtual int set_process_affinity(bool abort_on_error) const { return -1; }
771#endif
772 // Set the system's affinity to this affinity mask's value
773 virtual int set_system_affinity(bool abort_on_error) const { return -1; }
774 // Set this affinity mask to the current system affinity
775 virtual int get_system_affinity(bool abort_on_error) { return -1; }
776 // Only 1 DWORD in the mask should have any procs set.
777 // Return the appropriate index, or -1 for an invalid mask.
778 virtual int get_proc_group() const { return -1; }
779 int get_max_cpu() const {
780 int cpu;
781 int max_cpu = -1;
782 KMP_CPU_SET_ITERATE(cpu, this) {
783 if (cpu > max_cpu)
784 max_cpu = cpu;
785 }
786 return max_cpu;
787 }
788 };
789 void *operator new(size_t n);
790 void operator delete(void *p);
791 // Need virtual destructor
792 virtual ~KMPAffinity() = default;
793 // Determine if affinity is capable
794 virtual void determine_capable(const char *env_var) {}
795 // Bind the current thread to os proc
796 virtual void bind_thread(int proc) {}
797 // Factory functions to allocate/deallocate a mask
798 virtual Mask *allocate_mask() { return nullptr; }
799 virtual void deallocate_mask(Mask *m) {}
800 virtual Mask *allocate_mask_array(int num) { return nullptr; }
801 virtual void deallocate_mask_array(Mask *m) {}
802 virtual Mask *index_mask_array(Mask *m, int index) { return nullptr; }
803 static void pick_api();
804 static void destroy_api();
805 enum api_type {
806 NATIVE_OS
807#if KMP_USE_HWLOC
808 ,
809 HWLOC
810#endif
811 };
812 virtual api_type get_api_type() const {
813 KMP_ASSERT(0);
814 return NATIVE_OS;
815 }
816
817private:
818 static bool picked_api;
819};
820
821typedef KMPAffinity::Mask kmp_affin_mask_t;
822extern KMPAffinity *__kmp_affinity_dispatch;
823
824#ifndef KMP_OS_AIX
825class kmp_affinity_raii_t {
826 kmp_affin_mask_t *mask;
827 bool restored;
828
829public:
830 kmp_affinity_raii_t(const kmp_affin_mask_t *new_mask = nullptr)
831 : mask(nullptr), restored(false) {
832 if (KMP_AFFINITY_CAPABLE()) {
833 KMP_CPU_ALLOC(mask);
834 KMP_ASSERT(mask != NULL);
835 __kmp_get_system_affinity(mask, /*abort_on_error=*/true);
836 if (new_mask)
837 __kmp_set_system_affinity(new_mask, /*abort_on_error=*/true);
838 }
839 }
840 void restore() {
841 if (mask && KMP_AFFINITY_CAPABLE() && !restored) {
842 __kmp_set_system_affinity(mask, /*abort_on_error=*/true);
843 KMP_CPU_FREE(mask);
844 }
845 restored = true;
846 }
847 ~kmp_affinity_raii_t() { restore(); }
848};
849#endif // !KMP_OS_AIX
850
851// Declare local char buffers with this size for printing debug and info
852// messages, using __kmp_affinity_print_mask().
853#define KMP_AFFIN_MASK_PRINT_LEN 1024
854
855enum affinity_type {
856 affinity_none = 0,
857 affinity_physical,
858 affinity_logical,
859 affinity_compact,
860 affinity_scatter,
861 affinity_explicit,
862 affinity_balanced,
863 affinity_disabled, // not used outsize the env var parser
864 affinity_default
865};
866
867enum affinity_top_method {
868 affinity_top_method_all = 0, // try all (supported) methods, in order
869#if KMP_ARCH_X86 || KMP_ARCH_X86_64
870 affinity_top_method_apicid,
871 affinity_top_method_x2apicid,
872 affinity_top_method_x2apicid_1f,
873#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */
874 affinity_top_method_cpuinfo, // KMP_CPUINFO_FILE is usable on Windows* OS, too
875#if KMP_GROUP_AFFINITY
876 affinity_top_method_group,
877#endif /* KMP_GROUP_AFFINITY */
878 affinity_top_method_flat,
879#if KMP_USE_HWLOC
880 affinity_top_method_hwloc,
881#endif
882 affinity_top_method_default
883};
884
885#define affinity_respect_mask_default (2)
886
887typedef struct kmp_affinity_flags_t {
888 unsigned dups : 1;
889 unsigned verbose : 1;
890 unsigned warnings : 1;
891 unsigned respect : 2;
892 unsigned reset : 1;
893 unsigned initialized : 1;
894 unsigned core_types_gran : 1;
895 unsigned core_effs_gran : 1;
896 unsigned omp_places : 1;
897 unsigned reserved : 22;
898} kmp_affinity_flags_t;
899KMP_BUILD_ASSERT(sizeof(kmp_affinity_flags_t) == 4);
900
901typedef struct kmp_affinity_ids_t {
902 int os_id;
903 int ids[KMP_HW_LAST];
904} kmp_affinity_ids_t;
905
906typedef struct kmp_affinity_attrs_t {
907 int core_type : 8;
908 int core_eff : 8;
909 unsigned valid : 1;
910 unsigned reserved : 15;
911} kmp_affinity_attrs_t;
912#define KMP_AFFINITY_ATTRS_UNKNOWN \
913 { KMP_HW_CORE_TYPE_UNKNOWN, kmp_hw_attr_t::UNKNOWN_CORE_EFF, 0, 0 }
914
915typedef struct kmp_affinity_t {
916 char *proclist;
917 enum affinity_type type;
918 kmp_hw_t gran;
919 int gran_levels;
920 kmp_affinity_attrs_t core_attr_gran;
921 int compact;
922 int offset;
923 kmp_affinity_flags_t flags;
924 unsigned num_masks;
925 kmp_affin_mask_t *masks;
926 kmp_affinity_ids_t *ids;
927 kmp_affinity_attrs_t *attrs;
928 unsigned num_os_id_masks;
929 kmp_affin_mask_t *os_id_masks;
930 const char *env_var;
931} kmp_affinity_t;
932
933#define KMP_AFFINITY_INIT(env) \
934 { \
935 nullptr, affinity_default, KMP_HW_UNKNOWN, -1, KMP_AFFINITY_ATTRS_UNKNOWN, \
936 0, 0, \
937 {TRUE, FALSE, TRUE, affinity_respect_mask_default, FALSE, FALSE, \
938 FALSE, FALSE, FALSE}, \
939 0, nullptr, nullptr, nullptr, 0, nullptr, env \
940 }
941
942extern enum affinity_top_method __kmp_affinity_top_method;
943extern kmp_affinity_t __kmp_affinity;
944extern kmp_affinity_t __kmp_hh_affinity;
945extern kmp_affinity_t *__kmp_affinities[2];
946
947extern void __kmp_affinity_bind_thread(int which);
948
949extern kmp_affin_mask_t *__kmp_affin_fullMask;
950extern kmp_affin_mask_t *__kmp_affin_origMask;
951extern char *__kmp_cpuinfo_file;
952
953#if KMP_WEIGHTED_ITERATIONS_SUPPORTED
954extern int __kmp_first_osid_with_ecore;
955#endif
956
957#endif /* KMP_AFFINITY_SUPPORTED */
958
959// This needs to be kept in sync with the values in omp.h !!!
960typedef enum kmp_proc_bind_t {
961 proc_bind_false = 0,
962 proc_bind_true,
963 proc_bind_primary,
964 proc_bind_close,
965 proc_bind_spread,
966 proc_bind_intel, // use KMP_AFFINITY interface
967 proc_bind_default
968} kmp_proc_bind_t;
969
970typedef struct kmp_nested_proc_bind_t {
971 kmp_proc_bind_t *bind_types;
972 int size;
973 int used;
974} kmp_nested_proc_bind_t;
975
976extern kmp_nested_proc_bind_t __kmp_nested_proc_bind;
977extern kmp_proc_bind_t __kmp_teams_proc_bind;
978
979extern int __kmp_display_affinity;
980extern char *__kmp_affinity_format;
981static const size_t KMP_AFFINITY_FORMAT_SIZE = 512;
982#if OMPT_SUPPORT
983extern int __kmp_tool;
984extern char *__kmp_tool_libraries;
985#endif // OMPT_SUPPORT
986
987#if KMP_AFFINITY_SUPPORTED
988#define KMP_PLACE_ALL (-1)
989#define KMP_PLACE_UNDEFINED (-2)
990// Is KMP_AFFINITY is being used instead of OMP_PROC_BIND/OMP_PLACES?
991#define KMP_AFFINITY_NON_PROC_BIND \
992 ((__kmp_nested_proc_bind.bind_types[0] == proc_bind_false || \
993 __kmp_nested_proc_bind.bind_types[0] == proc_bind_intel) && \
994 (__kmp_affinity.num_masks > 0 || __kmp_affinity.type == affinity_balanced))
995#endif /* KMP_AFFINITY_SUPPORTED */
996
997extern int __kmp_affinity_num_places;
998
999typedef enum kmp_cancel_kind_t {
1000 cancel_noreq = 0,
1001 cancel_parallel = 1,
1002 cancel_loop = 2,
1003 cancel_sections = 3,
1004 cancel_taskgroup = 4
1005} kmp_cancel_kind_t;
1006
1007// KMP_HW_SUBSET support:
1008typedef struct kmp_hws_item {
1009 int num;
1010 int offset;
1011} kmp_hws_item_t;
1012
1013extern kmp_hws_item_t __kmp_hws_socket;
1014extern kmp_hws_item_t __kmp_hws_die;
1015extern kmp_hws_item_t __kmp_hws_node;
1016extern kmp_hws_item_t __kmp_hws_tile;
1017extern kmp_hws_item_t __kmp_hws_core;
1018extern kmp_hws_item_t __kmp_hws_proc;
1019extern int __kmp_hws_requested;
1020extern int __kmp_hws_abs_flag; // absolute or per-item number requested
1021
1022/* ------------------------------------------------------------------------ */
1023
1024#define KMP_PAD(type, sz) \
1025 (sizeof(type) + (sz - ((sizeof(type) - 1) % (sz)) - 1))
1026
1027// We need to avoid using -1 as a GTID as +1 is added to the gtid
1028// when storing it in a lock, and the value 0 is reserved.
1029#define KMP_GTID_DNE (-2) /* Does not exist */
1030#define KMP_GTID_SHUTDOWN (-3) /* Library is shutting down */
1031#define KMP_GTID_MONITOR (-4) /* Monitor thread ID */
1032#define KMP_GTID_UNKNOWN (-5) /* Is not known */
1033#define KMP_GTID_MIN (-6) /* Minimal gtid for low bound check in DEBUG */
1034
1035/* OpenMP 5.0 Memory Management support */
1036
1037#ifndef __OMP_H
1038// Duplicate type definitions from omp.h
1039typedef uintptr_t omp_uintptr_t;
1040
1041typedef enum {
1042 omp_atk_sync_hint = 1,
1043 omp_atk_alignment = 2,
1044 omp_atk_access = 3,
1045 omp_atk_pool_size = 4,
1046 omp_atk_fallback = 5,
1047 omp_atk_fb_data = 6,
1048 omp_atk_pinned = 7,
1049 omp_atk_partition = 8
1050} omp_alloctrait_key_t;
1051
1052typedef enum {
1053 omp_atv_false = 0,
1054 omp_atv_true = 1,
1055 omp_atv_contended = 3,
1056 omp_atv_uncontended = 4,
1057 omp_atv_serialized = 5,
1058 omp_atv_sequential = omp_atv_serialized, // (deprecated)
1059 omp_atv_private = 6,
1060 omp_atv_all = 7,
1061 omp_atv_thread = 8,
1062 omp_atv_pteam = 9,
1063 omp_atv_cgroup = 10,
1064 omp_atv_default_mem_fb = 11,
1065 omp_atv_null_fb = 12,
1066 omp_atv_abort_fb = 13,
1067 omp_atv_allocator_fb = 14,
1068 omp_atv_environment = 15,
1069 omp_atv_nearest = 16,
1070 omp_atv_blocked = 17,
1071 omp_atv_interleaved = 18
1072} omp_alloctrait_value_t;
1073#define omp_atv_default ((omp_uintptr_t)-1)
1074
1075typedef void *omp_memspace_handle_t;
1076extern omp_memspace_handle_t const omp_default_mem_space;
1077extern omp_memspace_handle_t const omp_large_cap_mem_space;
1078extern omp_memspace_handle_t const omp_const_mem_space;
1079extern omp_memspace_handle_t const omp_high_bw_mem_space;
1080extern omp_memspace_handle_t const omp_low_lat_mem_space;
1081extern omp_memspace_handle_t const llvm_omp_target_host_mem_space;
1082extern omp_memspace_handle_t const llvm_omp_target_shared_mem_space;
1083extern omp_memspace_handle_t const llvm_omp_target_device_mem_space;
1084
1085typedef struct {
1086 omp_alloctrait_key_t key;
1087 omp_uintptr_t value;
1088} omp_alloctrait_t;
1089
1090typedef void *omp_allocator_handle_t;
1091extern omp_allocator_handle_t const omp_null_allocator;
1092extern omp_allocator_handle_t const omp_default_mem_alloc;
1093extern omp_allocator_handle_t const omp_large_cap_mem_alloc;
1094extern omp_allocator_handle_t const omp_const_mem_alloc;
1095extern omp_allocator_handle_t const omp_high_bw_mem_alloc;
1096extern omp_allocator_handle_t const omp_low_lat_mem_alloc;
1097extern omp_allocator_handle_t const omp_cgroup_mem_alloc;
1098extern omp_allocator_handle_t const omp_pteam_mem_alloc;
1099extern omp_allocator_handle_t const omp_thread_mem_alloc;
1100extern omp_allocator_handle_t const llvm_omp_target_host_mem_alloc;
1101extern omp_allocator_handle_t const llvm_omp_target_shared_mem_alloc;
1102extern omp_allocator_handle_t const llvm_omp_target_device_mem_alloc;
1103extern omp_allocator_handle_t const kmp_max_mem_alloc;
1104extern omp_allocator_handle_t __kmp_def_allocator;
1105
1106// end of duplicate type definitions from omp.h
1107#endif
1108
1109extern int __kmp_memkind_available;
1110
1111typedef omp_memspace_handle_t kmp_memspace_t; // placeholder
1112
1113typedef struct kmp_allocator_t {
1114 omp_memspace_handle_t memspace;
1115 void **memkind; // pointer to memkind
1116 size_t alignment;
1117 omp_alloctrait_value_t fb;
1118 kmp_allocator_t *fb_data;
1119 kmp_uint64 pool_size;
1120 kmp_uint64 pool_used;
1121 bool pinned;
1122} kmp_allocator_t;
1123
1124extern omp_allocator_handle_t __kmpc_init_allocator(int gtid,
1125 omp_memspace_handle_t,
1126 int ntraits,
1127 omp_alloctrait_t traits[]);
1128extern void __kmpc_destroy_allocator(int gtid, omp_allocator_handle_t al);
1129extern void __kmpc_set_default_allocator(int gtid, omp_allocator_handle_t al);
1130extern omp_allocator_handle_t __kmpc_get_default_allocator(int gtid);
1131// external interfaces, may be used by compiler
1132extern void *__kmpc_alloc(int gtid, size_t sz, omp_allocator_handle_t al);
1133extern void *__kmpc_aligned_alloc(int gtid, size_t align, size_t sz,
1134 omp_allocator_handle_t al);
1135extern void *__kmpc_calloc(int gtid, size_t nmemb, size_t sz,
1136 omp_allocator_handle_t al);
1137extern void *__kmpc_realloc(int gtid, void *ptr, size_t sz,
1138 omp_allocator_handle_t al,
1139 omp_allocator_handle_t free_al);
1140extern void __kmpc_free(int gtid, void *ptr, omp_allocator_handle_t al);
1141// internal interfaces, contain real implementation
1142extern void *__kmp_alloc(int gtid, size_t align, size_t sz,
1143 omp_allocator_handle_t al);
1144extern void *__kmp_calloc(int gtid, size_t align, size_t nmemb, size_t sz,
1145 omp_allocator_handle_t al);
1146extern void *__kmp_realloc(int gtid, void *ptr, size_t sz,
1147 omp_allocator_handle_t al,
1148 omp_allocator_handle_t free_al);
1149extern void ___kmpc_free(int gtid, void *ptr, omp_allocator_handle_t al);
1150
1151extern void __kmp_init_memkind();
1152extern void __kmp_fini_memkind();
1153extern void __kmp_init_target_mem();
1154
1155/* ------------------------------------------------------------------------ */
1156
1157#if ENABLE_LIBOMPTARGET
1158extern void __kmp_init_target_task();
1159#endif
1160
1161/* ------------------------------------------------------------------------ */
1162
1163#define KMP_UINT64_MAX \
1164 (~((kmp_uint64)1 << ((sizeof(kmp_uint64) * (1 << 3)) - 1)))
1165
1166#define KMP_MIN_NTH 1
1167
1168#ifndef KMP_MAX_NTH
1169#if defined(PTHREAD_THREADS_MAX) && PTHREAD_THREADS_MAX < INT_MAX
1170#define KMP_MAX_NTH PTHREAD_THREADS_MAX
1171#else
1172#ifdef __ve__
1173// VE's pthread supports only up to 64 threads per a VE process.
1174// Please check p. 14 of following documentation for more details.
1175// https://sxauroratsubasa.sakura.ne.jp/documents/veos/en/VEOS_high_level_design.pdf
1176#define KMP_MAX_NTH 64
1177#else
1178#define KMP_MAX_NTH INT_MAX
1179#endif
1180#endif
1181#endif /* KMP_MAX_NTH */
1182
1183#ifdef PTHREAD_STACK_MIN
1184#define KMP_MIN_STKSIZE ((size_t)PTHREAD_STACK_MIN)
1185#else
1186#define KMP_MIN_STKSIZE ((size_t)(32 * 1024))
1187#endif
1188
1189#if KMP_OS_AIX && KMP_ARCH_PPC
1190#define KMP_MAX_STKSIZE 0x10000000 /* 256Mb max size on 32-bit AIX */
1191#else
1192#define KMP_MAX_STKSIZE (~((size_t)1 << ((sizeof(size_t) * (1 << 3)) - 1)))
1193#endif
1194
1195#if KMP_ARCH_X86
1196#define KMP_DEFAULT_STKSIZE ((size_t)(2 * 1024 * 1024))
1197#elif KMP_ARCH_X86_64
1198#define KMP_DEFAULT_STKSIZE ((size_t)(4 * 1024 * 1024))
1199#define KMP_BACKUP_STKSIZE ((size_t)(2 * 1024 * 1024))
1200#elif KMP_ARCH_VE
1201// Minimum stack size for pthread for VE is 4MB.
1202// https://www.hpc.nec/documents/veos/en/glibc/Difference_Points_glibc.htm
1203#define KMP_DEFAULT_STKSIZE ((size_t)(4 * 1024 * 1024))
1204#elif KMP_OS_AIX
1205// The default stack size for worker threads on AIX is 4MB.
1206#define KMP_DEFAULT_STKSIZE ((size_t)(4 * 1024 * 1024))
1207#else
1208#define KMP_DEFAULT_STKSIZE ((size_t)(1024 * 1024))
1209#endif
1210
1211#define KMP_DEFAULT_MALLOC_POOL_INCR ((size_t)(1024 * 1024))
1212#define KMP_MIN_MALLOC_POOL_INCR ((size_t)(4 * 1024))
1213#define KMP_MAX_MALLOC_POOL_INCR \
1214 (~((size_t)1 << ((sizeof(size_t) * (1 << 3)) - 1)))
1215
1216#define KMP_MIN_STKOFFSET (0)
1217#define KMP_MAX_STKOFFSET KMP_MAX_STKSIZE
1218#if KMP_OS_DARWIN
1219#define KMP_DEFAULT_STKOFFSET KMP_MIN_STKOFFSET
1220#else
1221#define KMP_DEFAULT_STKOFFSET CACHE_LINE
1222#endif
1223
1224#define KMP_MIN_STKPADDING (0)
1225#define KMP_MAX_STKPADDING (2 * 1024 * 1024)
1226
1227#define KMP_BLOCKTIME_MULTIPLIER \
1228 (1000000) /* number of blocktime units per second */
1229#define KMP_MIN_BLOCKTIME (0)
1230#define KMP_MAX_BLOCKTIME \
1231 (INT_MAX) /* Must be this for "infinite" setting the work */
1232
1233/* __kmp_blocktime is in microseconds */
1234#define KMP_DEFAULT_BLOCKTIME (__kmp_is_hybrid_cpu() ? (0) : (200000))
1235
1236#if KMP_USE_MONITOR
1237#define KMP_DEFAULT_MONITOR_STKSIZE ((size_t)(64 * 1024))
1238#define KMP_MIN_MONITOR_WAKEUPS (1) // min times monitor wakes up per second
1239#define KMP_MAX_MONITOR_WAKEUPS (1000) // max times monitor can wake up per sec
1240
1241/* Calculate new number of monitor wakeups for a specific block time based on
1242 previous monitor_wakeups. Only allow increasing number of wakeups */
1243#define KMP_WAKEUPS_FROM_BLOCKTIME(blocktime, monitor_wakeups) \
1244 (((blocktime) == KMP_MAX_BLOCKTIME) ? (monitor_wakeups) \
1245 : ((blocktime) == KMP_MIN_BLOCKTIME) ? KMP_MAX_MONITOR_WAKEUPS \
1246 : ((monitor_wakeups) > (KMP_BLOCKTIME_MULTIPLIER / (blocktime))) \
1247 ? (monitor_wakeups) \
1248 : (KMP_BLOCKTIME_MULTIPLIER) / (blocktime))
1249
1250/* Calculate number of intervals for a specific block time based on
1251 monitor_wakeups */
1252#define KMP_INTERVALS_FROM_BLOCKTIME(blocktime, monitor_wakeups) \
1253 (((blocktime) + (KMP_BLOCKTIME_MULTIPLIER / (monitor_wakeups)) - 1) / \
1254 (KMP_BLOCKTIME_MULTIPLIER / (monitor_wakeups)))
1255#else
1256#define KMP_BLOCKTIME(team, tid) \
1257 (get__bt_set(team, tid) ? get__blocktime(team, tid) : __kmp_dflt_blocktime)
1258#if KMP_OS_UNIX && (KMP_ARCH_X86 || KMP_ARCH_X86_64)
1259// HW TSC is used to reduce overhead (clock tick instead of nanosecond).
1260extern kmp_uint64 __kmp_ticks_per_msec;
1261extern kmp_uint64 __kmp_ticks_per_usec;
1262#if KMP_COMPILER_ICC || KMP_COMPILER_ICX
1263#define KMP_NOW() ((kmp_uint64)_rdtsc())
1264#else
1265#define KMP_NOW() __kmp_hardware_timestamp()
1266#endif
1267#define KMP_BLOCKTIME_INTERVAL(team, tid) \
1268 ((kmp_uint64)KMP_BLOCKTIME(team, tid) * __kmp_ticks_per_usec)
1269#define KMP_BLOCKING(goal, count) ((goal) > KMP_NOW())
1270#else
1271// System time is retrieved sporadically while blocking.
1272extern kmp_uint64 __kmp_now_nsec();
1273#define KMP_NOW() __kmp_now_nsec()
1274#define KMP_BLOCKTIME_INTERVAL(team, tid) \
1275 ((kmp_uint64)KMP_BLOCKTIME(team, tid) * (kmp_uint64)KMP_NSEC_PER_USEC)
1276#define KMP_BLOCKING(goal, count) ((count) % 1000 != 0 || (goal) > KMP_NOW())
1277#endif
1278#endif // KMP_USE_MONITOR
1279
1280#define KMP_MIN_STATSCOLS 40
1281#define KMP_MAX_STATSCOLS 4096
1282#define KMP_DEFAULT_STATSCOLS 80
1283
1284#define KMP_MIN_INTERVAL 0
1285#define KMP_MAX_INTERVAL (INT_MAX - 1)
1286#define KMP_DEFAULT_INTERVAL 0
1287
1288#define KMP_MIN_CHUNK 1
1289#define KMP_MAX_CHUNK (INT_MAX - 1)
1290#define KMP_DEFAULT_CHUNK 1
1291
1292#define KMP_MIN_DISP_NUM_BUFF 1
1293#define KMP_DFLT_DISP_NUM_BUFF 7
1294#define KMP_MAX_DISP_NUM_BUFF 4096
1295
1296#define KMP_MAX_ORDERED 8
1297
1298#define KMP_MAX_FIELDS 32
1299
1300#define KMP_MAX_BRANCH_BITS 31
1301
1302#define KMP_MAX_ACTIVE_LEVELS_LIMIT INT_MAX
1303
1304#define KMP_MAX_DEFAULT_DEVICE_LIMIT INT_MAX
1305
1306#define KMP_MAX_TASK_PRIORITY_LIMIT INT_MAX
1307
1308/* Minimum number of threads before switch to TLS gtid (experimentally
1309 determined) */
1310/* josh TODO: what about OS X* tuning? */
1311#if KMP_ARCH_X86 || KMP_ARCH_X86_64
1312#define KMP_TLS_GTID_MIN 5
1313#else
1314#define KMP_TLS_GTID_MIN INT_MAX
1315#endif
1316
1317#define KMP_MASTER_TID(tid) (0 == (tid))
1318#define KMP_WORKER_TID(tid) (0 != (tid))
1319
1320#define KMP_MASTER_GTID(gtid) (0 == __kmp_tid_from_gtid((gtid)))
1321#define KMP_WORKER_GTID(gtid) (0 != __kmp_tid_from_gtid((gtid)))
1322#define KMP_INITIAL_GTID(gtid) (0 == (gtid))
1323
1324#ifndef TRUE
1325#define FALSE 0
1326#define TRUE (!FALSE)
1327#endif
1328
1329/* NOTE: all of the following constants must be even */
1330
1331#if KMP_OS_WINDOWS
1332#define KMP_INIT_WAIT 64U /* initial number of spin-tests */
1333#define KMP_NEXT_WAIT 32U /* susequent number of spin-tests */
1334#elif KMP_OS_LINUX
1335#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1336#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1337#elif KMP_OS_DARWIN
1338/* TODO: tune for KMP_OS_DARWIN */
1339#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1340#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1341#elif KMP_OS_DRAGONFLY
1342/* TODO: tune for KMP_OS_DRAGONFLY */
1343#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1344#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1345#elif KMP_OS_FREEBSD
1346/* TODO: tune for KMP_OS_FREEBSD */
1347#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1348#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1349#elif KMP_OS_NETBSD
1350/* TODO: tune for KMP_OS_NETBSD */
1351#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1352#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1353#elif KMP_OS_OPENBSD
1354/* TODO: tune for KMP_OS_OPENBSD */
1355#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1356#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1357#elif KMP_OS_HURD
1358/* TODO: tune for KMP_OS_HURD */
1359#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1360#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1361#elif KMP_OS_SOLARIS
1362/* TODO: tune for KMP_OS_SOLARIS */
1363#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1364#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1365#elif KMP_OS_WASI
1366/* TODO: tune for KMP_OS_WASI */
1367#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1368#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1369#elif KMP_OS_AIX
1370/* TODO: tune for KMP_OS_AIX */
1371#define KMP_INIT_WAIT 1024U /* initial number of spin-tests */
1372#define KMP_NEXT_WAIT 512U /* susequent number of spin-tests */
1373#endif
1374
1375#if KMP_ARCH_X86 || KMP_ARCH_X86_64
1376typedef struct kmp_cpuid {
1377 kmp_uint32 eax;
1378 kmp_uint32 ebx;
1379 kmp_uint32 ecx;
1380 kmp_uint32 edx;
1381} kmp_cpuid_t;
1382
1383typedef struct kmp_cpuinfo_flags_t {
1384 unsigned sse2 : 1; // 0 if SSE2 instructions are not supported, 1 otherwise.
1385 unsigned rtm : 1; // 0 if RTM instructions are not supported, 1 otherwise.
1386 unsigned hybrid : 1;
1387 unsigned reserved : 29; // Ensure size of 32 bits
1388} kmp_cpuinfo_flags_t;
1389
1390typedef struct kmp_cpuinfo {
1391 int initialized; // If 0, other fields are not initialized.
1392 int signature; // CPUID(1).EAX
1393 int family; // CPUID(1).EAX[27:20]+CPUID(1).EAX[11:8] (Extended Family+Family)
1394 int model; // ( CPUID(1).EAX[19:16] << 4 ) + CPUID(1).EAX[7:4] ( ( Extended
1395 // Model << 4 ) + Model)
1396 int stepping; // CPUID(1).EAX[3:0] ( Stepping )
1397 kmp_cpuinfo_flags_t flags;
1398 int apic_id;
1399 kmp_uint64 frequency; // Nominal CPU frequency in Hz.
1400 char name[3 * sizeof(kmp_cpuid_t)]; // CPUID(0x80000002,0x80000003,0x80000004)
1401} kmp_cpuinfo_t;
1402
1403extern void __kmp_query_cpuid(kmp_cpuinfo_t *p);
1404
1405#if KMP_OS_UNIX
1406// subleaf is only needed for cache and topology discovery and can be set to
1407// zero in most cases
1408static inline void __kmp_x86_cpuid(int leaf, int subleaf, struct kmp_cpuid *p) {
1409 __asm__ __volatile__("cpuid"
1410 : "=a"(p->eax), "=b"(p->ebx), "=c"(p->ecx), "=d"(p->edx)
1411 : "a"(leaf), "c"(subleaf));
1412}
1413// Load p into FPU control word
1414static inline void __kmp_load_x87_fpu_control_word(const kmp_int16 *p) {
1415 __asm__ __volatile__("fldcw %0" : : "m"(*p));
1416}
1417// Store FPU control word into p
1418static inline void __kmp_store_x87_fpu_control_word(kmp_int16 *p) {
1419 __asm__ __volatile__("fstcw %0" : "=m"(*p));
1420}
1421static inline void __kmp_clear_x87_fpu_status_word() {
1422#if KMP_MIC
1423 // 32-bit protected mode x87 FPU state
1424 struct x87_fpu_state {
1425 unsigned cw;
1426 unsigned sw;
1427 unsigned tw;
1428 unsigned fip;
1429 unsigned fips;
1430 unsigned fdp;
1431 unsigned fds;
1432 };
1433 struct x87_fpu_state fpu_state = {0, 0, 0, 0, 0, 0, 0};
1434 __asm__ __volatile__("fstenv %0\n\t" // store FP env
1435 "andw $0x7f00, %1\n\t" // clear 0-7,15 bits of FP SW
1436 "fldenv %0\n\t" // load FP env back
1437 : "+m"(fpu_state), "+m"(fpu_state.sw));
1438#else
1439 __asm__ __volatile__("fnclex");
1440#endif // KMP_MIC
1441}
1442#if __SSE__
1443static inline void __kmp_load_mxcsr(const kmp_uint32 *p) { _mm_setcsr(*p); }
1444static inline void __kmp_store_mxcsr(kmp_uint32 *p) { *p = _mm_getcsr(); }
1445#else
1446static inline void __kmp_load_mxcsr(const kmp_uint32 *p) {}
1447static inline void __kmp_store_mxcsr(kmp_uint32 *p) { *p = 0; }
1448#endif
1449#else
1450// Windows still has these as external functions in assembly file
1451extern void __kmp_x86_cpuid(int mode, int mode2, struct kmp_cpuid *p);
1452extern void __kmp_load_x87_fpu_control_word(const kmp_int16 *p);
1453extern void __kmp_store_x87_fpu_control_word(kmp_int16 *p);
1454extern void __kmp_clear_x87_fpu_status_word();
1455static inline void __kmp_load_mxcsr(const kmp_uint32 *p) { _mm_setcsr(*p); }
1456static inline void __kmp_store_mxcsr(kmp_uint32 *p) { *p = _mm_getcsr(); }
1457#endif // KMP_OS_UNIX
1458
1459#define KMP_X86_MXCSR_MASK 0xffffffc0 /* ignore status flags (6 lsb) */
1460
1461// User-level Monitor/Mwait
1462#if KMP_HAVE_UMWAIT
1463// We always try for UMWAIT first
1464#if KMP_HAVE_WAITPKG_INTRINSICS
1465#if KMP_HAVE_IMMINTRIN_H
1466#include <immintrin.h>
1467#elif KMP_HAVE_INTRIN_H
1468#include <intrin.h>
1469#endif
1470#endif // KMP_HAVE_WAITPKG_INTRINSICS
1471
1472KMP_ATTRIBUTE_TARGET_WAITPKG
1473static inline int __kmp_tpause(uint32_t hint, uint64_t counter) {
1474#if !KMP_HAVE_WAITPKG_INTRINSICS
1475 uint32_t timeHi = uint32_t(counter >> 32);
1476 uint32_t timeLo = uint32_t(counter & 0xffffffff);
1477 char flag;
1478 __asm__ volatile("#tpause\n.byte 0x66, 0x0F, 0xAE, 0xF1\n"
1479 "setb %0"
1480 // The "=q" restraint means any register accessible as rl
1481 // in 32-bit mode: a, b, c, and d;
1482 // in 64-bit mode: any integer register
1483 : "=q"(flag)
1484 : "a"(timeLo), "d"(timeHi), "c"(hint)
1485 :);
1486 return flag;
1487#else
1488 return _tpause(hint, counter);
1489#endif
1490}
1491KMP_ATTRIBUTE_TARGET_WAITPKG
1492static inline void __kmp_umonitor(void *cacheline) {
1493#if !KMP_HAVE_WAITPKG_INTRINSICS
1494 __asm__ volatile("# umonitor\n.byte 0xF3, 0x0F, 0xAE, 0x01 "
1495 :
1496 : "a"(cacheline)
1497 :);
1498#else
1499 _umonitor(cacheline);
1500#endif
1501}
1502KMP_ATTRIBUTE_TARGET_WAITPKG
1503static inline int __kmp_umwait(uint32_t hint, uint64_t counter) {
1504#if !KMP_HAVE_WAITPKG_INTRINSICS
1505 uint32_t timeHi = uint32_t(counter >> 32);
1506 uint32_t timeLo = uint32_t(counter & 0xffffffff);
1507 char flag;
1508 __asm__ volatile("#umwait\n.byte 0xF2, 0x0F, 0xAE, 0xF1\n"
1509 "setb %0"
1510 // The "=q" restraint means any register accessible as rl
1511 // in 32-bit mode: a, b, c, and d;
1512 // in 64-bit mode: any integer register
1513 : "=q"(flag)
1514 : "a"(timeLo), "d"(timeHi), "c"(hint)
1515 :);
1516 return flag;
1517#else
1518 return _umwait(hint, counter);
1519#endif
1520}
1521#elif KMP_HAVE_MWAIT
1522#if KMP_OS_UNIX
1523#include <pmmintrin.h>
1524#else
1525#include <intrin.h>
1526#endif
1527#if KMP_OS_UNIX
1528__attribute__((target("sse3")))
1529#endif
1530static inline void
1531__kmp_mm_monitor(void *cacheline, unsigned extensions, unsigned hints) {
1532 _mm_monitor(cacheline, extensions, hints);
1533}
1534#if KMP_OS_UNIX
1535__attribute__((target("sse3")))
1536#endif
1537static inline void
1538__kmp_mm_mwait(unsigned extensions, unsigned hints) {
1539 _mm_mwait(extensions, hints);
1540}
1541#endif // KMP_HAVE_UMWAIT
1542
1543#if KMP_ARCH_X86
1544extern void __kmp_x86_pause(void);
1545#elif KMP_MIC
1546// Performance testing on KNC (C0QS-7120 P/A/X/D, 61-core, 16 GB Memory) showed
1547// regression after removal of extra PAUSE from spin loops. Changing
1548// the delay from 100 to 300 showed even better performance than double PAUSE
1549// on Spec OMP2001 and LCPC tasking tests, no regressions on EPCC.
1550static inline void __kmp_x86_pause(void) { _mm_delay_32(300); }
1551#else
1552static inline void __kmp_x86_pause(void) { _mm_pause(); }
1553#endif
1554#define KMP_CPU_PAUSE() __kmp_x86_pause()
1555#elif KMP_ARCH_PPC64
1556#define KMP_PPC64_PRI_LOW() __asm__ volatile("or 1, 1, 1")
1557#define KMP_PPC64_PRI_MED() __asm__ volatile("or 2, 2, 2")
1558#define KMP_PPC64_PRI_LOC_MB() __asm__ volatile("" : : : "memory")
1559#define KMP_CPU_PAUSE() \
1560 do { \
1561 KMP_PPC64_PRI_LOW(); \
1562 KMP_PPC64_PRI_MED(); \
1563 KMP_PPC64_PRI_LOC_MB(); \
1564 } while (0)
1565#else
1566#define KMP_CPU_PAUSE() /* nothing to do */
1567#endif
1568
1569#define KMP_INIT_YIELD(count) \
1570 { (count) = __kmp_yield_init; }
1571
1572#define KMP_INIT_BACKOFF(time) \
1573 { (time) = __kmp_pause_init; }
1574
1575#define KMP_OVERSUBSCRIBED \
1576 (TCR_4(__kmp_nth) > (__kmp_avail_proc ? __kmp_avail_proc : __kmp_xproc))
1577
1578#define KMP_TRY_YIELD \
1579 ((__kmp_use_yield == 1) || (__kmp_use_yield == 2 && (KMP_OVERSUBSCRIBED)))
1580
1581#define KMP_TRY_YIELD_OVERSUB \
1582 ((__kmp_use_yield == 1 || __kmp_use_yield == 2) && (KMP_OVERSUBSCRIBED))
1583
1584#define KMP_YIELD(cond) \
1585 { \
1586 KMP_CPU_PAUSE(); \
1587 if ((cond) && (KMP_TRY_YIELD)) \
1588 __kmp_yield(); \
1589 }
1590
1591#define KMP_YIELD_OVERSUB() \
1592 { \
1593 KMP_CPU_PAUSE(); \
1594 if ((KMP_TRY_YIELD_OVERSUB)) \
1595 __kmp_yield(); \
1596 }
1597
1598// Note the decrement of 2 in the following Macros. With KMP_LIBRARY=turnaround,
1599// there should be no yielding since initial value from KMP_INIT_YIELD() is odd.
1600#define KMP_YIELD_SPIN(count) \
1601 { \
1602 KMP_CPU_PAUSE(); \
1603 if (KMP_TRY_YIELD) { \
1604 (count) -= 2; \
1605 if (!(count)) { \
1606 __kmp_yield(); \
1607 (count) = __kmp_yield_next; \
1608 } \
1609 } \
1610 }
1611
1612// If TPAUSE is available & enabled, use it. If oversubscribed, use the slower
1613// (C0.2) state, which improves performance of other SMT threads on the same
1614// core, otherwise, use the fast (C0.1) default state, or whatever the user has
1615// requested. Uses a timed TPAUSE, and exponential backoff. If TPAUSE isn't
1616// available, fall back to the regular CPU pause and yield combination.
1617#if KMP_HAVE_UMWAIT
1618#define KMP_TPAUSE_MAX_MASK ((kmp_uint64)0xFFFF)
1619#define KMP_YIELD_OVERSUB_ELSE_SPIN(count, time) \
1620 { \
1621 if (__kmp_tpause_enabled) { \
1622 if (KMP_OVERSUBSCRIBED) { \
1623 __kmp_tpause(0, (time)); \
1624 } else { \
1625 __kmp_tpause(__kmp_tpause_hint, (time)); \
1626 } \
1627 (time) = (time << 1 | 1) & KMP_TPAUSE_MAX_MASK; \
1628 } else { \
1629 KMP_CPU_PAUSE(); \
1630 if ((KMP_TRY_YIELD_OVERSUB)) { \
1631 __kmp_yield(); \
1632 } else if (__kmp_use_yield == 1) { \
1633 (count) -= 2; \
1634 if (!(count)) { \
1635 __kmp_yield(); \
1636 (count) = __kmp_yield_next; \
1637 } \
1638 } \
1639 } \
1640 }
1641#else
1642#define KMP_YIELD_OVERSUB_ELSE_SPIN(count, time) \
1643 { \
1644 KMP_CPU_PAUSE(); \
1645 if ((KMP_TRY_YIELD_OVERSUB)) \
1646 __kmp_yield(); \
1647 else if (__kmp_use_yield == 1) { \
1648 (count) -= 2; \
1649 if (!(count)) { \
1650 __kmp_yield(); \
1651 (count) = __kmp_yield_next; \
1652 } \
1653 } \
1654 }
1655#endif // KMP_HAVE_UMWAIT
1656
1657/* ------------------------------------------------------------------------ */
1658/* Support datatypes for the orphaned construct nesting checks. */
1659/* ------------------------------------------------------------------------ */
1660
1661/* When adding to this enum, add its corresponding string in cons_text_c[]
1662 * array in kmp_error.cpp */
1663enum cons_type {
1664 ct_none,
1665 ct_parallel,
1666 ct_pdo,
1667 ct_pdo_ordered,
1668 ct_psections,
1669 ct_psingle,
1670 ct_critical,
1671 ct_ordered_in_parallel,
1672 ct_ordered_in_pdo,
1673 ct_master,
1674 ct_reduce,
1675 ct_barrier,
1676 ct_masked
1677};
1678
1679#define IS_CONS_TYPE_ORDERED(ct) ((ct) == ct_pdo_ordered)
1680
1681struct cons_data {
1682 ident_t const *ident;
1683 enum cons_type type;
1684 int prev;
1685 kmp_user_lock_p
1686 name; /* address exclusively for critical section name comparison */
1687};
1688
1689struct cons_header {
1690 int p_top, w_top, s_top;
1691 int stack_size, stack_top;
1692 struct cons_data *stack_data;
1693};
1694
1695struct kmp_region_info {
1696 char *text;
1697 int offset[KMP_MAX_FIELDS];
1698 int length[KMP_MAX_FIELDS];
1699};
1700
1701/* ---------------------------------------------------------------------- */
1702/* ---------------------------------------------------------------------- */
1703
1704#if KMP_OS_WINDOWS
1705typedef HANDLE kmp_thread_t;
1706typedef DWORD kmp_key_t;
1707#endif /* KMP_OS_WINDOWS */
1708
1709#if KMP_OS_UNIX
1710typedef pthread_t kmp_thread_t;
1711typedef pthread_key_t kmp_key_t;
1712#endif
1713
1714extern kmp_key_t __kmp_gtid_threadprivate_key;
1715
1716typedef struct kmp_sys_info {
1717 long maxrss; /* the maximum resident set size utilized (in kilobytes) */
1718 long minflt; /* the number of page faults serviced without any I/O */
1719 long majflt; /* the number of page faults serviced that required I/O */
1720 long nswap; /* the number of times a process was "swapped" out of memory */
1721 long inblock; /* the number of times the file system had to perform input */
1722 long oublock; /* the number of times the file system had to perform output */
1723 long nvcsw; /* the number of times a context switch was voluntarily */
1724 long nivcsw; /* the number of times a context switch was forced */
1725} kmp_sys_info_t;
1726
1727#if USE_ITT_BUILD
1728// We cannot include "kmp_itt.h" due to circular dependency. Declare the only
1729// required type here. Later we will check the type meets requirements.
1730typedef int kmp_itt_mark_t;
1731#define KMP_ITT_DEBUG 0
1732#endif /* USE_ITT_BUILD */
1733
1734typedef kmp_int32 kmp_critical_name[8];
1735
1745typedef void (*kmpc_micro)(kmp_int32 *global_tid, kmp_int32 *bound_tid, ...);
1746typedef void (*kmpc_micro_bound)(kmp_int32 *bound_tid, kmp_int32 *bound_nth,
1747 ...);
1748
1753/* ---------------------------------------------------------------------------
1754 */
1755/* Threadprivate initialization/finalization function declarations */
1756
1757/* for non-array objects: __kmpc_threadprivate_register() */
1758
1763typedef void *(*kmpc_ctor)(void *);
1764
1769typedef void (*kmpc_dtor)(
1770 void * /*, size_t */); /* 2nd arg: magic number for KCC unused by Intel
1771 compiler */
1776typedef void *(*kmpc_cctor)(void *, void *);
1777
1778/* for array objects: __kmpc_threadprivate_register_vec() */
1779/* First arg: "this" pointer */
1780/* Last arg: number of array elements */
1786typedef void *(*kmpc_ctor_vec)(void *, size_t);
1792typedef void (*kmpc_dtor_vec)(void *, size_t);
1798typedef void *(*kmpc_cctor_vec)(void *, void *,
1799 size_t); /* function unused by compiler */
1800
1805/* keeps tracked of threadprivate cache allocations for cleanup later */
1806typedef struct kmp_cached_addr {
1807 void **addr; /* address of allocated cache */
1808 void ***compiler_cache; /* pointer to compiler's cache */
1809 void *data; /* pointer to global data */
1810 struct kmp_cached_addr *next; /* pointer to next cached address */
1811} kmp_cached_addr_t;
1812
1813struct private_data {
1814 struct private_data *next; /* The next descriptor in the list */
1815 void *data; /* The data buffer for this descriptor */
1816 int more; /* The repeat count for this descriptor */
1817 size_t size; /* The data size for this descriptor */
1818};
1819
1820struct private_common {
1821 struct private_common *next;
1822 struct private_common *link;
1823 void *gbl_addr;
1824 void *par_addr; /* par_addr == gbl_addr for PRIMARY thread */
1825 size_t cmn_size;
1826};
1827
1828struct shared_common {
1829 struct shared_common *next;
1830 struct private_data *pod_init;
1831 void *obj_init;
1832 void *gbl_addr;
1833 union {
1834 kmpc_ctor ctor;
1835 kmpc_ctor_vec ctorv;
1836 } ct;
1837 union {
1838 kmpc_cctor cctor;
1839 kmpc_cctor_vec cctorv;
1840 } cct;
1841 union {
1842 kmpc_dtor dtor;
1843 kmpc_dtor_vec dtorv;
1844 } dt;
1845 size_t vec_len;
1846 int is_vec;
1847 size_t cmn_size;
1848};
1849
1850#define KMP_HASH_TABLE_LOG2 9 /* log2 of the hash table size */
1851#define KMP_HASH_TABLE_SIZE \
1852 (1 << KMP_HASH_TABLE_LOG2) /* size of the hash table */
1853#define KMP_HASH_SHIFT 3 /* throw away this many low bits from the address */
1854#define KMP_HASH(x) \
1855 ((((kmp_uintptr_t)x) >> KMP_HASH_SHIFT) & (KMP_HASH_TABLE_SIZE - 1))
1856
1857struct common_table {
1858 struct private_common *data[KMP_HASH_TABLE_SIZE];
1859};
1860
1861struct shared_table {
1862 struct shared_common *data[KMP_HASH_TABLE_SIZE];
1863};
1864
1865/* ------------------------------------------------------------------------ */
1866
1867#if KMP_USE_HIER_SCHED
1868// Shared barrier data that exists inside a single unit of the scheduling
1869// hierarchy
1870typedef struct kmp_hier_private_bdata_t {
1871 kmp_int32 num_active;
1872 kmp_uint64 index;
1873 kmp_uint64 wait_val[2];
1874} kmp_hier_private_bdata_t;
1875#endif
1876
1877typedef struct kmp_sched_flags {
1878 unsigned ordered : 1;
1879 unsigned nomerge : 1;
1880 unsigned contains_last : 1;
1881 unsigned use_hier : 1; // Used in KMP_USE_HIER_SCHED code
1882 unsigned use_hybrid : 1; // Used in KMP_WEIGHTED_ITERATIONS_SUPPORTED code
1883 unsigned unused : 27;
1884} kmp_sched_flags_t;
1885
1886KMP_BUILD_ASSERT(sizeof(kmp_sched_flags_t) == 4);
1887
1888#if KMP_STATIC_STEAL_ENABLED
1889typedef struct KMP_ALIGN_CACHE dispatch_private_info32 {
1890 kmp_int32 count;
1891 kmp_int32 ub;
1892 /* Adding KMP_ALIGN_CACHE here doesn't help / can hurt performance */
1893 kmp_int32 lb;
1894 kmp_int32 st;
1895 kmp_int32 tc;
1896 kmp_lock_t *steal_lock; // lock used for chunk stealing
1897
1898 kmp_uint32 ordered_lower;
1899 kmp_uint32 ordered_upper;
1900
1901 // KMP_ALIGN(32) ensures (if the KMP_ALIGN macro is turned on)
1902 // a) parm3 is properly aligned and
1903 // b) all parm1-4 are on the same cache line.
1904 // Because of parm1-4 are used together, performance seems to be better
1905 // if they are on the same cache line (not measured though).
1906
1907 struct KMP_ALIGN(32) {
1908 kmp_int32 parm1;
1909 kmp_int32 parm2;
1910 kmp_int32 parm3;
1911 kmp_int32 parm4;
1912 };
1913
1914#if KMP_WEIGHTED_ITERATIONS_SUPPORTED
1915 kmp_uint32 pchunks;
1916 kmp_uint32 num_procs_with_pcore;
1917 kmp_int32 first_thread_with_ecore;
1918#endif
1919#if KMP_OS_WINDOWS
1920 kmp_int32 last_upper;
1921#endif /* KMP_OS_WINDOWS */
1922} dispatch_private_info32_t;
1923
1924#if CACHE_LINE <= 128
1925KMP_BUILD_ASSERT(sizeof(dispatch_private_info32_t) <= 128);
1926#endif
1927
1928typedef struct KMP_ALIGN_CACHE dispatch_private_info64 {
1929 kmp_int64 count; // current chunk number for static & static-steal scheduling
1930 kmp_int64 ub; /* upper-bound */
1931 /* Adding KMP_ALIGN_CACHE here doesn't help / can hurt performance */
1932 kmp_int64 lb; /* lower-bound */
1933 kmp_int64 st; /* stride */
1934 kmp_int64 tc; /* trip count (number of iterations) */
1935 kmp_lock_t *steal_lock; // lock used for chunk stealing
1936
1937 kmp_uint64 ordered_lower;
1938 kmp_uint64 ordered_upper;
1939 /* parm[1-4] are used in different ways by different scheduling algorithms */
1940
1941 // KMP_ALIGN(32) ensures ( if the KMP_ALIGN macro is turned on )
1942 // a) parm3 is properly aligned and
1943 // b) all parm1-4 are in the same cache line.
1944 // Because of parm1-4 are used together, performance seems to be better
1945 // if they are in the same line (not measured though).
1946 struct KMP_ALIGN(32) {
1947 kmp_int64 parm1;
1948 kmp_int64 parm2;
1949 kmp_int64 parm3;
1950 kmp_int64 parm4;
1951 };
1952
1953#if KMP_WEIGHTED_ITERATIONS_SUPPORTED
1954 kmp_uint64 pchunks;
1955 kmp_uint64 num_procs_with_pcore;
1956 kmp_int64 first_thread_with_ecore;
1957#endif
1958
1959#if KMP_OS_WINDOWS
1960 kmp_int64 last_upper;
1961#endif /* KMP_OS_WINDOWS */
1962} dispatch_private_info64_t;
1963
1964#if CACHE_LINE <= 128
1965KMP_BUILD_ASSERT(sizeof(dispatch_private_info64_t) <= 128);
1966#endif
1967
1968#else /* KMP_STATIC_STEAL_ENABLED */
1969typedef struct KMP_ALIGN_CACHE dispatch_private_info32 {
1970 kmp_int32 lb;
1971 kmp_int32 ub;
1972 kmp_int32 st;
1973 kmp_int32 tc;
1974
1975 kmp_int32 parm1;
1976 kmp_int32 parm2;
1977 kmp_int32 parm3;
1978 kmp_int32 parm4;
1979
1980 kmp_int32 count;
1981
1982 kmp_uint32 ordered_lower;
1983 kmp_uint32 ordered_upper;
1984#if KMP_OS_WINDOWS
1985 kmp_int32 last_upper;
1986#endif /* KMP_OS_WINDOWS */
1987} dispatch_private_info32_t;
1988
1989typedef struct KMP_ALIGN_CACHE dispatch_private_info64 {
1990 kmp_int64 lb; /* lower-bound */
1991 kmp_int64 ub; /* upper-bound */
1992 kmp_int64 st; /* stride */
1993 kmp_int64 tc; /* trip count (number of iterations) */
1994
1995 /* parm[1-4] are used in different ways by different scheduling algorithms */
1996 kmp_int64 parm1;
1997 kmp_int64 parm2;
1998 kmp_int64 parm3;
1999 kmp_int64 parm4;
2000
2001 kmp_int64 count; /* current chunk number for static scheduling */
2002
2003 kmp_uint64 ordered_lower;
2004 kmp_uint64 ordered_upper;
2005#if KMP_OS_WINDOWS
2006 kmp_int64 last_upper;
2007#endif /* KMP_OS_WINDOWS */
2008} dispatch_private_info64_t;
2009#endif /* KMP_STATIC_STEAL_ENABLED */
2010
2011typedef struct KMP_ALIGN_CACHE dispatch_private_info {
2012 union private_info {
2013 dispatch_private_info32_t p32;
2014 dispatch_private_info64_t p64;
2015 } u;
2016 enum sched_type schedule; /* scheduling algorithm */
2017 kmp_sched_flags_t flags; /* flags (e.g., ordered, nomerge, etc.) */
2018 std::atomic<kmp_uint32> steal_flag; // static_steal only, state of a buffer
2019 kmp_int32 ordered_bumped;
2020 // Stack of buffers for nest of serial regions
2021 struct dispatch_private_info *next;
2022 kmp_int32 type_size; /* the size of types in private_info */
2023#if KMP_USE_HIER_SCHED
2024 kmp_int32 hier_id;
2025 void *parent; /* hierarchical scheduling parent pointer */
2026#endif
2027 enum cons_type pushed_ws;
2028} dispatch_private_info_t;
2029
2030typedef struct dispatch_shared_info32 {
2031 /* chunk index under dynamic, number of idle threads under static-steal;
2032 iteration index otherwise */
2033 volatile kmp_uint32 iteration;
2034 volatile kmp_int32 num_done;
2035 volatile kmp_uint32 ordered_iteration;
2036 // Dummy to retain the structure size after making ordered_iteration scalar
2037 kmp_int32 ordered_dummy[KMP_MAX_ORDERED - 1];
2038} dispatch_shared_info32_t;
2039
2040typedef struct dispatch_shared_info64 {
2041 /* chunk index under dynamic, number of idle threads under static-steal;
2042 iteration index otherwise */
2043 volatile kmp_uint64 iteration;
2044 volatile kmp_int64 num_done;
2045 volatile kmp_uint64 ordered_iteration;
2046 // Dummy to retain the structure size after making ordered_iteration scalar
2047 kmp_int64 ordered_dummy[KMP_MAX_ORDERED - 3];
2048} dispatch_shared_info64_t;
2049
2050typedef struct dispatch_shared_info {
2051 union shared_info {
2052 dispatch_shared_info32_t s32;
2053 dispatch_shared_info64_t s64;
2054 } u;
2055 volatile kmp_uint32 buffer_index;
2056 volatile kmp_int32 doacross_buf_idx; // teamwise index
2057 volatile kmp_uint32 *doacross_flags; // shared array of iteration flags (0/1)
2058 kmp_int32 doacross_num_done; // count finished threads
2059#if KMP_USE_HIER_SCHED
2060 void *hier;
2061#endif
2062#if KMP_USE_HWLOC
2063 // When linking with libhwloc, the ORDERED EPCC test slows down on big
2064 // machines (> 48 cores). Performance analysis showed that a cache thrash
2065 // was occurring and this padding helps alleviate the problem.
2066 char padding[64];
2067#endif
2068} dispatch_shared_info_t;
2069
2070typedef struct kmp_disp {
2071 /* Vector for ORDERED SECTION */
2072 void (*th_deo_fcn)(int *gtid, int *cid, ident_t *);
2073 /* Vector for END ORDERED SECTION */
2074 void (*th_dxo_fcn)(int *gtid, int *cid, ident_t *);
2075
2076 dispatch_shared_info_t *th_dispatch_sh_current;
2077 dispatch_private_info_t *th_dispatch_pr_current;
2078
2079 dispatch_private_info_t *th_disp_buffer;
2080 kmp_uint32 th_disp_index;
2081 kmp_int32 th_doacross_buf_idx; // thread's doacross buffer index
2082 volatile kmp_uint32 *th_doacross_flags; // pointer to shared array of flags
2083 kmp_int64 *th_doacross_info; // info on loop bounds
2084#if KMP_USE_INTERNODE_ALIGNMENT
2085 char more_padding[INTERNODE_CACHE_LINE];
2086#endif
2087} kmp_disp_t;
2088
2089/* ------------------------------------------------------------------------ */
2090/* Barrier stuff */
2091
2092/* constants for barrier state update */
2093#define KMP_INIT_BARRIER_STATE 0 /* should probably start from zero */
2094#define KMP_BARRIER_SLEEP_BIT 0 /* bit used for suspend/sleep part of state */
2095#define KMP_BARRIER_UNUSED_BIT 1 // bit that must never be set for valid state
2096#define KMP_BARRIER_BUMP_BIT 2 /* lsb used for bump of go/arrived state */
2097
2098#define KMP_BARRIER_SLEEP_STATE (1 << KMP_BARRIER_SLEEP_BIT)
2099#define KMP_BARRIER_UNUSED_STATE (1 << KMP_BARRIER_UNUSED_BIT)
2100#define KMP_BARRIER_STATE_BUMP (1 << KMP_BARRIER_BUMP_BIT)
2101
2102#if (KMP_BARRIER_SLEEP_BIT >= KMP_BARRIER_BUMP_BIT)
2103#error "Barrier sleep bit must be smaller than barrier bump bit"
2104#endif
2105#if (KMP_BARRIER_UNUSED_BIT >= KMP_BARRIER_BUMP_BIT)
2106#error "Barrier unused bit must be smaller than barrier bump bit"
2107#endif
2108
2109// Constants for release barrier wait state: currently, hierarchical only
2110#define KMP_BARRIER_NOT_WAITING 0 // Normal state; worker not in wait_sleep
2111#define KMP_BARRIER_OWN_FLAG \
2112 1 // Normal state; worker waiting on own b_go flag in release
2113#define KMP_BARRIER_PARENT_FLAG \
2114 2 // Special state; worker waiting on parent's b_go flag in release
2115#define KMP_BARRIER_SWITCH_TO_OWN_FLAG \
2116 3 // Special state; tells worker to shift from parent to own b_go
2117#define KMP_BARRIER_SWITCHING \
2118 4 // Special state; worker resets appropriate flag on wake-up
2119
2120#define KMP_NOT_SAFE_TO_REAP \
2121 0 // Thread th_reap_state: not safe to reap (tasking)
2122#define KMP_SAFE_TO_REAP 1 // Thread th_reap_state: safe to reap (not tasking)
2123
2124// The flag_type describes the storage used for the flag.
2125enum flag_type {
2126 flag32,
2127 flag64,
2128 atomic_flag64,
2129 flag_oncore,
2130 flag_unset
2131};
2132
2133enum barrier_type {
2134 bs_plain_barrier = 0, /* 0, All non-fork/join barriers (except reduction
2135 barriers if enabled) */
2136 bs_forkjoin_barrier, /* 1, All fork/join (parallel region) barriers */
2137#if KMP_FAST_REDUCTION_BARRIER
2138 bs_reduction_barrier, /* 2, All barriers that are used in reduction */
2139#endif // KMP_FAST_REDUCTION_BARRIER
2140 bs_last_barrier /* Just a placeholder to mark the end */
2141};
2142
2143// to work with reduction barriers just like with plain barriers
2144#if !KMP_FAST_REDUCTION_BARRIER
2145#define bs_reduction_barrier bs_plain_barrier
2146#endif // KMP_FAST_REDUCTION_BARRIER
2147
2148typedef enum kmp_bar_pat { /* Barrier communication patterns */
2149 bp_linear_bar =
2150 0, /* Single level (degenerate) tree */
2151 bp_tree_bar =
2152 1, /* Balanced tree with branching factor 2^n */
2153 bp_hyper_bar = 2, /* Hypercube-embedded tree with min
2154 branching factor 2^n */
2155 bp_hierarchical_bar = 3, /* Machine hierarchy tree */
2156 bp_dist_bar = 4, /* Distributed barrier */
2157 bp_last_bar /* Placeholder to mark the end */
2158} kmp_bar_pat_e;
2159
2160#define KMP_BARRIER_ICV_PUSH 1
2161
2162/* Record for holding the values of the internal controls stack records */
2163typedef struct kmp_internal_control {
2164 int serial_nesting_level; /* corresponds to the value of the
2165 th_team_serialized field */
2166 kmp_int8 dynamic; /* internal control for dynamic adjustment of threads (per
2167 thread) */
2168 kmp_int8
2169 bt_set; /* internal control for whether blocktime is explicitly set */
2170 int blocktime; /* internal control for blocktime */
2171#if KMP_USE_MONITOR
2172 int bt_intervals; /* internal control for blocktime intervals */
2173#endif
2174 int nproc; /* internal control for #threads for next parallel region (per
2175 thread) */
2176 int thread_limit; /* internal control for thread-limit-var */
2177 int task_thread_limit; /* internal control for thread-limit-var of a task*/
2178 int max_active_levels; /* internal control for max_active_levels */
2179 kmp_r_sched_t
2180 sched; /* internal control for runtime schedule {sched,chunk} pair */
2181 kmp_proc_bind_t proc_bind; /* internal control for affinity */
2182 kmp_int32 default_device; /* internal control for default device */
2183 struct kmp_internal_control *next;
2184} kmp_internal_control_t;
2185
2186static inline void copy_icvs(kmp_internal_control_t *dst,
2187 kmp_internal_control_t *src) {
2188 *dst = *src;
2189}
2190
2191/* Thread barrier needs volatile barrier fields */
2192typedef struct KMP_ALIGN_CACHE kmp_bstate {
2193 // th_fixed_icvs is aligned by virtue of kmp_bstate being aligned (and all
2194 // uses of it). It is not explicitly aligned below, because we *don't* want
2195 // it to be padded -- instead, we fit b_go into the same cache line with
2196 // th_fixed_icvs, enabling NGO cache lines stores in the hierarchical barrier.
2197 kmp_internal_control_t th_fixed_icvs; // Initial ICVs for the thread
2198 // Tuck b_go into end of th_fixed_icvs cache line, so it can be stored with
2199 // same NGO store
2200 volatile kmp_uint64 b_go; // STATE => task should proceed (hierarchical)
2201 KMP_ALIGN_CACHE volatile kmp_uint64
2202 b_arrived; // STATE => task reached synch point.
2203 kmp_uint32 *skip_per_level;
2204 kmp_uint32 my_level;
2205 kmp_int32 parent_tid;
2206 kmp_int32 old_tid;
2207 kmp_uint32 depth;
2208 struct kmp_bstate *parent_bar;
2209 kmp_team_t *team;
2210 kmp_uint64 leaf_state;
2211 kmp_uint32 nproc;
2212 kmp_uint8 base_leaf_kids;
2213 kmp_uint8 leaf_kids;
2214 kmp_uint8 offset;
2215 kmp_uint8 wait_flag;
2216 kmp_uint8 use_oncore_barrier;
2217#if USE_DEBUGGER
2218 // The following field is intended for the debugger solely. Only the worker
2219 // thread itself accesses this field: the worker increases it by 1 when it
2220 // arrives to a barrier.
2221 KMP_ALIGN_CACHE kmp_uint b_worker_arrived;
2222#endif /* USE_DEBUGGER */
2223} kmp_bstate_t;
2224
2225union KMP_ALIGN_CACHE kmp_barrier_union {
2226 double b_align; /* use worst case alignment */
2227 char b_pad[KMP_PAD(kmp_bstate_t, CACHE_LINE)];
2228 kmp_bstate_t bb;
2229};
2230
2231typedef union kmp_barrier_union kmp_balign_t;
2232
2233/* Team barrier needs only non-volatile arrived counter */
2234union KMP_ALIGN_CACHE kmp_barrier_team_union {
2235 double b_align; /* use worst case alignment */
2236 char b_pad[CACHE_LINE];
2237 struct {
2238 kmp_uint64 b_arrived; /* STATE => task reached synch point. */
2239#if USE_DEBUGGER
2240 // The following two fields are indended for the debugger solely. Only
2241 // primary thread of the team accesses these fields: the first one is
2242 // increased by 1 when the primary thread arrives to a barrier, the second
2243 // one is increased by one when all the threads arrived.
2244 kmp_uint b_master_arrived;
2245 kmp_uint b_team_arrived;
2246#endif
2247 };
2248};
2249
2250typedef union kmp_barrier_team_union kmp_balign_team_t;
2251
2252/* Padding for Linux* OS pthreads condition variables and mutexes used to signal
2253 threads when a condition changes. This is to workaround an NPTL bug where
2254 padding was added to pthread_cond_t which caused the initialization routine
2255 to write outside of the structure if compiled on pre-NPTL threads. */
2256#if KMP_OS_WINDOWS
2257typedef struct kmp_win32_mutex {
2258 /* The Lock */
2259 CRITICAL_SECTION cs;
2260} kmp_win32_mutex_t;
2261
2262typedef struct kmp_win32_cond {
2263 /* Count of the number of waiters. */
2264 int waiters_count_;
2265
2266 /* Serialize access to <waiters_count_> */
2267 kmp_win32_mutex_t waiters_count_lock_;
2268
2269 /* Number of threads to release via a <cond_broadcast> or a <cond_signal> */
2270 int release_count_;
2271
2272 /* Keeps track of the current "generation" so that we don't allow */
2273 /* one thread to steal all the "releases" from the broadcast. */
2274 int wait_generation_count_;
2275
2276 /* A manual-reset event that's used to block and release waiting threads. */
2277 HANDLE event_;
2278} kmp_win32_cond_t;
2279#endif
2280
2281#if KMP_OS_UNIX
2282
2283union KMP_ALIGN_CACHE kmp_cond_union {
2284 double c_align;
2285 char c_pad[CACHE_LINE];
2286 pthread_cond_t c_cond;
2287};
2288
2289typedef union kmp_cond_union kmp_cond_align_t;
2290
2291union KMP_ALIGN_CACHE kmp_mutex_union {
2292 double m_align;
2293 char m_pad[CACHE_LINE];
2294 pthread_mutex_t m_mutex;
2295};
2296
2297typedef union kmp_mutex_union kmp_mutex_align_t;
2298
2299#endif /* KMP_OS_UNIX */
2300
2301typedef struct kmp_desc_base {
2302 void *ds_stackbase;
2303 size_t ds_stacksize;
2304 int ds_stackgrow;
2305 kmp_thread_t ds_thread;
2306 volatile int ds_tid;
2307 int ds_gtid;
2308#if KMP_OS_WINDOWS
2309 volatile int ds_alive;
2310 DWORD ds_thread_id;
2311/* ds_thread keeps thread handle on Windows* OS. It is enough for RTL purposes.
2312 However, debugger support (libomp_db) cannot work with handles, because they
2313 uncomparable. For example, debugger requests info about thread with handle h.
2314 h is valid within debugger process, and meaningless within debugee process.
2315 Even if h is duped by call to DuplicateHandle(), so the result h' is valid
2316 within debugee process, but it is a *new* handle which does *not* equal to
2317 any other handle in debugee... The only way to compare handles is convert
2318 them to system-wide ids. GetThreadId() function is available only in
2319 Longhorn and Server 2003. :-( In contrast, GetCurrentThreadId() is available
2320 on all Windows* OS flavours (including Windows* 95). Thus, we have to get
2321 thread id by call to GetCurrentThreadId() from within the thread and save it
2322 to let libomp_db identify threads. */
2323#endif /* KMP_OS_WINDOWS */
2324} kmp_desc_base_t;
2325
2326typedef union KMP_ALIGN_CACHE kmp_desc {
2327 double ds_align; /* use worst case alignment */
2328 char ds_pad[KMP_PAD(kmp_desc_base_t, CACHE_LINE)];
2329 kmp_desc_base_t ds;
2330} kmp_desc_t;
2331
2332typedef struct kmp_local {
2333 volatile int this_construct; /* count of single's encountered by thread */
2334 void *reduce_data;
2335#if KMP_USE_BGET
2336 void *bget_data;
2337 void *bget_list;
2338#if !USE_CMP_XCHG_FOR_BGET
2339#ifdef USE_QUEUING_LOCK_FOR_BGET
2340 kmp_lock_t bget_lock; /* Lock for accessing bget free list */
2341#else
2342 kmp_bootstrap_lock_t bget_lock; // Lock for accessing bget free list. Must be
2343// bootstrap lock so we can use it at library
2344// shutdown.
2345#endif /* USE_LOCK_FOR_BGET */
2346#endif /* ! USE_CMP_XCHG_FOR_BGET */
2347#endif /* KMP_USE_BGET */
2348
2349 PACKED_REDUCTION_METHOD_T
2350 packed_reduction_method; /* stored by __kmpc_reduce*(), used by
2351 __kmpc_end_reduce*() */
2352
2353} kmp_local_t;
2354
2355#define KMP_CHECK_UPDATE(a, b) \
2356 if ((a) != (b)) \
2357 (a) = (b)
2358#define KMP_CHECK_UPDATE_SYNC(a, b) \
2359 if ((a) != (b)) \
2360 TCW_SYNC_PTR((a), (b))
2361
2362#define get__blocktime(xteam, xtid) \
2363 ((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.blocktime)
2364#define get__bt_set(xteam, xtid) \
2365 ((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.bt_set)
2366#if KMP_USE_MONITOR
2367#define get__bt_intervals(xteam, xtid) \
2368 ((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.bt_intervals)
2369#endif
2370
2371#define get__dynamic_2(xteam, xtid) \
2372 ((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.dynamic)
2373#define get__nproc_2(xteam, xtid) \
2374 ((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.nproc)
2375#define get__sched_2(xteam, xtid) \
2376 ((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.sched)
2377
2378#define set__blocktime_team(xteam, xtid, xval) \
2379 (((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.blocktime) = \
2380 (xval))
2381
2382#if KMP_USE_MONITOR
2383#define set__bt_intervals_team(xteam, xtid, xval) \
2384 (((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.bt_intervals) = \
2385 (xval))
2386#endif
2387
2388#define set__bt_set_team(xteam, xtid, xval) \
2389 (((xteam)->t.t_threads[(xtid)]->th.th_current_task->td_icvs.bt_set) = (xval))
2390
2391#define set__dynamic(xthread, xval) \
2392 (((xthread)->th.th_current_task->td_icvs.dynamic) = (xval))
2393#define get__dynamic(xthread) \
2394 (((xthread)->th.th_current_task->td_icvs.dynamic) ? (FTN_TRUE) : (FTN_FALSE))
2395
2396#define set__nproc(xthread, xval) \
2397 (((xthread)->th.th_current_task->td_icvs.nproc) = (xval))
2398
2399#define set__thread_limit(xthread, xval) \
2400 (((xthread)->th.th_current_task->td_icvs.thread_limit) = (xval))
2401
2402#define set__max_active_levels(xthread, xval) \
2403 (((xthread)->th.th_current_task->td_icvs.max_active_levels) = (xval))
2404
2405#define get__max_active_levels(xthread) \
2406 ((xthread)->th.th_current_task->td_icvs.max_active_levels)
2407
2408#define set__sched(xthread, xval) \
2409 (((xthread)->th.th_current_task->td_icvs.sched) = (xval))
2410
2411#define set__proc_bind(xthread, xval) \
2412 (((xthread)->th.th_current_task->td_icvs.proc_bind) = (xval))
2413#define get__proc_bind(xthread) \
2414 ((xthread)->th.th_current_task->td_icvs.proc_bind)
2415
2416// OpenMP tasking data structures
2417
2418typedef enum kmp_tasking_mode {
2419 tskm_immediate_exec = 0,
2420 tskm_extra_barrier = 1,
2421 tskm_task_teams = 2,
2422 tskm_max = 2
2423} kmp_tasking_mode_t;
2424
2425extern kmp_tasking_mode_t
2426 __kmp_tasking_mode; /* determines how/when to execute tasks */
2427extern int __kmp_task_stealing_constraint;
2428extern int __kmp_enable_task_throttling;
2429extern kmp_int32 __kmp_default_device; // Set via OMP_DEFAULT_DEVICE if
2430// specified, defaults to 0 otherwise
2431// Set via OMP_MAX_TASK_PRIORITY if specified, defaults to 0 otherwise
2432extern kmp_int32 __kmp_max_task_priority;
2433// Set via KMP_TASKLOOP_MIN_TASKS if specified, defaults to 0 otherwise
2434extern kmp_uint64 __kmp_taskloop_min_tasks;
2435
2436/* NOTE: kmp_taskdata_t and kmp_task_t structures allocated in single block with
2437 taskdata first */
2438#define KMP_TASK_TO_TASKDATA(task) (((kmp_taskdata_t *)task) - 1)
2439#define KMP_TASKDATA_TO_TASK(taskdata) (kmp_task_t *)(taskdata + 1)
2440
2441// The tt_found_tasks flag is a signal to all threads in the team that tasks
2442// were spawned and queued since the previous barrier release.
2443#define KMP_TASKING_ENABLED(task_team) \
2444 (TRUE == TCR_SYNC_4((task_team)->tt.tt_found_tasks))
2452typedef kmp_int32 (*kmp_routine_entry_t)(kmp_int32, void *);
2453
2454typedef union kmp_cmplrdata {
2455 kmp_int32 priority;
2456 kmp_routine_entry_t
2457 destructors; /* pointer to function to invoke deconstructors of
2458 firstprivate C++ objects */
2459 /* future data */
2460} kmp_cmplrdata_t;
2461
2462/* sizeof_kmp_task_t passed as arg to kmpc_omp_task call */
2465typedef struct kmp_task { /* GEH: Shouldn't this be aligned somehow? */
2466 void *shareds;
2467 kmp_routine_entry_t
2468 routine;
2469 kmp_int32 part_id;
2470 kmp_cmplrdata_t
2471 data1; /* Two known optional additions: destructors and priority */
2472 kmp_cmplrdata_t data2; /* Process destructors first, priority second */
2473 /* future data */
2474 /* private vars */
2475} kmp_task_t;
2476
2481typedef struct kmp_taskgroup {
2482 std::atomic<kmp_int32> count; // number of allocated and incomplete tasks
2483 std::atomic<kmp_int32>
2484 cancel_request; // request for cancellation of this taskgroup
2485 struct kmp_taskgroup *parent; // parent taskgroup
2486 // Block of data to perform task reduction
2487 void *reduce_data; // reduction related info
2488 kmp_int32 reduce_num_data; // number of data items to reduce
2489 uintptr_t *gomp_data; // gomp reduction data
2490} kmp_taskgroup_t;
2491
2492// forward declarations
2493typedef union kmp_depnode kmp_depnode_t;
2494typedef struct kmp_depnode_list kmp_depnode_list_t;
2495typedef struct kmp_dephash_entry kmp_dephash_entry_t;
2496
2497// macros for checking dep flag as an integer
2498#define KMP_DEP_IN 0x1
2499#define KMP_DEP_OUT 0x2
2500#define KMP_DEP_INOUT 0x3
2501#define KMP_DEP_MTX 0x4
2502#define KMP_DEP_SET 0x8
2503#define KMP_DEP_ALL 0x80
2504// Compiler sends us this info. Note: some test cases contain an explicit copy
2505// of this struct and should be in sync with any changes here.
2506typedef struct kmp_depend_info {
2507 kmp_intptr_t base_addr;
2508 size_t len;
2509 union {
2510 kmp_uint8 flag; // flag as an unsigned char
2511 struct { // flag as a set of 8 bits
2512#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
2513 /* Same fields as in the #else branch, but in reverse order */
2514 unsigned all : 1;
2515 unsigned unused : 3;
2516 unsigned set : 1;
2517 unsigned mtx : 1;
2518 unsigned out : 1;
2519 unsigned in : 1;
2520#else
2521 unsigned in : 1;
2522 unsigned out : 1;
2523 unsigned mtx : 1;
2524 unsigned set : 1;
2525 unsigned unused : 3;
2526 unsigned all : 1;
2527#endif
2528 } flags;
2529 };
2530} kmp_depend_info_t;
2531
2532// Internal structures to work with task dependencies:
2533struct kmp_depnode_list {
2534 kmp_depnode_t *node;
2535 kmp_depnode_list_t *next;
2536};
2537
2538// Max number of mutexinoutset dependencies per node
2539#define MAX_MTX_DEPS 4
2540
2541typedef struct kmp_base_depnode {
2542 kmp_depnode_list_t *successors; /* used under lock */
2543 kmp_task_t *task; /* non-NULL if depnode is active, used under lock */
2544 kmp_lock_t *mtx_locks[MAX_MTX_DEPS]; /* lock mutexinoutset dependent tasks */
2545 kmp_int32 mtx_num_locks; /* number of locks in mtx_locks array */
2546 kmp_lock_t lock; /* guards shared fields: task, successors */
2547#if KMP_SUPPORT_GRAPH_OUTPUT
2548 kmp_uint32 id;
2549#endif
2550 std::atomic<kmp_int32> npredecessors;
2551 std::atomic<kmp_int32> nrefs;
2552} kmp_base_depnode_t;
2553
2554union KMP_ALIGN_CACHE kmp_depnode {
2555 double dn_align; /* use worst case alignment */
2556 char dn_pad[KMP_PAD(kmp_base_depnode_t, CACHE_LINE)];
2557 kmp_base_depnode_t dn;
2558};
2559
2560struct kmp_dephash_entry {
2561 kmp_intptr_t addr;
2562 kmp_depnode_t *last_out;
2563 kmp_depnode_list_t *last_set;
2564 kmp_depnode_list_t *prev_set;
2565 kmp_uint8 last_flag;
2566 kmp_lock_t *mtx_lock; /* is referenced by depnodes w/mutexinoutset dep */
2567 kmp_dephash_entry_t *next_in_bucket;
2568};
2569
2570typedef struct kmp_dephash {
2571 kmp_dephash_entry_t **buckets;
2572 size_t size;
2573 kmp_depnode_t *last_all;
2574 size_t generation;
2575 kmp_uint32 nelements;
2576 kmp_uint32 nconflicts;
2577} kmp_dephash_t;
2578
2579typedef struct kmp_task_affinity_info {
2580 kmp_intptr_t base_addr;
2581 size_t len;
2582 struct {
2583 bool flag1 : 1;
2584 bool flag2 : 1;
2585 kmp_int32 reserved : 30;
2586 } flags;
2587} kmp_task_affinity_info_t;
2588
2589typedef enum kmp_event_type_t {
2590 KMP_EVENT_UNINITIALIZED = 0,
2591 KMP_EVENT_ALLOW_COMPLETION = 1
2592} kmp_event_type_t;
2593
2594typedef struct {
2595 kmp_event_type_t type;
2596 kmp_tas_lock_t lock;
2597 union {
2598 kmp_task_t *task;
2599 } ed;
2600} kmp_event_t;
2601
2602#if OMPX_TASKGRAPH
2603// Initial number of allocated nodes while recording
2604#define INIT_MAPSIZE 50
2605
2606typedef struct kmp_taskgraph_flags { /*This needs to be exactly 32 bits */
2607 unsigned nowait : 1;
2608 unsigned re_record : 1;
2609 unsigned reserved : 30;
2610} kmp_taskgraph_flags_t;
2611
2613typedef struct kmp_node_info {
2614 kmp_task_t *task; // Pointer to the actual task
2615 kmp_int32 *successors; // Array of the succesors ids
2616 kmp_int32 nsuccessors; // Number of succesors of the node
2617 std::atomic<kmp_int32>
2618 npredecessors_counter; // Number of predessors on the fly
2619 kmp_int32 npredecessors; // Total number of predecessors
2620 kmp_int32 successors_size; // Number of allocated succesors ids
2621 kmp_taskdata_t *parent_task; // Parent implicit task
2622} kmp_node_info_t;
2623
2625typedef enum kmp_tdg_status {
2626 KMP_TDG_NONE = 0,
2627 KMP_TDG_RECORDING = 1,
2628 KMP_TDG_READY = 2
2629} kmp_tdg_status_t;
2630
2632typedef struct kmp_tdg_info {
2633 kmp_int32 tdg_id; // Unique idenfifier of the TDG
2634 kmp_taskgraph_flags_t tdg_flags; // Flags related to a TDG
2635 kmp_int32 map_size; // Number of allocated TDG nodes
2636 kmp_int32 num_roots; // Number of roots tasks int the TDG
2637 kmp_int32 *root_tasks; // Array of tasks identifiers that are roots
2638 kmp_node_info_t *record_map; // Array of TDG nodes
2639 kmp_tdg_status_t tdg_status =
2640 KMP_TDG_NONE; // Status of the TDG (recording, ready...)
2641 std::atomic<kmp_int32> num_tasks; // Number of TDG nodes
2642 kmp_bootstrap_lock_t
2643 graph_lock; // Protect graph attributes when updated via taskloop_recur
2644 // Taskloop reduction related
2645 void *rec_taskred_data; // Data to pass to __kmpc_task_reduction_init or
2646 // __kmpc_taskred_init
2647 kmp_int32 rec_num_taskred;
2648} kmp_tdg_info_t;
2649
2650extern int __kmp_tdg_dot;
2651extern kmp_int32 __kmp_max_tdgs;
2652extern kmp_tdg_info_t **__kmp_global_tdgs;
2653extern kmp_int32 __kmp_curr_tdg_idx;
2654extern kmp_int32 __kmp_successors_size;
2655extern std::atomic<kmp_int32> __kmp_tdg_task_id;
2656extern kmp_int32 __kmp_num_tdg;
2657#endif
2658
2659#ifdef BUILD_TIED_TASK_STACK
2660
2661/* Tied Task stack definitions */
2662typedef struct kmp_stack_block {
2663 kmp_taskdata_t *sb_block[TASK_STACK_BLOCK_SIZE];
2664 struct kmp_stack_block *sb_next;
2665 struct kmp_stack_block *sb_prev;
2666} kmp_stack_block_t;
2667
2668typedef struct kmp_task_stack {
2669 kmp_stack_block_t ts_first_block; // first block of stack entries
2670 kmp_taskdata_t **ts_top; // pointer to the top of stack
2671 kmp_int32 ts_entries; // number of entries on the stack
2672} kmp_task_stack_t;
2673
2674#endif // BUILD_TIED_TASK_STACK
2675
2676typedef struct kmp_tasking_flags { /* Total struct must be exactly 32 bits */
2677#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
2678 /* Same fields as in the #else branch, but in reverse order */
2679#if OMPX_TASKGRAPH
2680 unsigned reserved31 : 5;
2681 unsigned onced : 1;
2682#else
2683 unsigned reserved31 : 6;
2684#endif
2685 unsigned target : 1;
2686 unsigned native : 1;
2687 unsigned freed : 1;
2688 unsigned complete : 1;
2689 unsigned executing : 1;
2690 unsigned started : 1;
2691 unsigned team_serial : 1;
2692 unsigned tasking_ser : 1;
2693 unsigned task_serial : 1;
2694 unsigned tasktype : 1;
2695 unsigned reserved : 8;
2696 unsigned hidden_helper : 1;
2697 unsigned detachable : 1;
2698 unsigned priority_specified : 1;
2699 unsigned proxy : 1;
2700 unsigned destructors_thunk : 1;
2701 unsigned merged_if0 : 1;
2702 unsigned final : 1;
2703 unsigned tiedness : 1;
2704#else
2705 /* Compiler flags */ /* Total compiler flags must be 16 bits */
2706 unsigned tiedness : 1; /* task is either tied (1) or untied (0) */
2707 unsigned final : 1; /* task is final(1) so execute immediately */
2708 unsigned merged_if0 : 1; /* no __kmpc_task_{begin/complete}_if0 calls in if0
2709 code path */
2710 unsigned destructors_thunk : 1; /* set if the compiler creates a thunk to
2711 invoke destructors from the runtime */
2712 unsigned proxy : 1; /* task is a proxy task (it will be executed outside the
2713 context of the RTL) */
2714 unsigned priority_specified : 1; /* set if the compiler provides priority
2715 setting for the task */
2716 unsigned detachable : 1; /* 1 == can detach */
2717 unsigned hidden_helper : 1; /* 1 == hidden helper task */
2718 unsigned reserved : 8; /* reserved for compiler use */
2719
2720 /* Library flags */ /* Total library flags must be 16 bits */
2721 unsigned tasktype : 1; /* task is either explicit(1) or implicit (0) */
2722 unsigned task_serial : 1; // task is executed immediately (1) or deferred (0)
2723 unsigned tasking_ser : 1; // all tasks in team are either executed immediately
2724 // (1) or may be deferred (0)
2725 unsigned team_serial : 1; // entire team is serial (1) [1 thread] or parallel
2726 // (0) [>= 2 threads]
2727 /* If either team_serial or tasking_ser is set, task team may be NULL */
2728 /* Task State Flags: */
2729 unsigned started : 1; /* 1==started, 0==not started */
2730 unsigned executing : 1; /* 1==executing, 0==not executing */
2731 unsigned complete : 1; /* 1==complete, 0==not complete */
2732 unsigned freed : 1; /* 1==freed, 0==allocated */
2733 unsigned native : 1; /* 1==gcc-compiled task, 0==intel */
2734 unsigned target : 1;
2735#if OMPX_TASKGRAPH
2736 unsigned onced : 1; /* 1==ran once already, 0==never ran, record & replay purposes */
2737 unsigned reserved31 : 5; /* reserved for library use */
2738#else
2739 unsigned reserved31 : 6; /* reserved for library use */
2740#endif
2741#endif
2742} kmp_tasking_flags_t;
2743
2744typedef struct kmp_target_data {
2745 void *async_handle; // libomptarget async handle for task completion query
2746} kmp_target_data_t;
2747
2748struct kmp_taskdata { /* aligned during dynamic allocation */
2749 kmp_int32 td_task_id; /* id, assigned by debugger */
2750 kmp_tasking_flags_t td_flags; /* task flags */
2751 kmp_team_t *td_team; /* team for this task */
2752 kmp_info_p *td_alloc_thread; /* thread that allocated data structures */
2753 /* Currently not used except for perhaps IDB */
2754 kmp_taskdata_t *td_parent; /* parent task */
2755 kmp_int32 td_level; /* task nesting level */
2756 std::atomic<kmp_int32> td_untied_count; // untied task active parts counter
2757 ident_t *td_ident; /* task identifier */
2758 // Taskwait data.
2759 ident_t *td_taskwait_ident;
2760 kmp_uint32 td_taskwait_counter;
2761 kmp_int32 td_taskwait_thread; /* gtid + 1 of thread encountered taskwait */
2762 KMP_ALIGN_CACHE kmp_internal_control_t
2763 td_icvs; /* Internal control variables for the task */
2764 KMP_ALIGN_CACHE std::atomic<kmp_int32>
2765 td_allocated_child_tasks; /* Child tasks (+ current task) not yet
2766 deallocated */
2767 std::atomic<kmp_int32>
2768 td_incomplete_child_tasks; /* Child tasks not yet complete */
2769 kmp_taskgroup_t
2770 *td_taskgroup; // Each task keeps pointer to its current taskgroup
2771 kmp_dephash_t
2772 *td_dephash; // Dependencies for children tasks are tracked from here
2773 kmp_depnode_t
2774 *td_depnode; // Pointer to graph node if this task has dependencies
2775 kmp_task_team_t *td_task_team;
2776 size_t td_size_alloc; // Size of task structure, including shareds etc.
2777#if defined(KMP_GOMP_COMPAT)
2778 // 4 or 8 byte integers for the loop bounds in GOMP_taskloop
2779 kmp_int32 td_size_loop_bounds;
2780#endif
2781 kmp_taskdata_t *td_last_tied; // keep tied task for task scheduling constraint
2782#if defined(KMP_GOMP_COMPAT)
2783 // GOMP sends in a copy function for copy constructors
2784 void (*td_copy_func)(void *, void *);
2785#endif
2786 kmp_event_t td_allow_completion_event;
2787#if OMPT_SUPPORT
2788 ompt_task_info_t ompt_task_info;
2789#endif
2790#if OMPX_TASKGRAPH
2791 bool is_taskgraph = 0; // whether the task is within a TDG
2792 kmp_tdg_info_t *tdg; // used to associate task with a TDG
2793 kmp_int32 td_tdg_task_id; // local task id in its TDG
2794#endif
2795 kmp_target_data_t td_target_data;
2796}; // struct kmp_taskdata
2797
2798// Make sure padding above worked
2799KMP_BUILD_ASSERT(sizeof(kmp_taskdata_t) % sizeof(void *) == 0);
2800
2801// Data for task team but per thread
2802typedef struct kmp_base_thread_data {
2803 kmp_info_p *td_thr; // Pointer back to thread info
2804 // Used only in __kmp_execute_tasks_template, maybe not avail until task is
2805 // queued?
2806 kmp_bootstrap_lock_t td_deque_lock; // Lock for accessing deque
2807 kmp_taskdata_t *
2808 *td_deque; // Deque of tasks encountered by td_thr, dynamically allocated
2809 kmp_int32 td_deque_size; // Size of deck
2810 kmp_uint32 td_deque_head; // Head of deque (will wrap)
2811 kmp_uint32 td_deque_tail; // Tail of deque (will wrap)
2812 kmp_int32 td_deque_ntasks; // Number of tasks in deque
2813 // GEH: shouldn't this be volatile since used in while-spin?
2814 kmp_int32 td_deque_last_stolen; // Thread number of last successful steal
2815#ifdef BUILD_TIED_TASK_STACK
2816 kmp_task_stack_t td_susp_tied_tasks; // Stack of suspended tied tasks for task
2817// scheduling constraint
2818#endif // BUILD_TIED_TASK_STACK
2819} kmp_base_thread_data_t;
2820
2821#define TASK_DEQUE_BITS 8 // Used solely to define INITIAL_TASK_DEQUE_SIZE
2822#define INITIAL_TASK_DEQUE_SIZE (1 << TASK_DEQUE_BITS)
2823
2824#define TASK_DEQUE_SIZE(td) ((td).td_deque_size)
2825#define TASK_DEQUE_MASK(td) ((td).td_deque_size - 1)
2826
2827typedef union KMP_ALIGN_CACHE kmp_thread_data {
2828 kmp_base_thread_data_t td;
2829 double td_align; /* use worst case alignment */
2830 char td_pad[KMP_PAD(kmp_base_thread_data_t, CACHE_LINE)];
2831} kmp_thread_data_t;
2832
2833typedef struct kmp_task_pri {
2834 kmp_thread_data_t td;
2835 kmp_int32 priority;
2836 kmp_task_pri *next;
2837} kmp_task_pri_t;
2838
2839// Data for task teams which are used when tasking is enabled for the team
2840typedef struct kmp_base_task_team {
2841 kmp_bootstrap_lock_t
2842 tt_threads_lock; /* Lock used to allocate per-thread part of task team */
2843 /* must be bootstrap lock since used at library shutdown*/
2844
2845 // TODO: check performance vs kmp_tas_lock_t
2846 kmp_bootstrap_lock_t tt_task_pri_lock; /* Lock to access priority tasks */
2847 kmp_task_pri_t *tt_task_pri_list;
2848
2849 kmp_task_team_t *tt_next; /* For linking the task team free list */
2850 kmp_thread_data_t
2851 *tt_threads_data; /* Array of per-thread structures for task team */
2852 /* Data survives task team deallocation */
2853 kmp_int32 tt_found_tasks; /* Have we found tasks and queued them while
2854 executing this team? */
2855 /* TRUE means tt_threads_data is set up and initialized */
2856 kmp_int32 tt_nproc; /* #threads in team */
2857 kmp_int32 tt_max_threads; // # entries allocated for threads_data array
2858 kmp_int32 tt_found_proxy_tasks; // found proxy tasks since last barrier
2859 kmp_int32 tt_untied_task_encountered;
2860 std::atomic<kmp_int32> tt_num_task_pri; // number of priority tasks enqueued
2861 // There is hidden helper thread encountered in this task team so that we must
2862 // wait when waiting on task team
2863 kmp_int32 tt_hidden_helper_task_encountered;
2864
2865 KMP_ALIGN_CACHE
2866 std::atomic<kmp_int32> tt_unfinished_threads; /* #threads still active */
2867
2868 KMP_ALIGN_CACHE
2869 volatile kmp_uint32
2870 tt_active; /* is the team still actively executing tasks */
2871} kmp_base_task_team_t;
2872
2873union KMP_ALIGN_CACHE kmp_task_team {
2874 kmp_base_task_team_t tt;
2875 double tt_align; /* use worst case alignment */
2876 char tt_pad[KMP_PAD(kmp_base_task_team_t, CACHE_LINE)];
2877};
2878
2879typedef struct kmp_task_team_list_t {
2880 kmp_task_team_t *task_team;
2881 kmp_task_team_list_t *next;
2882} kmp_task_team_list_t;
2883
2884#if (USE_FAST_MEMORY == 3) || (USE_FAST_MEMORY == 5)
2885// Free lists keep same-size free memory slots for fast memory allocation
2886// routines
2887typedef struct kmp_free_list {
2888 void *th_free_list_self; // Self-allocated tasks free list
2889 void *th_free_list_sync; // Self-allocated tasks stolen/returned by other
2890 // threads
2891 void *th_free_list_other; // Non-self free list (to be returned to owner's
2892 // sync list)
2893} kmp_free_list_t;
2894#endif
2895#if KMP_NESTED_HOT_TEAMS
2896// Hot teams array keeps hot teams and their sizes for given thread. Hot teams
2897// are not put in teams pool, and they don't put threads in threads pool.
2898typedef struct kmp_hot_team_ptr {
2899 kmp_team_p *hot_team; // pointer to hot_team of given nesting level
2900 kmp_int32 hot_team_nth; // number of threads allocated for the hot_team
2901} kmp_hot_team_ptr_t;
2902#endif
2903typedef struct kmp_teams_size {
2904 kmp_int32 nteams; // number of teams in a league
2905 kmp_int32 nth; // number of threads in each team of the league
2906} kmp_teams_size_t;
2907
2908// This struct stores a thread that acts as a "root" for a contention
2909// group. Contention groups are rooted at kmp_root threads, but also at
2910// each primary thread of each team created in the teams construct.
2911// This struct therefore also stores a thread_limit associated with
2912// that contention group, and a counter to track the number of threads
2913// active in that contention group. Each thread has a list of these: CG
2914// root threads have an entry in their list in which cg_root refers to
2915// the thread itself, whereas other workers in the CG will have a
2916// single entry where cg_root is same as the entry containing their CG
2917// root. When a thread encounters a teams construct, it will add a new
2918// entry to the front of its list, because it now roots a new CG.
2919typedef struct kmp_cg_root {
2920 kmp_info_p *cg_root; // "root" thread for a contention group
2921 // The CG root's limit comes from OMP_THREAD_LIMIT for root threads, or
2922 // thread_limit clause for teams primary threads
2923 kmp_int32 cg_thread_limit;
2924 kmp_int32 cg_nthreads; // Count of active threads in CG rooted at cg_root
2925 struct kmp_cg_root *up; // pointer to higher level CG root in list
2926} kmp_cg_root_t;
2927
2928// OpenMP thread data structures
2929
2930typedef struct KMP_ALIGN_CACHE kmp_base_info {
2931 /* Start with the readonly data which is cache aligned and padded. This is
2932 written before the thread starts working by the primary thread. Uber
2933 masters may update themselves later. Usage does not consider serialized
2934 regions. */
2935 kmp_desc_t th_info;
2936 kmp_team_p *th_team; /* team we belong to */
2937 kmp_root_p *th_root; /* pointer to root of task hierarchy */
2938 kmp_info_p *th_next_pool; /* next available thread in the pool */
2939 kmp_disp_t *th_dispatch; /* thread's dispatch data */
2940 int th_in_pool; /* in thread pool (32 bits for TCR/TCW) */
2941
2942 /* The following are cached from the team info structure */
2943 /* TODO use these in more places as determined to be needed via profiling */
2944 int th_team_nproc; /* number of threads in a team */
2945 kmp_info_p *th_team_master; /* the team's primary thread */
2946 int th_team_serialized; /* team is serialized */
2947 microtask_t th_teams_microtask; /* save entry address for teams construct */
2948 int th_teams_level; /* save initial level of teams construct */
2949/* it is 0 on device but may be any on host */
2950
2951/* The blocktime info is copied from the team struct to the thread struct */
2952/* at the start of a barrier, and the values stored in the team are used */
2953/* at points in the code where the team struct is no longer guaranteed */
2954/* to exist (from the POV of worker threads). */
2955#if KMP_USE_MONITOR
2956 int th_team_bt_intervals;
2957 int th_team_bt_set;
2958#else
2959 kmp_uint64 th_team_bt_intervals;
2960#endif
2961
2962#if KMP_AFFINITY_SUPPORTED
2963 kmp_affin_mask_t *th_affin_mask; /* thread's current affinity mask */
2964 kmp_affinity_ids_t th_topology_ids; /* thread's current topology ids */
2965 kmp_affinity_attrs_t th_topology_attrs; /* thread's current topology attrs */
2966#endif
2967 omp_allocator_handle_t th_def_allocator; /* default allocator */
2968 /* The data set by the primary thread at reinit, then R/W by the worker */
2969 KMP_ALIGN_CACHE int
2970 th_set_nproc; /* if > 0, then only use this request for the next fork */
2971 int *th_set_nested_nth;
2972 bool th_nt_strict; // num_threads clause has strict modifier
2973 ident_t *th_nt_loc; // loc for strict modifier
2974 int th_nt_sev; // error severity for strict modifier
2975 const char *th_nt_msg; // error message for strict modifier
2976 int th_set_nested_nth_sz;
2977#if KMP_NESTED_HOT_TEAMS
2978 kmp_hot_team_ptr_t *th_hot_teams; /* array of hot teams */
2979#endif
2980 kmp_proc_bind_t
2981 th_set_proc_bind; /* if != proc_bind_default, use request for next fork */
2982 kmp_teams_size_t
2983 th_teams_size; /* number of teams/threads in teams construct */
2984#if KMP_AFFINITY_SUPPORTED
2985 int th_current_place; /* place currently bound to */
2986 int th_new_place; /* place to bind to in par reg */
2987 int th_first_place; /* first place in partition */
2988 int th_last_place; /* last place in partition */
2989#endif
2990 int th_prev_level; /* previous level for affinity format */
2991 int th_prev_num_threads; /* previous num_threads for affinity format */
2992#if USE_ITT_BUILD
2993 kmp_uint64 th_bar_arrive_time; /* arrival to barrier timestamp */
2994 kmp_uint64 th_bar_min_time; /* minimum arrival time at the barrier */
2995 kmp_uint64 th_frame_time; /* frame timestamp */
2996#endif /* USE_ITT_BUILD */
2997 kmp_local_t th_local;
2998 struct private_common *th_pri_head;
2999
3000 /* Now the data only used by the worker (after initial allocation) */
3001 /* TODO the first serial team should actually be stored in the info_t
3002 structure. this will help reduce initial allocation overhead */
3003 KMP_ALIGN_CACHE kmp_team_p
3004 *th_serial_team; /*serialized team held in reserve*/
3005
3006#if OMPT_SUPPORT
3007 ompt_thread_info_t ompt_thread_info;
3008#endif
3009
3010 /* The following are also read by the primary thread during reinit */
3011 struct common_table *th_pri_common;
3012
3013 volatile kmp_uint32 th_spin_here; /* thread-local location for spinning */
3014 /* while awaiting queuing lock acquire */
3015
3016 volatile void *th_sleep_loc; // this points at a kmp_flag<T>
3017 flag_type th_sleep_loc_type; // enum type of flag stored in th_sleep_loc
3018
3019 ident_t *th_ident;
3020 unsigned th_x; // Random number generator data
3021 unsigned th_a; // Random number generator data
3022
3023 /* Tasking-related data for the thread */
3024 kmp_task_team_t *th_task_team; // Task team struct
3025 kmp_taskdata_t *th_current_task; // Innermost Task being executed
3026 kmp_uint8 th_task_state; // alternating 0/1 for task team identification
3027 kmp_uint32 th_reap_state; // Non-zero indicates thread is not
3028 // tasking, thus safe to reap
3029
3030 /* More stuff for keeping track of active/sleeping threads (this part is
3031 written by the worker thread) */
3032 kmp_uint8 th_active_in_pool; // included in count of #active threads in pool
3033 int th_active; // ! sleeping; 32 bits for TCR/TCW
3034 std::atomic<kmp_uint32> th_used_in_team; // Flag indicating use in team
3035 // 0 = not used in team; 1 = used in team;
3036 // 2 = transitioning to not used in team; 3 = transitioning to used in team
3037 struct cons_header *th_cons; // used for consistency check
3038#if KMP_USE_HIER_SCHED
3039 // used for hierarchical scheduling
3040 kmp_hier_private_bdata_t *th_hier_bar_data;
3041#endif
3042
3043 /* Add the syncronizing data which is cache aligned and padded. */
3044 KMP_ALIGN_CACHE kmp_balign_t th_bar[bs_last_barrier];
3045
3046 KMP_ALIGN_CACHE volatile kmp_int32
3047 th_next_waiting; /* gtid+1 of next thread on lock wait queue, 0 if none */
3048
3049#if (USE_FAST_MEMORY == 3) || (USE_FAST_MEMORY == 5)
3050#define NUM_LISTS 4
3051 kmp_free_list_t th_free_lists[NUM_LISTS]; // Free lists for fast memory
3052// allocation routines
3053#endif
3054
3055#if KMP_OS_WINDOWS
3056 kmp_win32_cond_t th_suspend_cv;
3057 kmp_win32_mutex_t th_suspend_mx;
3058 std::atomic<int> th_suspend_init;
3059#endif
3060#if KMP_OS_UNIX
3061 kmp_cond_align_t th_suspend_cv;
3062 kmp_mutex_align_t th_suspend_mx;
3063 std::atomic<int> th_suspend_init_count;
3064#endif
3065
3066#if USE_ITT_BUILD
3067 kmp_itt_mark_t th_itt_mark_single;
3068// alignment ???
3069#endif /* USE_ITT_BUILD */
3070#if KMP_STATS_ENABLED
3071 kmp_stats_list *th_stats;
3072#endif
3073#if KMP_OS_UNIX
3074 std::atomic<bool> th_blocking;
3075#endif
3076 kmp_cg_root_t *th_cg_roots; // list of cg_roots associated with this thread
3077} kmp_base_info_t;
3078
3079typedef union KMP_ALIGN_CACHE kmp_info {
3080 double th_align; /* use worst case alignment */
3081 char th_pad[KMP_PAD(kmp_base_info_t, CACHE_LINE)];
3082 kmp_base_info_t th;
3083} kmp_info_t;
3084
3085// OpenMP thread team data structures
3086
3087typedef struct kmp_base_data {
3088 volatile kmp_uint32 t_value;
3089} kmp_base_data_t;
3090
3091typedef union KMP_ALIGN_CACHE kmp_sleep_team {
3092 double dt_align; /* use worst case alignment */
3093 char dt_pad[KMP_PAD(kmp_base_data_t, CACHE_LINE)];
3094 kmp_base_data_t dt;
3095} kmp_sleep_team_t;
3096
3097typedef union KMP_ALIGN_CACHE kmp_ordered_team {
3098 double dt_align; /* use worst case alignment */
3099 char dt_pad[KMP_PAD(kmp_base_data_t, CACHE_LINE)];
3100 kmp_base_data_t dt;
3101} kmp_ordered_team_t;
3102
3103typedef int (*launch_t)(int gtid);
3104
3105/* Minimum number of ARGV entries to malloc if necessary */
3106#define KMP_MIN_MALLOC_ARGV_ENTRIES 100
3107
3108// Set up how many argv pointers will fit in cache lines containing
3109// t_inline_argv. Historically, we have supported at least 96 bytes. Using a
3110// larger value for more space between the primary write/worker read section and
3111// read/write by all section seems to buy more performance on EPCC PARALLEL.
3112#if KMP_ARCH_X86 || KMP_ARCH_X86_64
3113#define KMP_INLINE_ARGV_BYTES \
3114 (4 * CACHE_LINE - \
3115 ((3 * KMP_PTR_SKIP + 2 * sizeof(int) + 2 * sizeof(kmp_int8) + \
3116 sizeof(kmp_int16) + sizeof(kmp_uint32)) % \
3117 CACHE_LINE))
3118#else
3119#define KMP_INLINE_ARGV_BYTES \
3120 (2 * CACHE_LINE - ((3 * KMP_PTR_SKIP + 2 * sizeof(int)) % CACHE_LINE))
3121#endif
3122#define KMP_INLINE_ARGV_ENTRIES (int)(KMP_INLINE_ARGV_BYTES / KMP_PTR_SKIP)
3123
3124typedef struct KMP_ALIGN_CACHE kmp_base_team {
3125 // Synchronization Data
3126 // ---------------------------------------------------------------------------
3127 KMP_ALIGN_CACHE kmp_ordered_team_t t_ordered;
3128 kmp_balign_team_t t_bar[bs_last_barrier];
3129 std::atomic<int> t_construct; // count of single directive encountered by team
3130 char pad[sizeof(kmp_lock_t)]; // padding to maintain performance on big iron
3131
3132 // [0] - parallel / [1] - worksharing task reduction data shared by taskgroups
3133 std::atomic<void *> t_tg_reduce_data[2]; // to support task modifier
3134 std::atomic<int> t_tg_fini_counter[2]; // sync end of task reductions
3135
3136 // Primary thread only
3137 // ---------------------------------------------------------------------------
3138 KMP_ALIGN_CACHE int t_master_tid; // tid of primary thread in parent team
3139 int t_master_this_cons; // "this_construct" single counter of primary thread
3140 // in parent team
3141 ident_t *t_ident; // if volatile, have to change too much other crud to
3142 // volatile too
3143 kmp_team_p *t_parent; // parent team
3144 kmp_team_p *t_next_pool; // next free team in the team pool
3145 kmp_disp_t *t_dispatch; // thread's dispatch data
3146 kmp_task_team_t *t_task_team[2]; // Task team struct; switch between 2
3147 kmp_proc_bind_t t_proc_bind; // bind type for par region
3148 int t_primary_task_state; // primary thread's task state saved
3149#if USE_ITT_BUILD
3150 kmp_uint64 t_region_time; // region begin timestamp
3151#endif /* USE_ITT_BUILD */
3152
3153 // Primary thread write, workers read
3154 // --------------------------------------------------------------------------
3155 KMP_ALIGN_CACHE void **t_argv;
3156 int t_argc;
3157 int t_nproc; // number of threads in team
3158 microtask_t t_pkfn;
3159 launch_t t_invoke; // procedure to launch the microtask
3160
3161#if OMPT_SUPPORT
3162 ompt_team_info_t ompt_team_info;
3163 ompt_lw_taskteam_t *ompt_serialized_team_info;
3164#endif
3165
3166#if KMP_ARCH_X86 || KMP_ARCH_X86_64
3167 kmp_int8 t_fp_control_saved;
3168 kmp_int8 t_pad2b;
3169 kmp_int16 t_x87_fpu_control_word; // FP control regs
3170 kmp_uint32 t_mxcsr;
3171#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */
3172
3173 void *t_inline_argv[KMP_INLINE_ARGV_ENTRIES];
3174
3175 KMP_ALIGN_CACHE kmp_info_t **t_threads;
3176 kmp_taskdata_t
3177 *t_implicit_task_taskdata; // Taskdata for the thread's implicit task
3178 int t_level; // nested parallel level
3179
3180 KMP_ALIGN_CACHE int t_max_argc;
3181 int t_max_nproc; // max threads this team can handle (dynamically expandable)
3182 int t_serialized; // levels deep of serialized teams
3183 dispatch_shared_info_t *t_disp_buffer; // buffers for dispatch system
3184 int t_id; // team's id, assigned by debugger.
3185 int t_active_level; // nested active parallel level
3186 kmp_r_sched_t t_sched; // run-time schedule for the team
3187#if KMP_AFFINITY_SUPPORTED
3188 int t_first_place; // first & last place in parent thread's partition.
3189 int t_last_place; // Restore these values to primary thread after par region.
3190#endif // KMP_AFFINITY_SUPPORTED
3191 int t_display_affinity;
3192 int t_size_changed; // team size was changed?: 0: no, 1: yes, -1: changed via
3193 // omp_set_num_threads() call
3194 omp_allocator_handle_t t_def_allocator; /* default allocator */
3195
3196// Read/write by workers as well
3197#if (KMP_ARCH_X86 || KMP_ARCH_X86_64)
3198 // Using CACHE_LINE=64 reduces memory footprint, but causes a big perf
3199 // regression of epcc 'parallel' and 'barrier' on fxe256lin01. This extra
3200 // padding serves to fix the performance of epcc 'parallel' and 'barrier' when
3201 // CACHE_LINE=64. TODO: investigate more and get rid if this padding.
3202 char dummy_padding[1024];
3203#endif
3204 // Internal control stack for additional nested teams.
3205 KMP_ALIGN_CACHE kmp_internal_control_t *t_control_stack_top;
3206 // for SERIALIZED teams nested 2 or more levels deep
3207 // typed flag to store request state of cancellation
3208 std::atomic<kmp_int32> t_cancel_request;
3209 int t_master_active; // save on fork, restore on join
3210 void *t_copypriv_data; // team specific pointer to copyprivate data array
3211#if KMP_OS_WINDOWS
3212 std::atomic<kmp_uint32> t_copyin_counter;
3213#endif
3214#if USE_ITT_BUILD
3215 void *t_stack_id; // team specific stack stitching id (for ittnotify)
3216#endif /* USE_ITT_BUILD */
3217 distributedBarrier *b; // Distributed barrier data associated with team
3218 kmp_nested_nthreads_t *t_nested_nth;
3219} kmp_base_team_t;
3220
3221// Assert that the list structure fits and aligns within
3222// the double task team pointer
3223KMP_BUILD_ASSERT(sizeof(kmp_task_team_t *[2]) == sizeof(kmp_task_team_list_t));
3224KMP_BUILD_ASSERT(alignof(kmp_task_team_t *[2]) ==
3225 alignof(kmp_task_team_list_t));
3226
3227union KMP_ALIGN_CACHE kmp_team {
3228 kmp_base_team_t t;
3229 double t_align; /* use worst case alignment */
3230 char t_pad[KMP_PAD(kmp_base_team_t, CACHE_LINE)];
3231};
3232
3233typedef union KMP_ALIGN_CACHE kmp_time_global {
3234 double dt_align; /* use worst case alignment */
3235 char dt_pad[KMP_PAD(kmp_base_data_t, CACHE_LINE)];
3236 kmp_base_data_t dt;
3237} kmp_time_global_t;
3238
3239typedef struct kmp_base_global {
3240 /* cache-aligned */
3241 kmp_time_global_t g_time;
3242
3243 /* non cache-aligned */
3244 volatile int g_abort;
3245 volatile int g_done;
3246
3247 int g_dynamic;
3248 enum dynamic_mode g_dynamic_mode;
3249} kmp_base_global_t;
3250
3251typedef union KMP_ALIGN_CACHE kmp_global {
3252 kmp_base_global_t g;
3253 double g_align; /* use worst case alignment */
3254 char g_pad[KMP_PAD(kmp_base_global_t, CACHE_LINE)];
3255} kmp_global_t;
3256
3257typedef struct kmp_base_root {
3258 // TODO: GEH - combine r_active with r_in_parallel then r_active ==
3259 // (r_in_parallel>= 0)
3260 // TODO: GEH - then replace r_active with t_active_levels if we can to reduce
3261 // the synch overhead or keeping r_active
3262 volatile int r_active; /* TRUE if some region in a nest has > 1 thread */
3263 // keeps a count of active parallel regions per root
3264 std::atomic<int> r_in_parallel;
3265 // GEH: This is misnamed, should be r_active_levels
3266 kmp_team_t *r_root_team;
3267 kmp_team_t *r_hot_team;
3268 kmp_info_t *r_uber_thread;
3269 kmp_lock_t r_begin_lock;
3270 volatile int r_begin;
3271 int r_blocktime; /* blocktime for this root and descendants */
3272#if KMP_AFFINITY_SUPPORTED
3273 int r_affinity_assigned;
3274#endif // KMP_AFFINITY_SUPPORTED
3275} kmp_base_root_t;
3276
3277typedef union KMP_ALIGN_CACHE kmp_root {
3278 kmp_base_root_t r;
3279 double r_align; /* use worst case alignment */
3280 char r_pad[KMP_PAD(kmp_base_root_t, CACHE_LINE)];
3281} kmp_root_t;
3282
3283struct fortran_inx_info {
3284 kmp_int32 data;
3285};
3286
3287// This list type exists to hold old __kmp_threads arrays so that
3288// old references to them may complete while reallocation takes place when
3289// expanding the array. The items in this list are kept alive until library
3290// shutdown.
3291typedef struct kmp_old_threads_list_t {
3292 kmp_info_t **threads;
3293 struct kmp_old_threads_list_t *next;
3294} kmp_old_threads_list_t;
3295
3296/* ------------------------------------------------------------------------ */
3297
3298extern int __kmp_settings;
3299extern int __kmp_duplicate_library_ok;
3300#if USE_ITT_BUILD
3301extern int __kmp_forkjoin_frames;
3302extern int __kmp_forkjoin_frames_mode;
3303#endif
3304extern PACKED_REDUCTION_METHOD_T __kmp_force_reduction_method;
3305extern int __kmp_determ_red;
3306
3307#ifdef KMP_DEBUG
3308extern int kmp_a_debug;
3309extern int kmp_b_debug;
3310extern int kmp_c_debug;
3311extern int kmp_d_debug;
3312extern int kmp_e_debug;
3313extern int kmp_f_debug;
3314#endif /* KMP_DEBUG */
3315
3316/* For debug information logging using rotating buffer */
3317#define KMP_DEBUG_BUF_LINES_INIT 512
3318#define KMP_DEBUG_BUF_LINES_MIN 1
3319
3320#define KMP_DEBUG_BUF_CHARS_INIT 128
3321#define KMP_DEBUG_BUF_CHARS_MIN 2
3322
3323extern int
3324 __kmp_debug_buf; /* TRUE means use buffer, FALSE means print to stderr */
3325extern int __kmp_debug_buf_lines; /* How many lines of debug stored in buffer */
3326extern int
3327 __kmp_debug_buf_chars; /* How many characters allowed per line in buffer */
3328extern int __kmp_debug_buf_atomic; /* TRUE means use atomic update of buffer
3329 entry pointer */
3330
3331extern char *__kmp_debug_buffer; /* Debug buffer itself */
3332extern std::atomic<int> __kmp_debug_count; /* Counter for number of lines
3333 printed in buffer so far */
3334extern int __kmp_debug_buf_warn_chars; /* Keep track of char increase
3335 recommended in warnings */
3336/* end rotating debug buffer */
3337
3338#ifdef KMP_DEBUG
3339extern int __kmp_par_range; /* +1 => only go par for constructs in range */
3340
3341#define KMP_PAR_RANGE_ROUTINE_LEN 1024
3342extern char __kmp_par_range_routine[KMP_PAR_RANGE_ROUTINE_LEN];
3343#define KMP_PAR_RANGE_FILENAME_LEN 1024
3344extern char __kmp_par_range_filename[KMP_PAR_RANGE_FILENAME_LEN];
3345extern int __kmp_par_range_lb;
3346extern int __kmp_par_range_ub;
3347#endif
3348
3349/* For printing out dynamic storage map for threads and teams */
3350extern int
3351 __kmp_storage_map; /* True means print storage map for threads and teams */
3352extern int __kmp_storage_map_verbose; /* True means storage map includes
3353 placement info */
3354extern int __kmp_storage_map_verbose_specified;
3355
3356#if KMP_ARCH_X86 || KMP_ARCH_X86_64
3357extern kmp_cpuinfo_t __kmp_cpuinfo;
3358static inline bool __kmp_is_hybrid_cpu() { return __kmp_cpuinfo.flags.hybrid; }
3359#elif KMP_OS_DARWIN && KMP_ARCH_AARCH64
3360static inline bool __kmp_is_hybrid_cpu() { return true; }
3361#else
3362static inline bool __kmp_is_hybrid_cpu() { return false; }
3363#endif
3364
3365extern volatile int __kmp_init_serial;
3366extern volatile int __kmp_init_gtid;
3367extern volatile int __kmp_init_common;
3368extern volatile int __kmp_need_register_serial;
3369extern volatile int __kmp_init_middle;
3370extern volatile int __kmp_init_parallel;
3371#if KMP_USE_MONITOR
3372extern volatile int __kmp_init_monitor;
3373#endif
3374extern volatile int __kmp_init_user_locks;
3375extern volatile int __kmp_init_hidden_helper_threads;
3376extern int __kmp_init_counter;
3377extern int __kmp_root_counter;
3378extern int __kmp_version;
3379
3380/* list of address of allocated caches for commons */
3381extern kmp_cached_addr_t *__kmp_threadpriv_cache_list;
3382
3383/* Barrier algorithm types and options */
3384extern kmp_uint32 __kmp_barrier_gather_bb_dflt;
3385extern kmp_uint32 __kmp_barrier_release_bb_dflt;
3386extern kmp_bar_pat_e __kmp_barrier_gather_pat_dflt;
3387extern kmp_bar_pat_e __kmp_barrier_release_pat_dflt;
3388extern kmp_uint32 __kmp_barrier_gather_branch_bits[bs_last_barrier];
3389extern kmp_uint32 __kmp_barrier_release_branch_bits[bs_last_barrier];
3390extern kmp_bar_pat_e __kmp_barrier_gather_pattern[bs_last_barrier];
3391extern kmp_bar_pat_e __kmp_barrier_release_pattern[bs_last_barrier];
3392extern char const *__kmp_barrier_branch_bit_env_name[bs_last_barrier];
3393extern char const *__kmp_barrier_pattern_env_name[bs_last_barrier];
3394extern char const *__kmp_barrier_type_name[bs_last_barrier];
3395extern char const *__kmp_barrier_pattern_name[bp_last_bar];
3396
3397/* Global Locks */
3398extern kmp_bootstrap_lock_t __kmp_initz_lock; /* control initialization */
3399extern kmp_bootstrap_lock_t __kmp_forkjoin_lock; /* control fork/join access */
3400extern kmp_bootstrap_lock_t __kmp_task_team_lock;
3401extern kmp_bootstrap_lock_t
3402 __kmp_exit_lock; /* exit() is not always thread-safe */
3403#if KMP_USE_MONITOR
3404extern kmp_bootstrap_lock_t
3405 __kmp_monitor_lock; /* control monitor thread creation */
3406#endif
3407extern kmp_bootstrap_lock_t
3408 __kmp_tp_cached_lock; /* used for the hack to allow threadprivate cache and
3409 __kmp_threads expansion to co-exist */
3410
3411extern kmp_lock_t __kmp_global_lock; /* control OS/global access */
3412
3413extern enum library_type __kmp_library;
3414
3415extern enum sched_type __kmp_sched; /* default runtime scheduling */
3416extern enum sched_type __kmp_static; /* default static scheduling method */
3417extern enum sched_type __kmp_guided; /* default guided scheduling method */
3418extern enum sched_type __kmp_auto; /* default auto scheduling method */
3419extern int __kmp_chunk; /* default runtime chunk size */
3420extern int __kmp_force_monotonic; /* whether monotonic scheduling forced */
3421
3422extern size_t __kmp_stksize; /* stack size per thread */
3423#if KMP_USE_MONITOR
3424extern size_t __kmp_monitor_stksize; /* stack size for monitor thread */
3425#endif
3426extern size_t __kmp_stkoffset; /* stack offset per thread */
3427extern int __kmp_stkpadding; /* Should we pad root thread(s) stack */
3428
3429extern size_t
3430 __kmp_malloc_pool_incr; /* incremental size of pool for kmp_malloc() */
3431extern int __kmp_env_stksize; /* was KMP_STACKSIZE specified? */
3432extern int __kmp_env_blocktime; /* was KMP_BLOCKTIME specified? */
3433extern int __kmp_env_checks; /* was KMP_CHECKS specified? */
3434extern int __kmp_env_consistency_check; // was KMP_CONSISTENCY_CHECK specified?
3435extern int __kmp_generate_warnings; /* should we issue warnings? */
3436extern int __kmp_reserve_warn; /* have we issued reserve_threads warning? */
3437
3438#ifdef DEBUG_SUSPEND
3439extern int __kmp_suspend_count; /* count inside __kmp_suspend_template() */
3440#endif
3441
3442extern kmp_int32 __kmp_use_yield;
3443extern kmp_int32 __kmp_use_yield_exp_set;
3444extern kmp_uint32 __kmp_yield_init;
3445extern kmp_uint32 __kmp_yield_next;
3446extern kmp_uint64 __kmp_pause_init;
3447
3448/* ------------------------------------------------------------------------- */
3449extern int __kmp_allThreadsSpecified;
3450
3451extern size_t __kmp_align_alloc;
3452/* following data protected by initialization routines */
3453extern int __kmp_xproc; /* number of processors in the system */
3454extern int __kmp_avail_proc; /* number of processors available to the process */
3455extern size_t __kmp_sys_min_stksize; /* system-defined minimum stack size */
3456extern int __kmp_sys_max_nth; /* system-imposed maximum number of threads */
3457// maximum total number of concurrently-existing threads on device
3458extern int __kmp_max_nth;
3459// maximum total number of concurrently-existing threads in a contention group
3460extern int __kmp_cg_max_nth;
3461extern int __kmp_task_max_nth; // max threads used in a task
3462extern int __kmp_teams_max_nth; // max threads used in a teams construct
3463extern int __kmp_threads_capacity; /* capacity of the arrays __kmp_threads and
3464 __kmp_root */
3465extern int __kmp_dflt_team_nth; /* default number of threads in a parallel
3466 region a la OMP_NUM_THREADS */
3467extern int __kmp_dflt_team_nth_ub; /* upper bound on "" determined at serial
3468 initialization */
3469extern int __kmp_tp_capacity; /* capacity of __kmp_threads if threadprivate is
3470 used (fixed) */
3471extern int __kmp_tp_cached; /* whether threadprivate cache has been created
3472 (__kmpc_threadprivate_cached()) */
3473extern int __kmp_dflt_blocktime; /* number of microseconds to wait before
3474 blocking (env setting) */
3475extern char __kmp_blocktime_units; /* 'm' or 'u' to note units specified */
3476extern bool __kmp_wpolicy_passive; /* explicitly set passive wait policy */
3477
3478// Convert raw blocktime from ms to us if needed.
3479static inline void __kmp_aux_convert_blocktime(int *bt) {
3480 if (__kmp_blocktime_units == 'm') {
3481 if (*bt > INT_MAX / 1000) {
3482 *bt = INT_MAX / 1000;
3483 KMP_INFORM(MaxValueUsing, "kmp_set_blocktime(ms)", bt);
3484 }
3485 *bt = *bt * 1000;
3486 }
3487}
3488
3489#if KMP_USE_MONITOR
3490extern int
3491 __kmp_monitor_wakeups; /* number of times monitor wakes up per second */
3492extern int __kmp_bt_intervals; /* number of monitor timestamp intervals before
3493 blocking */
3494#endif
3495#ifdef KMP_ADJUST_BLOCKTIME
3496extern int __kmp_zero_bt; /* whether blocktime has been forced to zero */
3497#endif /* KMP_ADJUST_BLOCKTIME */
3498#ifdef KMP_DFLT_NTH_CORES
3499extern int __kmp_ncores; /* Total number of cores for threads placement */
3500#endif
3501/* Number of millisecs to delay on abort for Intel(R) VTune(TM) tools */
3502extern int __kmp_abort_delay;
3503
3504extern int __kmp_need_register_atfork_specified;
3505extern int __kmp_need_register_atfork; /* At initialization, call pthread_atfork
3506 to install fork handler */
3507extern int __kmp_gtid_mode; /* Method of getting gtid, values:
3508 0 - not set, will be set at runtime
3509 1 - using stack search
3510 2 - dynamic TLS (pthread_getspecific(Linux* OS/OS
3511 X*) or TlsGetValue(Windows* OS))
3512 3 - static TLS (__declspec(thread) __kmp_gtid),
3513 Linux* OS .so only. */
3514extern int
3515 __kmp_adjust_gtid_mode; /* If true, adjust method based on #threads */
3516#ifdef KMP_TDATA_GTID
3517extern KMP_THREAD_LOCAL int __kmp_gtid;
3518#endif
3519extern int __kmp_tls_gtid_min; /* #threads below which use sp search for gtid */
3520extern int __kmp_foreign_tp; // If true, separate TP var for each foreign thread
3521#if KMP_ARCH_X86 || KMP_ARCH_X86_64
3522extern int __kmp_inherit_fp_control; // copy fp creg(s) parent->workers at fork
3523extern kmp_int16 __kmp_init_x87_fpu_control_word; // init thread's FP ctrl reg
3524extern kmp_uint32 __kmp_init_mxcsr; /* init thread's mxscr */
3525#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */
3526
3527// max_active_levels for nested parallelism enabled by default via
3528// OMP_MAX_ACTIVE_LEVELS, OMP_NESTED, OMP_NUM_THREADS, and OMP_PROC_BIND
3529extern int __kmp_dflt_max_active_levels;
3530// Indicates whether value of __kmp_dflt_max_active_levels was already
3531// explicitly set by OMP_MAX_ACTIVE_LEVELS or OMP_NESTED=false
3532extern bool __kmp_dflt_max_active_levels_set;
3533extern int __kmp_dispatch_num_buffers; /* max possible dynamic loops in
3534 concurrent execution per team */
3535#if KMP_NESTED_HOT_TEAMS
3536extern int __kmp_hot_teams_mode;
3537extern int __kmp_hot_teams_max_level;
3538#endif
3539
3540#if KMP_MIC_SUPPORTED
3541extern enum mic_type __kmp_mic_type;
3542#endif
3543
3544#ifdef USE_LOAD_BALANCE
3545extern double __kmp_load_balance_interval; // load balance algorithm interval
3546#endif /* USE_LOAD_BALANCE */
3547
3548#if KMP_USE_ADAPTIVE_LOCKS
3549
3550// Parameters for the speculative lock backoff system.
3551struct kmp_adaptive_backoff_params_t {
3552 // Number of soft retries before it counts as a hard retry.
3553 kmp_uint32 max_soft_retries;
3554 // Badness is a bit mask : 0,1,3,7,15,... on each hard failure we move one to
3555 // the right
3556 kmp_uint32 max_badness;
3557};
3558
3559extern kmp_adaptive_backoff_params_t __kmp_adaptive_backoff_params;
3560
3561#if KMP_DEBUG_ADAPTIVE_LOCKS
3562extern const char *__kmp_speculative_statsfile;
3563#endif
3564
3565#endif // KMP_USE_ADAPTIVE_LOCKS
3566
3567extern int __kmp_display_env; /* TRUE or FALSE */
3568extern int __kmp_display_env_verbose; /* TRUE if OMP_DISPLAY_ENV=VERBOSE */
3569extern int __kmp_omp_cancellation; /* TRUE or FALSE */
3570extern int __kmp_nteams;
3571extern int __kmp_teams_thread_limit;
3572
3573/* ------------------------------------------------------------------------- */
3574
3575/* the following are protected by the fork/join lock */
3576/* write: lock read: anytime */
3577extern kmp_info_t **__kmp_threads; /* Descriptors for the threads */
3578/* Holds old arrays of __kmp_threads until library shutdown */
3579extern kmp_old_threads_list_t *__kmp_old_threads_list;
3580/* read/write: lock */
3581extern volatile kmp_team_t *__kmp_team_pool;
3582extern volatile kmp_info_t *__kmp_thread_pool;
3583extern kmp_info_t *__kmp_thread_pool_insert_pt;
3584
3585// total num threads reachable from some root thread including all root threads
3586extern volatile int __kmp_nth;
3587/* total number of threads reachable from some root thread including all root
3588 threads, and those in the thread pool */
3589extern volatile int __kmp_all_nth;
3590extern std::atomic<int> __kmp_thread_pool_active_nth;
3591
3592extern kmp_root_t **__kmp_root; /* root of thread hierarchy */
3593/* end data protected by fork/join lock */
3594/* ------------------------------------------------------------------------- */
3595
3596#define __kmp_get_gtid() __kmp_get_global_thread_id()
3597#define __kmp_entry_gtid() __kmp_get_global_thread_id_reg()
3598#define __kmp_get_tid() (__kmp_tid_from_gtid(__kmp_get_gtid()))
3599#define __kmp_get_team() (__kmp_threads[(__kmp_get_gtid())]->th.th_team)
3600#define __kmp_get_thread() (__kmp_thread_from_gtid(__kmp_get_gtid()))
3601
3602// AT: Which way is correct?
3603// AT: 1. nproc = __kmp_threads[ ( gtid ) ] -> th.th_team -> t.t_nproc;
3604// AT: 2. nproc = __kmp_threads[ ( gtid ) ] -> th.th_team_nproc;
3605#define __kmp_get_team_num_threads(gtid) \
3606 (__kmp_threads[(gtid)]->th.th_team->t.t_nproc)
3607
3608static inline bool KMP_UBER_GTID(int gtid) {
3609 KMP_DEBUG_ASSERT(gtid >= KMP_GTID_MIN);
3610 KMP_DEBUG_ASSERT(gtid < __kmp_threads_capacity);
3611 return (gtid >= 0 && __kmp_root[gtid] && __kmp_threads[gtid] &&
3612 __kmp_threads[gtid] == __kmp_root[gtid]->r.r_uber_thread);
3613}
3614
3615static inline int __kmp_tid_from_gtid(int gtid) {
3616 KMP_DEBUG_ASSERT(gtid >= 0);
3617 return __kmp_threads[gtid]->th.th_info.ds.ds_tid;
3618}
3619
3620static inline int __kmp_gtid_from_tid(int tid, const kmp_team_t *team) {
3621 KMP_DEBUG_ASSERT(tid >= 0 && team);
3622 return team->t.t_threads[tid]->th.th_info.ds.ds_gtid;
3623}
3624
3625static inline int __kmp_gtid_from_thread(const kmp_info_t *thr) {
3626 KMP_DEBUG_ASSERT(thr);
3627 return thr->th.th_info.ds.ds_gtid;
3628}
3629
3630static inline kmp_info_t *__kmp_thread_from_gtid(int gtid) {
3631 KMP_DEBUG_ASSERT(gtid >= 0);
3632 return __kmp_threads[gtid];
3633}
3634
3635static inline kmp_team_t *__kmp_team_from_gtid(int gtid) {
3636 KMP_DEBUG_ASSERT(gtid >= 0);
3637 return __kmp_threads[gtid]->th.th_team;
3638}
3639
3640static inline void __kmp_assert_valid_gtid(kmp_int32 gtid) {
3641 if (UNLIKELY(gtid < 0 || gtid >= __kmp_threads_capacity))
3642 KMP_FATAL(ThreadIdentInvalid);
3643}
3644
3645#if KMP_HAVE_MWAIT || KMP_HAVE_UMWAIT
3646extern int __kmp_user_level_mwait; // TRUE or FALSE; from KMP_USER_LEVEL_MWAIT
3647extern int __kmp_umwait_enabled; // Runtime check if user-level mwait enabled
3648extern int __kmp_mwait_enabled; // Runtime check if ring3 mwait is enabled
3649extern int __kmp_mwait_hints; // Hints to pass in to mwait
3650#endif
3651
3652#if KMP_HAVE_UMWAIT
3653extern int __kmp_waitpkg_enabled; // Runtime check if waitpkg exists
3654extern int __kmp_tpause_state; // 0 (default), 1=C0.1, 2=C0.2; from KMP_TPAUSE
3655extern int __kmp_tpause_hint; // 1=C0.1 (default), 0=C0.2; from KMP_TPAUSE
3656extern int __kmp_tpause_enabled; // 0 (default), 1 (KMP_TPAUSE is non-zero)
3657#endif
3658
3659/* ------------------------------------------------------------------------- */
3660
3661extern kmp_global_t __kmp_global; /* global status */
3662
3663extern kmp_info_t __kmp_monitor;
3664// For Debugging Support Library
3665extern std::atomic<kmp_int32> __kmp_team_counter;
3666// For Debugging Support Library
3667extern std::atomic<kmp_int32> __kmp_task_counter;
3668
3669#if USE_DEBUGGER
3670#define _KMP_GEN_ID(counter) \
3671 (__kmp_debugging ? KMP_ATOMIC_INC(&counter) + 1 : ~0)
3672#else
3673#define _KMP_GEN_ID(counter) (~0)
3674#endif /* USE_DEBUGGER */
3675
3676#define KMP_GEN_TASK_ID() _KMP_GEN_ID(__kmp_task_counter)
3677#define KMP_GEN_TEAM_ID() _KMP_GEN_ID(__kmp_team_counter)
3678
3679/* ------------------------------------------------------------------------ */
3680
3681extern void __kmp_print_storage_map_gtid(int gtid, void *p1, void *p2,
3682 size_t size, char const *format, ...);
3683
3684extern void __kmp_serial_initialize(void);
3685extern void __kmp_middle_initialize(void);
3686extern void __kmp_parallel_initialize(void);
3687
3688extern void __kmp_internal_begin(void);
3689extern void __kmp_internal_end_library(int gtid);
3690extern void __kmp_internal_end_thread(int gtid);
3691extern void __kmp_internal_end_atexit(void);
3692extern void __kmp_internal_end_dtor(void);
3693extern void __kmp_internal_end_dest(void *);
3694
3695extern int __kmp_register_root(int initial_thread);
3696extern void __kmp_unregister_root(int gtid);
3697extern void __kmp_unregister_library(void); // called by __kmp_internal_end()
3698
3699extern int __kmp_ignore_mppbeg(void);
3700extern int __kmp_ignore_mppend(void);
3701
3702extern int __kmp_enter_single(int gtid, ident_t *id_ref, int push_ws);
3703extern void __kmp_exit_single(int gtid);
3704
3705extern void __kmp_parallel_deo(int *gtid_ref, int *cid_ref, ident_t *loc_ref);
3706extern void __kmp_parallel_dxo(int *gtid_ref, int *cid_ref, ident_t *loc_ref);
3707
3708#ifdef USE_LOAD_BALANCE
3709extern int __kmp_get_load_balance(int);
3710#endif
3711
3712extern int __kmp_get_global_thread_id(void);
3713extern int __kmp_get_global_thread_id_reg(void);
3714extern void __kmp_exit_thread(int exit_status);
3715extern void __kmp_abort(char const *format, ...);
3716extern void __kmp_abort_thread(void);
3717KMP_NORETURN extern void __kmp_abort_process(void);
3718extern void __kmp_warn(char const *format, ...);
3719
3720extern void __kmp_set_num_threads(int new_nth, int gtid);
3721
3722extern bool __kmp_detect_shm();
3723extern bool __kmp_detect_tmp();
3724
3725// Returns current thread (pointer to kmp_info_t). Current thread *must* be
3726// registered.
3727static inline kmp_info_t *__kmp_entry_thread() {
3728 int gtid = __kmp_entry_gtid();
3729
3730 return __kmp_threads[gtid];
3731}
3732
3733extern void __kmp_set_max_active_levels(int gtid, int new_max_active_levels);
3734extern int __kmp_get_max_active_levels(int gtid);
3735extern int __kmp_get_ancestor_thread_num(int gtid, int level);
3736extern int __kmp_get_team_size(int gtid, int level);
3737extern void __kmp_set_schedule(int gtid, kmp_sched_t new_sched, int chunk);
3738extern void __kmp_get_schedule(int gtid, kmp_sched_t *sched, int *chunk);
3739
3740extern unsigned short __kmp_get_random(kmp_info_t *thread);
3741extern void __kmp_init_random(kmp_info_t *thread);
3742
3743extern kmp_r_sched_t __kmp_get_schedule_global(void);
3744extern void __kmp_adjust_num_threads(int new_nproc);
3745extern void __kmp_check_stksize(size_t *val);
3746
3747extern void *___kmp_allocate(size_t size KMP_SRC_LOC_DECL);
3748extern void *___kmp_page_allocate(size_t size KMP_SRC_LOC_DECL);
3749extern void ___kmp_free(void *ptr KMP_SRC_LOC_DECL);
3750#define __kmp_allocate(size) ___kmp_allocate((size)KMP_SRC_LOC_CURR)
3751#define __kmp_page_allocate(size) ___kmp_page_allocate((size)KMP_SRC_LOC_CURR)
3752#define __kmp_free(ptr) ___kmp_free((ptr)KMP_SRC_LOC_CURR)
3753
3754#if USE_FAST_MEMORY
3755extern void *___kmp_fast_allocate(kmp_info_t *this_thr,
3756 size_t size KMP_SRC_LOC_DECL);
3757extern void ___kmp_fast_free(kmp_info_t *this_thr, void *ptr KMP_SRC_LOC_DECL);
3758extern void __kmp_free_fast_memory(kmp_info_t *this_thr);
3759extern void __kmp_initialize_fast_memory(kmp_info_t *this_thr);
3760#define __kmp_fast_allocate(this_thr, size) \
3761 ___kmp_fast_allocate((this_thr), (size)KMP_SRC_LOC_CURR)
3762#define __kmp_fast_free(this_thr, ptr) \
3763 ___kmp_fast_free((this_thr), (ptr)KMP_SRC_LOC_CURR)
3764#endif
3765
3766extern void *___kmp_thread_malloc(kmp_info_t *th, size_t size KMP_SRC_LOC_DECL);
3767extern void *___kmp_thread_calloc(kmp_info_t *th, size_t nelem,
3768 size_t elsize KMP_SRC_LOC_DECL);
3769extern void *___kmp_thread_realloc(kmp_info_t *th, void *ptr,
3770 size_t size KMP_SRC_LOC_DECL);
3771extern void ___kmp_thread_free(kmp_info_t *th, void *ptr KMP_SRC_LOC_DECL);
3772#define __kmp_thread_malloc(th, size) \
3773 ___kmp_thread_malloc((th), (size)KMP_SRC_LOC_CURR)
3774#define __kmp_thread_calloc(th, nelem, elsize) \
3775 ___kmp_thread_calloc((th), (nelem), (elsize)KMP_SRC_LOC_CURR)
3776#define __kmp_thread_realloc(th, ptr, size) \
3777 ___kmp_thread_realloc((th), (ptr), (size)KMP_SRC_LOC_CURR)
3778#define __kmp_thread_free(th, ptr) \
3779 ___kmp_thread_free((th), (ptr)KMP_SRC_LOC_CURR)
3780
3781extern void __kmp_push_num_threads(ident_t *loc, int gtid, int num_threads);
3782extern void __kmp_push_num_threads_list(ident_t *loc, int gtid,
3783 kmp_uint32 list_length,
3784 int *num_threads_list);
3785extern void __kmp_set_strict_num_threads(ident_t *loc, int gtid, int sev,
3786 const char *msg);
3787
3788extern void __kmp_push_proc_bind(ident_t *loc, int gtid,
3789 kmp_proc_bind_t proc_bind);
3790extern void __kmp_push_num_teams(ident_t *loc, int gtid, int num_teams,
3791 int num_threads);
3792extern void __kmp_push_num_teams_51(ident_t *loc, int gtid, int num_teams_lb,
3793 int num_teams_ub, int num_threads);
3794
3795extern void __kmp_yield();
3796
3797extern void __kmpc_dispatch_init_4(ident_t *loc, kmp_int32 gtid,
3798 enum sched_type schedule, kmp_int32 lb,
3799 kmp_int32 ub, kmp_int32 st, kmp_int32 chunk);
3800extern void __kmpc_dispatch_init_4u(ident_t *loc, kmp_int32 gtid,
3801 enum sched_type schedule, kmp_uint32 lb,
3802 kmp_uint32 ub, kmp_int32 st,
3803 kmp_int32 chunk);
3804extern void __kmpc_dispatch_init_8(ident_t *loc, kmp_int32 gtid,
3805 enum sched_type schedule, kmp_int64 lb,
3806 kmp_int64 ub, kmp_int64 st, kmp_int64 chunk);
3807extern void __kmpc_dispatch_init_8u(ident_t *loc, kmp_int32 gtid,
3808 enum sched_type schedule, kmp_uint64 lb,
3809 kmp_uint64 ub, kmp_int64 st,
3810 kmp_int64 chunk);
3811
3812extern int __kmpc_dispatch_next_4(ident_t *loc, kmp_int32 gtid,
3813 kmp_int32 *p_last, kmp_int32 *p_lb,
3814 kmp_int32 *p_ub, kmp_int32 *p_st);
3815extern int __kmpc_dispatch_next_4u(ident_t *loc, kmp_int32 gtid,
3816 kmp_int32 *p_last, kmp_uint32 *p_lb,
3817 kmp_uint32 *p_ub, kmp_int32 *p_st);
3818extern int __kmpc_dispatch_next_8(ident_t *loc, kmp_int32 gtid,
3819 kmp_int32 *p_last, kmp_int64 *p_lb,
3820 kmp_int64 *p_ub, kmp_int64 *p_st);
3821extern int __kmpc_dispatch_next_8u(ident_t *loc, kmp_int32 gtid,
3822 kmp_int32 *p_last, kmp_uint64 *p_lb,
3823 kmp_uint64 *p_ub, kmp_int64 *p_st);
3824
3825extern void __kmpc_dispatch_fini_4(ident_t *loc, kmp_int32 gtid);
3826extern void __kmpc_dispatch_fini_8(ident_t *loc, kmp_int32 gtid);
3827extern void __kmpc_dispatch_fini_4u(ident_t *loc, kmp_int32 gtid);
3828extern void __kmpc_dispatch_fini_8u(ident_t *loc, kmp_int32 gtid);
3829
3830extern void __kmpc_dispatch_deinit(ident_t *loc, kmp_int32 gtid);
3831
3832#ifdef KMP_GOMP_COMPAT
3833
3834extern void __kmp_aux_dispatch_init_4(ident_t *loc, kmp_int32 gtid,
3835 enum sched_type schedule, kmp_int32 lb,
3836 kmp_int32 ub, kmp_int32 st,
3837 kmp_int32 chunk, int push_ws);
3838extern void __kmp_aux_dispatch_init_4u(ident_t *loc, kmp_int32 gtid,
3839 enum sched_type schedule, kmp_uint32 lb,
3840 kmp_uint32 ub, kmp_int32 st,
3841 kmp_int32 chunk, int push_ws);
3842extern void __kmp_aux_dispatch_init_8(ident_t *loc, kmp_int32 gtid,
3843 enum sched_type schedule, kmp_int64 lb,
3844 kmp_int64 ub, kmp_int64 st,
3845 kmp_int64 chunk, int push_ws);
3846extern void __kmp_aux_dispatch_init_8u(ident_t *loc, kmp_int32 gtid,
3847 enum sched_type schedule, kmp_uint64 lb,
3848 kmp_uint64 ub, kmp_int64 st,
3849 kmp_int64 chunk, int push_ws);
3850extern void __kmp_aux_dispatch_fini_chunk_4(ident_t *loc, kmp_int32 gtid);
3851extern void __kmp_aux_dispatch_fini_chunk_8(ident_t *loc, kmp_int32 gtid);
3852extern void __kmp_aux_dispatch_fini_chunk_4u(ident_t *loc, kmp_int32 gtid);
3853extern void __kmp_aux_dispatch_fini_chunk_8u(ident_t *loc, kmp_int32 gtid);
3854
3855#endif /* KMP_GOMP_COMPAT */
3856
3857extern kmp_uint32 __kmp_eq_4(kmp_uint32 value, kmp_uint32 checker);
3858extern kmp_uint32 __kmp_neq_4(kmp_uint32 value, kmp_uint32 checker);
3859extern kmp_uint32 __kmp_lt_4(kmp_uint32 value, kmp_uint32 checker);
3860extern kmp_uint32 __kmp_ge_4(kmp_uint32 value, kmp_uint32 checker);
3861extern kmp_uint32 __kmp_le_4(kmp_uint32 value, kmp_uint32 checker);
3862extern kmp_uint32 __kmp_wait_4(kmp_uint32 volatile *spinner, kmp_uint32 checker,
3863 kmp_uint32 (*pred)(kmp_uint32, kmp_uint32),
3864 void *obj);
3865extern void __kmp_wait_4_ptr(void *spinner, kmp_uint32 checker,
3866 kmp_uint32 (*pred)(void *, kmp_uint32), void *obj);
3867
3868extern void __kmp_wait_64(kmp_info_t *this_thr, kmp_flag_64<> *flag,
3869 int final_spin
3870#if USE_ITT_BUILD
3871 ,
3872 void *itt_sync_obj
3873#endif
3874);
3875extern void __kmp_release_64(kmp_flag_64<> *flag);
3876
3877extern void __kmp_infinite_loop(void);
3878
3879extern void __kmp_cleanup(void);
3880
3881#if KMP_HANDLE_SIGNALS
3882extern int __kmp_handle_signals;
3883extern void __kmp_install_signals(int parallel_init);
3884extern void __kmp_remove_signals(void);
3885#endif
3886
3887extern void __kmp_clear_system_time(void);
3888extern void __kmp_read_system_time(double *delta);
3889
3890extern void __kmp_check_stack_overlap(kmp_info_t *thr);
3891
3892extern void __kmp_expand_host_name(char *buffer, size_t size);
3893extern void __kmp_expand_file_name(char *result, size_t rlen, char *pattern);
3894
3895#if KMP_ARCH_X86 || KMP_ARCH_X86_64 || (KMP_OS_WINDOWS && (KMP_ARCH_AARCH64 || KMP_ARCH_ARM))
3896extern void
3897__kmp_initialize_system_tick(void); /* Initialize timer tick value */
3898#endif
3899
3900extern void
3901__kmp_runtime_initialize(void); /* machine specific initialization */
3902extern void __kmp_runtime_destroy(void);
3903
3904#if KMP_AFFINITY_SUPPORTED
3905extern char *__kmp_affinity_print_mask(char *buf, int buf_len,
3906 kmp_affin_mask_t *mask);
3907extern kmp_str_buf_t *__kmp_affinity_str_buf_mask(kmp_str_buf_t *buf,
3908 kmp_affin_mask_t *mask);
3909extern void __kmp_affinity_initialize(kmp_affinity_t &affinity);
3910extern void __kmp_affinity_uninitialize(void);
3911extern void __kmp_affinity_set_init_mask(
3912 int gtid, int isa_root); /* set affinity according to KMP_AFFINITY */
3913void __kmp_affinity_bind_init_mask(int gtid);
3914extern void __kmp_affinity_bind_place(int gtid);
3915extern void __kmp_affinity_determine_capable(const char *env_var);
3916extern int __kmp_aux_set_affinity(void **mask);
3917extern int __kmp_aux_get_affinity(void **mask);
3918extern int __kmp_aux_get_affinity_max_proc();
3919extern int __kmp_aux_set_affinity_mask_proc(int proc, void **mask);
3920extern int __kmp_aux_unset_affinity_mask_proc(int proc, void **mask);
3921extern int __kmp_aux_get_affinity_mask_proc(int proc, void **mask);
3922extern void __kmp_balanced_affinity(kmp_info_t *th, int team_size);
3923#if KMP_WEIGHTED_ITERATIONS_SUPPORTED
3924extern int __kmp_get_first_osid_with_ecore(void);
3925#endif
3926#if KMP_OS_LINUX || KMP_OS_FREEBSD || KMP_OS_NETBSD || KMP_OS_DRAGONFLY || \
3927 KMP_OS_AIX
3928extern int kmp_set_thread_affinity_mask_initial(void);
3929#endif
3930static inline void __kmp_assign_root_init_mask() {
3931 int gtid = __kmp_entry_gtid();
3932 kmp_root_t *r = __kmp_threads[gtid]->th.th_root;
3933 if (r->r.r_uber_thread == __kmp_threads[gtid] && !r->r.r_affinity_assigned) {
3934 __kmp_affinity_set_init_mask(gtid, /*isa_root=*/TRUE);
3935 __kmp_affinity_bind_init_mask(gtid);
3936 r->r.r_affinity_assigned = TRUE;
3937 }
3938}
3939static inline void __kmp_reset_root_init_mask(int gtid) {
3940 if (!KMP_AFFINITY_CAPABLE())
3941 return;
3942 kmp_info_t *th = __kmp_threads[gtid];
3943 kmp_root_t *r = th->th.th_root;
3944 if (r->r.r_uber_thread == th && r->r.r_affinity_assigned) {
3945 __kmp_set_system_affinity(__kmp_affin_origMask, FALSE);
3946 KMP_CPU_COPY(th->th.th_affin_mask, __kmp_affin_origMask);
3947 r->r.r_affinity_assigned = FALSE;
3948 }
3949}
3950#else /* KMP_AFFINITY_SUPPORTED */
3951#define __kmp_assign_root_init_mask() /* Nothing */
3952static inline void __kmp_reset_root_init_mask(int gtid) {}
3953#endif /* KMP_AFFINITY_SUPPORTED */
3954// No need for KMP_AFFINITY_SUPPORTED guard as only one field in the
3955// format string is for affinity, so platforms that do not support
3956// affinity can still use the other fields, e.g., %n for num_threads
3957extern size_t __kmp_aux_capture_affinity(int gtid, const char *format,
3958 kmp_str_buf_t *buffer);
3959extern void __kmp_aux_display_affinity(int gtid, const char *format);
3960
3961extern void __kmp_cleanup_hierarchy();
3962extern void __kmp_get_hierarchy(kmp_uint32 nproc, kmp_bstate_t *thr_bar);
3963
3964#if KMP_USE_FUTEX
3965
3966extern int __kmp_futex_determine_capable(void);
3967
3968#endif // KMP_USE_FUTEX
3969
3970extern void __kmp_gtid_set_specific(int gtid);
3971extern int __kmp_gtid_get_specific(void);
3972
3973extern double __kmp_read_cpu_time(void);
3974
3975extern int __kmp_read_system_info(struct kmp_sys_info *info);
3976
3977#if KMP_USE_MONITOR
3978extern void __kmp_create_monitor(kmp_info_t *th);
3979#endif
3980
3981extern void *__kmp_launch_thread(kmp_info_t *thr);
3982
3983extern void __kmp_create_worker(int gtid, kmp_info_t *th, size_t stack_size);
3984
3985#if KMP_OS_WINDOWS
3986extern int __kmp_still_running(kmp_info_t *th);
3987extern int __kmp_is_thread_alive(kmp_info_t *th, DWORD *exit_val);
3988extern void __kmp_free_handle(kmp_thread_t tHandle);
3989#endif
3990
3991#if KMP_USE_MONITOR
3992extern void __kmp_reap_monitor(kmp_info_t *th);
3993#endif
3994extern void __kmp_reap_worker(kmp_info_t *th);
3995extern void __kmp_terminate_thread(int gtid);
3996
3997extern int __kmp_try_suspend_mx(kmp_info_t *th);
3998extern void __kmp_lock_suspend_mx(kmp_info_t *th);
3999extern void __kmp_unlock_suspend_mx(kmp_info_t *th);
4000
4001extern void __kmp_elapsed(double *);
4002extern void __kmp_elapsed_tick(double *);
4003
4004extern void __kmp_enable(int old_state);
4005extern void __kmp_disable(int *old_state);
4006
4007extern void __kmp_thread_sleep(int millis);
4008
4009extern void __kmp_common_initialize(void);
4010extern void __kmp_common_destroy(void);
4011extern void __kmp_common_destroy_gtid(int gtid);
4012
4013#if KMP_OS_UNIX
4014extern void __kmp_register_atfork(void);
4015#endif
4016extern void __kmp_suspend_initialize(void);
4017extern void __kmp_suspend_initialize_thread(kmp_info_t *th);
4018extern void __kmp_suspend_uninitialize_thread(kmp_info_t *th);
4019
4020extern kmp_info_t *__kmp_allocate_thread(kmp_root_t *root, kmp_team_t *team,
4021 int tid);
4022extern kmp_team_t *
4023__kmp_allocate_team(kmp_root_t *root, int new_nproc, int max_nproc,
4024#if OMPT_SUPPORT
4025 ompt_data_t ompt_parallel_data,
4026#endif
4027 kmp_proc_bind_t proc_bind, kmp_internal_control_t *new_icvs,
4028 int argc USE_NESTED_HOT_ARG(kmp_info_t *thr));
4029extern void __kmp_free_thread(kmp_info_t *);
4030extern void __kmp_free_team(kmp_root_t *,
4031 kmp_team_t *USE_NESTED_HOT_ARG(kmp_info_t *));
4032extern kmp_team_t *__kmp_reap_team(kmp_team_t *);
4033
4034/* ------------------------------------------------------------------------ */
4035
4036extern void __kmp_initialize_bget(kmp_info_t *th);
4037extern void __kmp_finalize_bget(kmp_info_t *th);
4038
4039KMP_EXPORT void *kmpc_malloc(size_t size);
4040KMP_EXPORT void *kmpc_aligned_malloc(size_t size, size_t alignment);
4041KMP_EXPORT void *kmpc_calloc(size_t nelem, size_t elsize);
4042KMP_EXPORT void *kmpc_realloc(void *ptr, size_t size);
4043KMP_EXPORT void kmpc_free(void *ptr);
4044
4045/* declarations for internal use */
4046
4047extern int __kmp_barrier(enum barrier_type bt, int gtid, int is_split,
4048 size_t reduce_size, void *reduce_data,
4049 void (*reduce)(void *, void *));
4050extern void __kmp_end_split_barrier(enum barrier_type bt, int gtid);
4051extern int __kmp_barrier_gomp_cancel(int gtid);
4052
4057enum fork_context_e {
4058 fork_context_gnu,
4060 fork_context_intel,
4061 fork_context_last
4062};
4063extern int __kmp_fork_call(ident_t *loc, int gtid,
4064 enum fork_context_e fork_context, kmp_int32 argc,
4065 microtask_t microtask, launch_t invoker,
4066 kmp_va_list ap);
4067
4068extern void __kmp_join_call(ident_t *loc, int gtid
4069#if OMPT_SUPPORT
4070 ,
4071 enum fork_context_e fork_context
4072#endif
4073 ,
4074 int exit_teams = 0);
4075
4076extern void __kmp_serialized_parallel(ident_t *id, kmp_int32 gtid);
4077extern void __kmp_internal_fork(ident_t *id, int gtid, kmp_team_t *team);
4078extern void __kmp_internal_join(ident_t *id, int gtid, kmp_team_t *team);
4079extern int __kmp_invoke_task_func(int gtid);
4080extern void __kmp_run_before_invoked_task(int gtid, int tid,
4081 kmp_info_t *this_thr,
4082 kmp_team_t *team);
4083extern void __kmp_run_after_invoked_task(int gtid, int tid,
4084 kmp_info_t *this_thr,
4085 kmp_team_t *team);
4086
4087// should never have been exported
4088KMP_EXPORT int __kmpc_invoke_task_func(int gtid);
4089extern int __kmp_invoke_teams_master(int gtid);
4090extern void __kmp_teams_master(int gtid);
4091extern int __kmp_aux_get_team_num();
4092extern int __kmp_aux_get_num_teams();
4093extern void __kmp_save_internal_controls(kmp_info_t *thread);
4094extern void __kmp_user_set_library(enum library_type arg);
4095extern void __kmp_aux_set_library(enum library_type arg);
4096extern void __kmp_aux_set_stacksize(size_t arg);
4097extern void __kmp_aux_set_blocktime(int arg, kmp_info_t *thread, int tid);
4098extern void __kmp_aux_set_defaults(char const *str, size_t len);
4099
4100/* Functions called from __kmp_aux_env_initialize() in kmp_settings.cpp */
4101void kmpc_set_blocktime(int arg);
4102void ompc_set_nested(int flag);
4103void ompc_set_dynamic(int flag);
4104void ompc_set_num_threads(int arg);
4105
4106extern void __kmp_push_current_task_to_thread(kmp_info_t *this_thr,
4107 kmp_team_t *team, int tid);
4108extern void __kmp_pop_current_task_from_thread(kmp_info_t *this_thr);
4109extern kmp_task_t *__kmp_task_alloc(ident_t *loc_ref, kmp_int32 gtid,
4110 kmp_tasking_flags_t *flags,
4111 size_t sizeof_kmp_task_t,
4112 size_t sizeof_shareds,
4113 kmp_routine_entry_t task_entry);
4114extern void __kmp_init_implicit_task(ident_t *loc_ref, kmp_info_t *this_thr,
4115 kmp_team_t *team, int tid,
4116 int set_curr_task);
4117extern void __kmp_finish_implicit_task(kmp_info_t *this_thr);
4118extern void __kmp_free_implicit_task(kmp_info_t *this_thr);
4119
4120extern kmp_event_t *__kmpc_task_allow_completion_event(ident_t *loc_ref,
4121 int gtid,
4122 kmp_task_t *task);
4123extern void __kmp_fulfill_event(kmp_event_t *event);
4124
4125extern void __kmp_free_task_team(kmp_info_t *thread,
4126 kmp_task_team_t *task_team);
4127extern void __kmp_reap_task_teams(void);
4128extern void __kmp_push_task_team_node(kmp_info_t *thread, kmp_team_t *team);
4129extern void __kmp_pop_task_team_node(kmp_info_t *thread, kmp_team_t *team);
4130extern void __kmp_wait_to_unref_task_teams(void);
4131extern void __kmp_task_team_setup(kmp_info_t *this_thr, kmp_team_t *team);
4132extern void __kmp_task_team_sync(kmp_info_t *this_thr, kmp_team_t *team);
4133extern void __kmp_task_team_wait(kmp_info_t *this_thr, kmp_team_t *team
4134#if USE_ITT_BUILD
4135 ,
4136 void *itt_sync_obj
4137#endif /* USE_ITT_BUILD */
4138 ,
4139 int wait = 1);
4140extern void __kmp_tasking_barrier(kmp_team_t *team, kmp_info_t *thread,
4141 int gtid);
4142#if KMP_DEBUG
4143#define KMP_DEBUG_ASSERT_TASKTEAM_INVARIANT(team, thr) \
4144 KMP_DEBUG_ASSERT( \
4145 __kmp_tasking_mode != tskm_task_teams || team->t.t_nproc == 1 || \
4146 thr->th.th_task_team == team->t.t_task_team[thr->th.th_task_state])
4147#else
4148#define KMP_DEBUG_ASSERT_TASKTEAM_INVARIANT(team, thr) /* Nothing */
4149#endif
4150
4151extern int __kmp_is_address_mapped(void *addr);
4152extern kmp_uint64 __kmp_hardware_timestamp(void);
4153
4154#if KMP_OS_UNIX
4155extern int __kmp_read_from_file(char const *path, char const *format, ...);
4156#endif
4157
4158/* ------------------------------------------------------------------------ */
4159//
4160// Assembly routines that have no compiler intrinsic replacement
4161//
4162
4163extern int __kmp_invoke_microtask(microtask_t pkfn, int gtid, int npr, int argc,
4164 void *argv[]
4165#if OMPT_SUPPORT
4166 ,
4167 void **exit_frame_ptr
4168#endif
4169);
4170
4171/* ------------------------------------------------------------------------ */
4172
4173KMP_EXPORT void __kmpc_begin(ident_t *, kmp_int32 flags);
4174KMP_EXPORT void __kmpc_end(ident_t *);
4175
4176KMP_EXPORT void __kmpc_threadprivate_register_vec(ident_t *, void *data,
4177 kmpc_ctor_vec ctor,
4178 kmpc_cctor_vec cctor,
4179 kmpc_dtor_vec dtor,
4180 size_t vector_length);
4181KMP_EXPORT void __kmpc_threadprivate_register(ident_t *, void *data,
4182 kmpc_ctor ctor, kmpc_cctor cctor,
4183 kmpc_dtor dtor);
4184KMP_EXPORT void *__kmpc_threadprivate(ident_t *, kmp_int32 global_tid,
4185 void *data, size_t size);
4186
4187KMP_EXPORT kmp_int32 __kmpc_global_thread_num(ident_t *);
4188KMP_EXPORT kmp_int32 __kmpc_global_num_threads(ident_t *);
4189KMP_EXPORT kmp_int32 __kmpc_bound_thread_num(ident_t *);
4190KMP_EXPORT kmp_int32 __kmpc_bound_num_threads(ident_t *);
4191
4192KMP_EXPORT kmp_int32 __kmpc_ok_to_fork(ident_t *);
4193KMP_EXPORT void __kmpc_fork_call(ident_t *, kmp_int32 nargs,
4194 kmpc_micro microtask, ...);
4195KMP_EXPORT void __kmpc_fork_call_if(ident_t *loc, kmp_int32 nargs,
4196 kmpc_micro microtask, kmp_int32 cond,
4197 void *args);
4198
4199KMP_EXPORT void __kmpc_serialized_parallel(ident_t *, kmp_int32 global_tid);
4200KMP_EXPORT void __kmpc_end_serialized_parallel(ident_t *, kmp_int32 global_tid);
4201
4202KMP_EXPORT void __kmpc_flush(ident_t *);
4203KMP_EXPORT void __kmpc_barrier(ident_t *, kmp_int32 global_tid);
4204KMP_EXPORT kmp_int32 __kmpc_master(ident_t *, kmp_int32 global_tid);
4205KMP_EXPORT void __kmpc_end_master(ident_t *, kmp_int32 global_tid);
4206KMP_EXPORT kmp_int32 __kmpc_masked(ident_t *, kmp_int32 global_tid,
4207 kmp_int32 filter);
4208KMP_EXPORT void __kmpc_end_masked(ident_t *, kmp_int32 global_tid);
4209KMP_EXPORT void __kmpc_ordered(ident_t *, kmp_int32 global_tid);
4210KMP_EXPORT void __kmpc_end_ordered(ident_t *, kmp_int32 global_tid);
4211KMP_EXPORT void __kmpc_critical(ident_t *, kmp_int32 global_tid,
4212 kmp_critical_name *);
4213KMP_EXPORT void __kmpc_end_critical(ident_t *, kmp_int32 global_tid,
4214 kmp_critical_name *);
4215KMP_EXPORT void __kmpc_critical_with_hint(ident_t *, kmp_int32 global_tid,
4216 kmp_critical_name *, uint32_t hint);
4217
4218KMP_EXPORT kmp_int32 __kmpc_barrier_master(ident_t *, kmp_int32 global_tid);
4219KMP_EXPORT void __kmpc_end_barrier_master(ident_t *, kmp_int32 global_tid);
4220
4221KMP_EXPORT kmp_int32 __kmpc_barrier_master_nowait(ident_t *,
4222 kmp_int32 global_tid);
4223
4224KMP_EXPORT kmp_int32 __kmpc_single(ident_t *, kmp_int32 global_tid);
4225KMP_EXPORT void __kmpc_end_single(ident_t *, kmp_int32 global_tid);
4226
4227KMP_EXPORT kmp_int32 __kmpc_sections_init(ident_t *loc, kmp_int32 global_tid);
4228KMP_EXPORT kmp_int32 __kmpc_next_section(ident_t *loc, kmp_int32 global_tid,
4229 kmp_int32 numberOfSections);
4230KMP_EXPORT void __kmpc_end_sections(ident_t *loc, kmp_int32 global_tid);
4231
4232KMP_EXPORT void KMPC_FOR_STATIC_INIT(ident_t *loc, kmp_int32 global_tid,
4233 kmp_int32 schedtype, kmp_int32 *plastiter,
4234 kmp_int *plower, kmp_int *pupper,
4235 kmp_int *pstride, kmp_int incr,
4236 kmp_int chunk);
4237
4238KMP_EXPORT void __kmpc_for_static_fini(ident_t *loc, kmp_int32 global_tid);
4239
4240KMP_EXPORT void __kmpc_copyprivate(ident_t *loc, kmp_int32 global_tid,
4241 size_t cpy_size, void *cpy_data,
4242 void (*cpy_func)(void *, void *),
4243 kmp_int32 didit);
4244
4245KMP_EXPORT void *__kmpc_copyprivate_light(ident_t *loc, kmp_int32 gtid,
4246 void *cpy_data);
4247
4248extern void KMPC_SET_NUM_THREADS(int arg);
4249extern void KMPC_SET_DYNAMIC(int flag);
4250extern void KMPC_SET_NESTED(int flag);
4251
4252/* OMP 3.0 tasking interface routines */
4253KMP_EXPORT kmp_int32 __kmpc_omp_task(ident_t *loc_ref, kmp_int32 gtid,
4254 kmp_task_t *new_task);
4255KMP_EXPORT kmp_task_t *__kmpc_omp_task_alloc(ident_t *loc_ref, kmp_int32 gtid,
4256 kmp_int32 flags,
4257 size_t sizeof_kmp_task_t,
4258 size_t sizeof_shareds,
4259 kmp_routine_entry_t task_entry);
4260KMP_EXPORT kmp_task_t *__kmpc_omp_target_task_alloc(
4261 ident_t *loc_ref, kmp_int32 gtid, kmp_int32 flags, size_t sizeof_kmp_task_t,
4262 size_t sizeof_shareds, kmp_routine_entry_t task_entry, kmp_int64 device_id);
4263KMP_EXPORT void __kmpc_omp_task_begin_if0(ident_t *loc_ref, kmp_int32 gtid,
4264 kmp_task_t *task);
4265KMP_EXPORT void __kmpc_omp_task_complete_if0(ident_t *loc_ref, kmp_int32 gtid,
4266 kmp_task_t *task);
4267KMP_EXPORT kmp_int32 __kmpc_omp_task_parts(ident_t *loc_ref, kmp_int32 gtid,
4268 kmp_task_t *new_task);
4269KMP_EXPORT kmp_int32 __kmpc_omp_taskwait(ident_t *loc_ref, kmp_int32 gtid);
4270KMP_EXPORT kmp_int32 __kmpc_omp_taskyield(ident_t *loc_ref, kmp_int32 gtid,
4271 int end_part);
4272
4273#if TASK_UNUSED
4274void __kmpc_omp_task_begin(ident_t *loc_ref, kmp_int32 gtid, kmp_task_t *task);
4275void __kmpc_omp_task_complete(ident_t *loc_ref, kmp_int32 gtid,
4276 kmp_task_t *task);
4277#endif // TASK_UNUSED
4278
4279/* ------------------------------------------------------------------------ */
4280
4281KMP_EXPORT void __kmpc_taskgroup(ident_t *loc, int gtid);
4282KMP_EXPORT void __kmpc_end_taskgroup(ident_t *loc, int gtid);
4283
4284KMP_EXPORT kmp_int32 __kmpc_omp_task_with_deps(
4285 ident_t *loc_ref, kmp_int32 gtid, kmp_task_t *new_task, kmp_int32 ndeps,
4286 kmp_depend_info_t *dep_list, kmp_int32 ndeps_noalias,
4287 kmp_depend_info_t *noalias_dep_list);
4288
4289KMP_EXPORT kmp_base_depnode_t *__kmpc_task_get_depnode(kmp_task_t *task);
4290
4291KMP_EXPORT kmp_depnode_list_t *__kmpc_task_get_successors(kmp_task_t *task);
4292
4293KMP_EXPORT void __kmpc_omp_wait_deps(ident_t *loc_ref, kmp_int32 gtid,
4294 kmp_int32 ndeps,
4295 kmp_depend_info_t *dep_list,
4296 kmp_int32 ndeps_noalias,
4297 kmp_depend_info_t *noalias_dep_list);
4298/* __kmpc_omp_taskwait_deps_51 : Function for OpenMP 5.1 nowait clause.
4299 * Placeholder for taskwait with nowait clause.*/
4300KMP_EXPORT void __kmpc_omp_taskwait_deps_51(ident_t *loc_ref, kmp_int32 gtid,
4301 kmp_int32 ndeps,
4302 kmp_depend_info_t *dep_list,
4303 kmp_int32 ndeps_noalias,
4304 kmp_depend_info_t *noalias_dep_list,
4305 kmp_int32 has_no_wait);
4306
4307extern kmp_int32 __kmp_omp_task(kmp_int32 gtid, kmp_task_t *new_task,
4308 bool serialize_immediate);
4309
4310KMP_EXPORT kmp_int32 __kmpc_cancel(ident_t *loc_ref, kmp_int32 gtid,
4311 kmp_int32 cncl_kind);
4312KMP_EXPORT kmp_int32 __kmpc_cancellationpoint(ident_t *loc_ref, kmp_int32 gtid,
4313 kmp_int32 cncl_kind);
4314KMP_EXPORT kmp_int32 __kmpc_cancel_barrier(ident_t *loc_ref, kmp_int32 gtid);
4315KMP_EXPORT int __kmp_get_cancellation_status(int cancel_kind);
4316
4317KMP_EXPORT void __kmpc_proxy_task_completed(kmp_int32 gtid, kmp_task_t *ptask);
4318KMP_EXPORT void __kmpc_proxy_task_completed_ooo(kmp_task_t *ptask);
4319KMP_EXPORT void __kmpc_taskloop(ident_t *loc, kmp_int32 gtid, kmp_task_t *task,
4320 kmp_int32 if_val, kmp_uint64 *lb,
4321 kmp_uint64 *ub, kmp_int64 st, kmp_int32 nogroup,
4322 kmp_int32 sched, kmp_uint64 grainsize,
4323 void *task_dup);
4324KMP_EXPORT void __kmpc_taskloop_5(ident_t *loc, kmp_int32 gtid,
4325 kmp_task_t *task, kmp_int32 if_val,
4326 kmp_uint64 *lb, kmp_uint64 *ub, kmp_int64 st,
4327 kmp_int32 nogroup, kmp_int32 sched,
4328 kmp_uint64 grainsize, kmp_int32 modifier,
4329 void *task_dup);
4330KMP_EXPORT void *__kmpc_task_reduction_init(int gtid, int num_data, void *data);
4331KMP_EXPORT void *__kmpc_taskred_init(int gtid, int num_data, void *data);
4332KMP_EXPORT void *__kmpc_task_reduction_get_th_data(int gtid, void *tg, void *d);
4333KMP_EXPORT void *__kmpc_task_reduction_modifier_init(ident_t *loc, int gtid,
4334 int is_ws, int num,
4335 void *data);
4336KMP_EXPORT void *__kmpc_taskred_modifier_init(ident_t *loc, int gtid, int is_ws,
4337 int num, void *data);
4338KMP_EXPORT void __kmpc_task_reduction_modifier_fini(ident_t *loc, int gtid,
4339 int is_ws);
4340KMP_EXPORT kmp_int32 __kmpc_omp_reg_task_with_affinity(
4341 ident_t *loc_ref, kmp_int32 gtid, kmp_task_t *new_task, kmp_int32 naffins,
4342 kmp_task_affinity_info_t *affin_list);
4343KMP_EXPORT void __kmp_set_num_teams(int num_teams);
4344KMP_EXPORT int __kmp_get_max_teams(void);
4345KMP_EXPORT void __kmp_set_teams_thread_limit(int limit);
4346KMP_EXPORT int __kmp_get_teams_thread_limit(void);
4347
4348/* Interface target task integration */
4349KMP_EXPORT void **__kmpc_omp_get_target_async_handle_ptr(kmp_int32 gtid);
4350KMP_EXPORT bool __kmpc_omp_has_task_team(kmp_int32 gtid);
4351
4352/* Lock interface routines (fast versions with gtid passed in) */
4353KMP_EXPORT void __kmpc_init_lock(ident_t *loc, kmp_int32 gtid,
4354 void **user_lock);
4355KMP_EXPORT void __kmpc_init_nest_lock(ident_t *loc, kmp_int32 gtid,
4356 void **user_lock);
4357KMP_EXPORT void __kmpc_destroy_lock(ident_t *loc, kmp_int32 gtid,
4358 void **user_lock);
4359KMP_EXPORT void __kmpc_destroy_nest_lock(ident_t *loc, kmp_int32 gtid,
4360 void **user_lock);
4361KMP_EXPORT void __kmpc_set_lock(ident_t *loc, kmp_int32 gtid, void **user_lock);
4362KMP_EXPORT void __kmpc_set_nest_lock(ident_t *loc, kmp_int32 gtid,
4363 void **user_lock);
4364KMP_EXPORT void __kmpc_unset_lock(ident_t *loc, kmp_int32 gtid,
4365 void **user_lock);
4366KMP_EXPORT void __kmpc_unset_nest_lock(ident_t *loc, kmp_int32 gtid,
4367 void **user_lock);
4368KMP_EXPORT int __kmpc_test_lock(ident_t *loc, kmp_int32 gtid, void **user_lock);
4369KMP_EXPORT int __kmpc_test_nest_lock(ident_t *loc, kmp_int32 gtid,
4370 void **user_lock);
4371
4372KMP_EXPORT void __kmpc_init_lock_with_hint(ident_t *loc, kmp_int32 gtid,
4373 void **user_lock, uintptr_t hint);
4374KMP_EXPORT void __kmpc_init_nest_lock_with_hint(ident_t *loc, kmp_int32 gtid,
4375 void **user_lock,
4376 uintptr_t hint);
4377
4378#if OMPX_TASKGRAPH
4379// Taskgraph's Record & Replay mechanism
4380// __kmp_tdg_is_recording: check whether a given TDG is recording
4381// status: the tdg's current status
4382static inline bool __kmp_tdg_is_recording(kmp_tdg_status_t status) {
4383 return status == KMP_TDG_RECORDING;
4384}
4385
4386KMP_EXPORT kmp_int32 __kmpc_start_record_task(ident_t *loc, kmp_int32 gtid,
4387 kmp_int32 input_flags,
4388 kmp_int32 tdg_id);
4389KMP_EXPORT void __kmpc_end_record_task(ident_t *loc, kmp_int32 gtid,
4390 kmp_int32 input_flags, kmp_int32 tdg_id);
4391#endif
4392/* Interface to fast scalable reduce methods routines */
4393
4394KMP_EXPORT kmp_int32 __kmpc_reduce_nowait(
4395 ident_t *loc, kmp_int32 global_tid, kmp_int32 num_vars, size_t reduce_size,
4396 void *reduce_data, void (*reduce_func)(void *lhs_data, void *rhs_data),
4397 kmp_critical_name *lck);
4398KMP_EXPORT void __kmpc_end_reduce_nowait(ident_t *loc, kmp_int32 global_tid,
4399 kmp_critical_name *lck);
4400KMP_EXPORT kmp_int32 __kmpc_reduce(
4401 ident_t *loc, kmp_int32 global_tid, kmp_int32 num_vars, size_t reduce_size,
4402 void *reduce_data, void (*reduce_func)(void *lhs_data, void *rhs_data),
4403 kmp_critical_name *lck);
4404KMP_EXPORT void __kmpc_end_reduce(ident_t *loc, kmp_int32 global_tid,
4405 kmp_critical_name *lck);
4406
4407/* Internal fast reduction routines */
4408
4409extern PACKED_REDUCTION_METHOD_T __kmp_determine_reduction_method(
4410 ident_t *loc, kmp_int32 global_tid, kmp_int32 num_vars, size_t reduce_size,
4411 void *reduce_data, void (*reduce_func)(void *lhs_data, void *rhs_data),
4412 kmp_critical_name *lck);
4413
4414// this function is for testing set/get/determine reduce method
4415KMP_EXPORT kmp_int32 __kmp_get_reduce_method(void);
4416
4417KMP_EXPORT kmp_uint64 __kmpc_get_taskid();
4418KMP_EXPORT kmp_uint64 __kmpc_get_parent_taskid();
4419
4420// C++ port
4421// missing 'extern "C"' declarations
4422
4423KMP_EXPORT kmp_int32 __kmpc_in_parallel(ident_t *loc);
4424KMP_EXPORT void __kmpc_pop_num_threads(ident_t *loc, kmp_int32 global_tid);
4425KMP_EXPORT void __kmpc_push_num_threads(ident_t *loc, kmp_int32 global_tid,
4426 kmp_int32 num_threads);
4427KMP_EXPORT void __kmpc_push_num_threads_strict(ident_t *loc,
4428 kmp_int32 global_tid,
4429 kmp_int32 num_threads,
4430 int severity,
4431 const char *message);
4432
4433KMP_EXPORT void __kmpc_push_num_threads_list(ident_t *loc, kmp_int32 global_tid,
4434 kmp_uint32 list_length,
4435 kmp_int32 *num_threads_list);
4436KMP_EXPORT void __kmpc_push_num_threads_list_strict(
4437 ident_t *loc, kmp_int32 global_tid, kmp_uint32 list_length,
4438 kmp_int32 *num_threads_list, int severity, const char *message);
4439
4440KMP_EXPORT void __kmpc_push_proc_bind(ident_t *loc, kmp_int32 global_tid,
4441 int proc_bind);
4442KMP_EXPORT void __kmpc_push_num_teams(ident_t *loc, kmp_int32 global_tid,
4443 kmp_int32 num_teams,
4444 kmp_int32 num_threads);
4445KMP_EXPORT void __kmpc_set_thread_limit(ident_t *loc, kmp_int32 global_tid,
4446 kmp_int32 thread_limit);
4447/* Function for OpenMP 5.1 num_teams clause */
4448KMP_EXPORT void __kmpc_push_num_teams_51(ident_t *loc, kmp_int32 global_tid,
4449 kmp_int32 num_teams_lb,
4450 kmp_int32 num_teams_ub,
4451 kmp_int32 num_threads);
4452KMP_EXPORT void __kmpc_fork_teams(ident_t *loc, kmp_int32 argc,
4453 kmpc_micro microtask, ...);
4454struct kmp_dim { // loop bounds info casted to kmp_int64
4455 kmp_int64 lo; // lower
4456 kmp_int64 up; // upper
4457 kmp_int64 st; // stride
4458};
4459KMP_EXPORT void __kmpc_doacross_init(ident_t *loc, kmp_int32 gtid,
4460 kmp_int32 num_dims,
4461 const struct kmp_dim *dims);
4462KMP_EXPORT void __kmpc_doacross_wait(ident_t *loc, kmp_int32 gtid,
4463 const kmp_int64 *vec);
4464KMP_EXPORT void __kmpc_doacross_post(ident_t *loc, kmp_int32 gtid,
4465 const kmp_int64 *vec);
4466KMP_EXPORT void __kmpc_doacross_fini(ident_t *loc, kmp_int32 gtid);
4467
4468KMP_EXPORT void *__kmpc_threadprivate_cached(ident_t *loc, kmp_int32 global_tid,
4469 void *data, size_t size,
4470 void ***cache);
4471
4472// The routines below are not exported.
4473// Consider making them 'static' in corresponding source files.
4474void kmp_threadprivate_insert_private_data(int gtid, void *pc_addr,
4475 void *data_addr, size_t pc_size);
4476struct private_common *kmp_threadprivate_insert(int gtid, void *pc_addr,
4477 void *data_addr,
4478 size_t pc_size);
4479void __kmp_threadprivate_resize_cache(int newCapacity);
4480void __kmp_cleanup_threadprivate_caches();
4481
4482// ompc_, kmpc_ entries moved from omp.h.
4483#if KMP_OS_WINDOWS
4484#define KMPC_CONVENTION __cdecl
4485#else
4486#define KMPC_CONVENTION
4487#endif
4488
4489#ifndef __OMP_H
4490typedef enum omp_sched_t {
4491 omp_sched_static = 1,
4492 omp_sched_dynamic = 2,
4493 omp_sched_guided = 3,
4494 omp_sched_auto = 4
4495} omp_sched_t;
4496typedef void *kmp_affinity_mask_t;
4497#endif
4498
4499KMP_EXPORT void KMPC_CONVENTION ompc_set_max_active_levels(int);
4500KMP_EXPORT void KMPC_CONVENTION ompc_set_schedule(omp_sched_t, int);
4501KMP_EXPORT int KMPC_CONVENTION ompc_get_ancestor_thread_num(int);
4502KMP_EXPORT int KMPC_CONVENTION ompc_get_team_size(int);
4503KMP_EXPORT int KMPC_CONVENTION
4504kmpc_set_affinity_mask_proc(int, kmp_affinity_mask_t *);
4505KMP_EXPORT int KMPC_CONVENTION
4506kmpc_unset_affinity_mask_proc(int, kmp_affinity_mask_t *);
4507KMP_EXPORT int KMPC_CONVENTION
4508kmpc_get_affinity_mask_proc(int, kmp_affinity_mask_t *);
4509
4510KMP_EXPORT void KMPC_CONVENTION kmpc_set_stacksize(int);
4511KMP_EXPORT void KMPC_CONVENTION kmpc_set_stacksize_s(size_t);
4512KMP_EXPORT void KMPC_CONVENTION kmpc_set_library(int);
4513KMP_EXPORT void KMPC_CONVENTION kmpc_set_defaults(char const *);
4514KMP_EXPORT void KMPC_CONVENTION kmpc_set_disp_num_buffers(int);
4515void KMP_EXPAND_NAME(ompc_set_affinity_format)(char const *format);
4516size_t KMP_EXPAND_NAME(ompc_get_affinity_format)(char *buffer, size_t size);
4517void KMP_EXPAND_NAME(ompc_display_affinity)(char const *format);
4518size_t KMP_EXPAND_NAME(ompc_capture_affinity)(char *buffer, size_t buf_size,
4519 char const *format);
4520
4521enum kmp_target_offload_kind {
4522 tgt_disabled = 0,
4523 tgt_default = 1,
4524 tgt_mandatory = 2
4525};
4526typedef enum kmp_target_offload_kind kmp_target_offload_kind_t;
4527// Set via OMP_TARGET_OFFLOAD if specified, defaults to tgt_default otherwise
4528extern kmp_target_offload_kind_t __kmp_target_offload;
4529extern int __kmpc_get_target_offload();
4530
4531// Constants used in libomptarget
4532#define KMP_DEVICE_DEFAULT -1 // This is libomptarget's default device.
4533#define KMP_DEVICE_ALL -11 // This is libomptarget's "all devices".
4534
4535// OMP Pause Resource
4536
4537// The following enum is used both to set the status in __kmp_pause_status, and
4538// as the internal equivalent of the externally-visible omp_pause_resource_t.
4539typedef enum kmp_pause_status_t {
4540 kmp_not_paused = 0, // status is not paused, or, requesting resume
4541 kmp_soft_paused = 1, // status is soft-paused, or, requesting soft pause
4542 kmp_hard_paused = 2, // status is hard-paused, or, requesting hard pause
4543 kmp_stop_tool_paused = 3 // requesting stop_tool pause
4544} kmp_pause_status_t;
4545
4546// This stores the pause state of the runtime
4547extern kmp_pause_status_t __kmp_pause_status;
4548extern int __kmpc_pause_resource(kmp_pause_status_t level);
4549extern int __kmp_pause_resource(kmp_pause_status_t level);
4550// Soft resume sets __kmp_pause_status, and wakes up all threads.
4551extern void __kmp_resume_if_soft_paused();
4552// Hard resume simply resets the status to not paused. Library will appear to
4553// be uninitialized after hard pause. Let OMP constructs trigger required
4554// initializations.
4555static inline void __kmp_resume_if_hard_paused() {
4556 if (__kmp_pause_status == kmp_hard_paused) {
4557 __kmp_pause_status = kmp_not_paused;
4558 }
4559}
4560
4561extern void __kmp_omp_display_env(int verbose);
4562
4563// 1: it is initializing hidden helper team
4564extern volatile int __kmp_init_hidden_helper;
4565// 1: the hidden helper team is done
4566extern volatile int __kmp_hidden_helper_team_done;
4567// 1: enable hidden helper task
4568extern kmp_int32 __kmp_enable_hidden_helper;
4569// Main thread of hidden helper team
4570extern kmp_info_t *__kmp_hidden_helper_main_thread;
4571// Descriptors for the hidden helper threads
4572extern kmp_info_t **__kmp_hidden_helper_threads;
4573// Number of hidden helper threads
4574extern kmp_int32 __kmp_hidden_helper_threads_num;
4575// Number of hidden helper tasks that have not been executed yet
4576extern std::atomic<kmp_int32> __kmp_unexecuted_hidden_helper_tasks;
4577
4578extern void __kmp_hidden_helper_initialize();
4579extern void __kmp_hidden_helper_threads_initz_routine();
4580extern void __kmp_do_initialize_hidden_helper_threads();
4581extern void __kmp_hidden_helper_threads_initz_wait();
4582extern void __kmp_hidden_helper_initz_release();
4583extern void __kmp_hidden_helper_threads_deinitz_wait();
4584extern void __kmp_hidden_helper_threads_deinitz_release();
4585extern void __kmp_hidden_helper_main_thread_wait();
4586extern void __kmp_hidden_helper_worker_thread_wait();
4587extern void __kmp_hidden_helper_worker_thread_signal();
4588extern void __kmp_hidden_helper_main_thread_release();
4589
4590// Check whether a given thread is a hidden helper thread
4591#define KMP_HIDDEN_HELPER_THREAD(gtid) \
4592 ((gtid) >= 1 && (gtid) <= __kmp_hidden_helper_threads_num)
4593
4594#define KMP_HIDDEN_HELPER_WORKER_THREAD(gtid) \
4595 ((gtid) > 1 && (gtid) <= __kmp_hidden_helper_threads_num)
4596
4597#define KMP_HIDDEN_HELPER_MAIN_THREAD(gtid) \
4598 ((gtid) == 1 && (gtid) <= __kmp_hidden_helper_threads_num)
4599
4600#define KMP_HIDDEN_HELPER_TEAM(team) \
4601 (team->t.t_threads[0] == __kmp_hidden_helper_main_thread)
4602
4603// Map a gtid to a hidden helper thread. The first hidden helper thread, a.k.a
4604// main thread, is skipped.
4605#define KMP_GTID_TO_SHADOW_GTID(gtid) \
4606 ((gtid) % (__kmp_hidden_helper_threads_num - 1) + 2)
4607
4608// Return the adjusted gtid value by subtracting from gtid the number
4609// of hidden helper threads. This adjusted value is the gtid the thread would
4610// have received if there were no hidden helper threads.
4611static inline int __kmp_adjust_gtid_for_hidden_helpers(int gtid) {
4612 int adjusted_gtid = gtid;
4613 if (__kmp_hidden_helper_threads_num > 0 && gtid > 0 &&
4614 gtid - __kmp_hidden_helper_threads_num >= 0) {
4615 adjusted_gtid -= __kmp_hidden_helper_threads_num;
4616 }
4617 return adjusted_gtid;
4618}
4619
4620// Support for error directive
4621typedef enum kmp_severity_t {
4622 severity_warning = 1,
4623 severity_fatal = 2
4624} kmp_severity_t;
4625extern void __kmpc_error(ident_t *loc, int severity, const char *message);
4626
4627// Support for scope directive
4628KMP_EXPORT void __kmpc_scope(ident_t *loc, kmp_int32 gtid, void *reserved);
4629KMP_EXPORT void __kmpc_end_scope(ident_t *loc, kmp_int32 gtid, void *reserved);
4630
4631#ifdef __cplusplus
4632}
4633#endif
4634
4635template <bool C, bool S>
4636extern void __kmp_suspend_32(int th_gtid, kmp_flag_32<C, S> *flag);
4637template <bool C, bool S>
4638extern void __kmp_suspend_64(int th_gtid, kmp_flag_64<C, S> *flag);
4639template <bool C, bool S>
4640extern void __kmp_atomic_suspend_64(int th_gtid,
4641 kmp_atomic_flag_64<C, S> *flag);
4642extern void __kmp_suspend_oncore(int th_gtid, kmp_flag_oncore *flag);
4643#if KMP_HAVE_MWAIT || KMP_HAVE_UMWAIT
4644template <bool C, bool S>
4645extern void __kmp_mwait_32(int th_gtid, kmp_flag_32<C, S> *flag);
4646template <bool C, bool S>
4647extern void __kmp_mwait_64(int th_gtid, kmp_flag_64<C, S> *flag);
4648template <bool C, bool S>
4649extern void __kmp_atomic_mwait_64(int th_gtid, kmp_atomic_flag_64<C, S> *flag);
4650extern void __kmp_mwait_oncore(int th_gtid, kmp_flag_oncore *flag);
4651#endif
4652template <bool C, bool S>
4653extern void __kmp_resume_32(int target_gtid, kmp_flag_32<C, S> *flag);
4654template <bool C, bool S>
4655extern void __kmp_resume_64(int target_gtid, kmp_flag_64<C, S> *flag);
4656template <bool C, bool S>
4657extern void __kmp_atomic_resume_64(int target_gtid,
4658 kmp_atomic_flag_64<C, S> *flag);
4659extern void __kmp_resume_oncore(int target_gtid, kmp_flag_oncore *flag);
4660
4661template <bool C, bool S>
4662int __kmp_execute_tasks_32(kmp_info_t *thread, kmp_int32 gtid,
4663 kmp_flag_32<C, S> *flag, int final_spin,
4664 int *thread_finished,
4665#if USE_ITT_BUILD
4666 void *itt_sync_obj,
4667#endif /* USE_ITT_BUILD */
4668 kmp_int32 is_constrained);
4669template <bool C, bool S>
4670int __kmp_execute_tasks_64(kmp_info_t *thread, kmp_int32 gtid,
4671 kmp_flag_64<C, S> *flag, int final_spin,
4672 int *thread_finished,
4673#if USE_ITT_BUILD
4674 void *itt_sync_obj,
4675#endif /* USE_ITT_BUILD */
4676 kmp_int32 is_constrained);
4677template <bool C, bool S>
4678int __kmp_atomic_execute_tasks_64(kmp_info_t *thread, kmp_int32 gtid,
4679 kmp_atomic_flag_64<C, S> *flag,
4680 int final_spin, int *thread_finished,
4681#if USE_ITT_BUILD
4682 void *itt_sync_obj,
4683#endif /* USE_ITT_BUILD */
4684 kmp_int32 is_constrained);
4685int __kmp_execute_tasks_oncore(kmp_info_t *thread, kmp_int32 gtid,
4686 kmp_flag_oncore *flag, int final_spin,
4687 int *thread_finished,
4688#if USE_ITT_BUILD
4689 void *itt_sync_obj,
4690#endif /* USE_ITT_BUILD */
4691 kmp_int32 is_constrained);
4692
4693extern int __kmp_nesting_mode;
4694extern int __kmp_nesting_mode_nlevels;
4695extern int *__kmp_nesting_nth_level;
4696extern void __kmp_init_nesting_mode();
4697extern void __kmp_set_nesting_mode_threads();
4698
4706 FILE *f;
4707
4708 void close() {
4709 if (f && f != stdout && f != stderr) {
4710 fclose(f);
4711 f = nullptr;
4712 }
4713 }
4714
4715public:
4716 kmp_safe_raii_file_t() : f(nullptr) {}
4717 kmp_safe_raii_file_t(const char *filename, const char *mode,
4718 const char *env_var = nullptr)
4719 : f(nullptr) {
4720 open(filename, mode, env_var);
4721 }
4722 kmp_safe_raii_file_t(const kmp_safe_raii_file_t &other) = delete;
4723 kmp_safe_raii_file_t &operator=(const kmp_safe_raii_file_t &other) = delete;
4724 ~kmp_safe_raii_file_t() { close(); }
4725
4729 void open(const char *filename, const char *mode,
4730 const char *env_var = nullptr) {
4731 KMP_ASSERT(!f);
4732 f = fopen(filename, mode);
4733 if (!f) {
4734 int code = errno;
4735 if (env_var) {
4736 __kmp_fatal(KMP_MSG(CantOpenFileForReading, filename), KMP_ERR(code),
4737 KMP_HNT(CheckEnvVar, env_var, filename), __kmp_msg_null);
4738 } else {
4739 __kmp_fatal(KMP_MSG(CantOpenFileForReading, filename), KMP_ERR(code),
4740 __kmp_msg_null);
4741 }
4742 }
4743 }
4746 int try_open(const char *filename, const char *mode) {
4747 KMP_ASSERT(!f);
4748 f = fopen(filename, mode);
4749 if (!f)
4750 return errno;
4751 return 0;
4752 }
4755 void set_stdout() {
4756 KMP_ASSERT(!f);
4757 f = stdout;
4758 }
4761 void set_stderr() {
4762 KMP_ASSERT(!f);
4763 f = stderr;
4764 }
4765 operator bool() { return bool(f); }
4766 operator FILE *() { return f; }
4767};
4768
4769template <typename SourceType, typename TargetType,
4770 bool isSourceSmaller = (sizeof(SourceType) < sizeof(TargetType)),
4771 bool isSourceEqual = (sizeof(SourceType) == sizeof(TargetType)),
4772 bool isSourceSigned = std::is_signed<SourceType>::value,
4773 bool isTargetSigned = std::is_signed<TargetType>::value>
4774struct kmp_convert {};
4775
4776// Both types are signed; Source smaller
4777template <typename SourceType, typename TargetType>
4778struct kmp_convert<SourceType, TargetType, true, false, true, true> {
4779 static TargetType to(SourceType src) { return (TargetType)src; }
4780};
4781// Source equal
4782template <typename SourceType, typename TargetType>
4783struct kmp_convert<SourceType, TargetType, false, true, true, true> {
4784 static TargetType to(SourceType src) { return src; }
4785};
4786// Source bigger
4787template <typename SourceType, typename TargetType>
4788struct kmp_convert<SourceType, TargetType, false, false, true, true> {
4789 static TargetType to(SourceType src) {
4790 KMP_ASSERT(src <= static_cast<SourceType>(
4791 (std::numeric_limits<TargetType>::max)()));
4792 KMP_ASSERT(src >= static_cast<SourceType>(
4793 (std::numeric_limits<TargetType>::min)()));
4794 return (TargetType)src;
4795 }
4796};
4797
4798// Source signed, Target unsigned
4799// Source smaller
4800template <typename SourceType, typename TargetType>
4801struct kmp_convert<SourceType, TargetType, true, false, true, false> {
4802 static TargetType to(SourceType src) {
4803 KMP_ASSERT(src >= 0);
4804 return (TargetType)src;
4805 }
4806};
4807// Source equal
4808template <typename SourceType, typename TargetType>
4809struct kmp_convert<SourceType, TargetType, false, true, true, false> {
4810 static TargetType to(SourceType src) {
4811 KMP_ASSERT(src >= 0);
4812 return (TargetType)src;
4813 }
4814};
4815// Source bigger
4816template <typename SourceType, typename TargetType>
4817struct kmp_convert<SourceType, TargetType, false, false, true, false> {
4818 static TargetType to(SourceType src) {
4819 KMP_ASSERT(src >= 0);
4820 KMP_ASSERT(src <= static_cast<SourceType>(
4821 (std::numeric_limits<TargetType>::max)()));
4822 return (TargetType)src;
4823 }
4824};
4825
4826// Source unsigned, Target signed
4827// Source smaller
4828template <typename SourceType, typename TargetType>
4829struct kmp_convert<SourceType, TargetType, true, false, false, true> {
4830 static TargetType to(SourceType src) { return (TargetType)src; }
4831};
4832// Source equal
4833template <typename SourceType, typename TargetType>
4834struct kmp_convert<SourceType, TargetType, false, true, false, true> {
4835 static TargetType to(SourceType src) {
4836 KMP_ASSERT(src <= static_cast<SourceType>(
4837 (std::numeric_limits<TargetType>::max)()));
4838 return (TargetType)src;
4839 }
4840};
4841// Source bigger
4842template <typename SourceType, typename TargetType>
4843struct kmp_convert<SourceType, TargetType, false, false, false, true> {
4844 static TargetType to(SourceType src) {
4845 KMP_ASSERT(src <= static_cast<SourceType>(
4846 (std::numeric_limits<TargetType>::max)()));
4847 return (TargetType)src;
4848 }
4849};
4850
4851// Source unsigned, Target unsigned
4852// Source smaller
4853template <typename SourceType, typename TargetType>
4854struct kmp_convert<SourceType, TargetType, true, false, false, false> {
4855 static TargetType to(SourceType src) { return (TargetType)src; }
4856};
4857// Source equal
4858template <typename SourceType, typename TargetType>
4859struct kmp_convert<SourceType, TargetType, false, true, false, false> {
4860 static TargetType to(SourceType src) { return src; }
4861};
4862// Source bigger
4863template <typename SourceType, typename TargetType>
4864struct kmp_convert<SourceType, TargetType, false, false, false, false> {
4865 static TargetType to(SourceType src) {
4866 KMP_ASSERT(src <= static_cast<SourceType>(
4867 (std::numeric_limits<TargetType>::max)()));
4868 return (TargetType)src;
4869 }
4870};
4871
4872template <typename T1, typename T2>
4873static inline void __kmp_type_convert(T1 src, T2 *dest) {
4874 *dest = kmp_convert<T1, T2>::to(src);
4875}
4876
4877#endif /* KMP_H */
void set_stdout()
Definition kmp.h:4755
void set_stderr()
Definition kmp.h:4761
int try_open(const char *filename, const char *mode)
Definition kmp.h:4746
void open(const char *filename, const char *mode, const char *env_var=nullptr)
Definition kmp.h:4729
struct ident ident_t
@ KMP_IDENT_KMPC
Definition kmp.h:209
@ KMP_IDENT_IMB
Definition kmp.h:207
@ KMP_IDENT_WORK_LOOP
Definition kmp.h:227
@ KMP_IDENT_BARRIER_IMPL
Definition kmp.h:218
@ KMP_IDENT_WORK_SECTIONS
Definition kmp.h:229
@ KMP_IDENT_AUTOPAR
Definition kmp.h:212
@ KMP_IDENT_ATOMIC_HINT_MASK
Definition kmp.h:236
@ KMP_IDENT_WORK_DISTRIBUTE
Definition kmp.h:231
@ KMP_IDENT_BARRIER_EXPL
Definition kmp.h:216
@ KMP_IDENT_ATOMIC_REDUCE
Definition kmp.h:214
KMP_EXPORT kmp_int32 __kmpc_ok_to_fork(ident_t *)
KMP_EXPORT void __kmpc_fork_teams(ident_t *loc, kmp_int32 argc, kmpc_micro microtask,...)
KMP_EXPORT void __kmpc_fork_call_if(ident_t *loc, kmp_int32 nargs, kmpc_micro microtask, kmp_int32 cond, void *args)
KMP_EXPORT void __kmpc_push_num_threads(ident_t *loc, kmp_int32 global_tid, kmp_int32 num_threads)
KMP_EXPORT void __kmpc_set_thread_limit(ident_t *loc, kmp_int32 global_tid, kmp_int32 thread_limit)
KMP_EXPORT void __kmpc_serialized_parallel(ident_t *, kmp_int32 global_tid)
KMP_EXPORT void __kmpc_push_num_threads_list(ident_t *loc, kmp_int32 global_tid, kmp_uint32 list_length, kmp_int32 *num_threads_list)
KMP_EXPORT void __kmpc_push_num_teams(ident_t *loc, kmp_int32 global_tid, kmp_int32 num_teams, kmp_int32 num_threads)
KMP_EXPORT void __kmpc_fork_call(ident_t *, kmp_int32 nargs, kmpc_micro microtask,...)
KMP_EXPORT void __kmpc_end_serialized_parallel(ident_t *, kmp_int32 global_tid)
void(* kmpc_micro)(kmp_int32 *global_tid, kmp_int32 *bound_tid,...)
Definition kmp.h:1745
KMP_EXPORT void __kmpc_push_num_teams_51(ident_t *loc, kmp_int32 global_tid, kmp_int32 num_teams_lb, kmp_int32 num_teams_ub, kmp_int32 num_threads)
KMP_EXPORT void __kmpc_begin(ident_t *, kmp_int32 flags)
KMP_EXPORT void __kmpc_end(ident_t *)
KMP_EXPORT void __kmpc_end_reduce(ident_t *loc, kmp_int32 global_tid, kmp_critical_name *lck)
KMP_EXPORT void __kmpc_end_barrier_master(ident_t *, kmp_int32 global_tid)
KMP_EXPORT kmp_int32 __kmpc_barrier_master_nowait(ident_t *, kmp_int32 global_tid)
KMP_EXPORT void __kmpc_end_reduce_nowait(ident_t *loc, kmp_int32 global_tid, kmp_critical_name *lck)
KMP_EXPORT kmp_int32 __kmpc_reduce(ident_t *loc, kmp_int32 global_tid, kmp_int32 num_vars, size_t reduce_size, void *reduce_data, void(*reduce_func)(void *lhs_data, void *rhs_data), kmp_critical_name *lck)
KMP_EXPORT void __kmpc_barrier(ident_t *, kmp_int32 global_tid)
KMP_EXPORT void __kmpc_flush(ident_t *)
KMP_EXPORT kmp_int32 __kmpc_barrier_master(ident_t *, kmp_int32 global_tid)
KMP_EXPORT kmp_int32 __kmpc_reduce_nowait(ident_t *loc, kmp_int32 global_tid, kmp_int32 num_vars, size_t reduce_size, void *reduce_data, void(*reduce_func)(void *lhs_data, void *rhs_data), kmp_critical_name *lck)
KMP_EXPORT void * __kmpc_task_reduction_get_th_data(int gtid, void *tg, void *d)
KMP_EXPORT void * __kmpc_task_reduction_modifier_init(ident_t *loc, int gtid, int is_ws, int num, void *data)
KMP_EXPORT void * __kmpc_taskred_modifier_init(ident_t *loc, int gtid, int is_ws, int num, void *data)
KMP_EXPORT bool __kmpc_omp_has_task_team(kmp_int32 gtid)
KMP_EXPORT void __kmpc_proxy_task_completed_ooo(kmp_task_t *ptask)
KMP_EXPORT void __kmpc_task_reduction_modifier_fini(ident_t *loc, int gtid, int is_ws)
KMP_EXPORT kmp_int32 __kmpc_omp_task_with_deps(ident_t *loc_ref, kmp_int32 gtid, kmp_task_t *new_task, kmp_int32 ndeps, kmp_depend_info_t *dep_list, kmp_int32 ndeps_noalias, kmp_depend_info_t *noalias_dep_list)
KMP_EXPORT kmp_int32 __kmpc_omp_reg_task_with_affinity(ident_t *loc_ref, kmp_int32 gtid, kmp_task_t *new_task, kmp_int32 naffins, kmp_task_affinity_info_t *affin_list)
KMP_EXPORT void * __kmpc_task_reduction_init(int gtid, int num_data, void *data)
KMP_EXPORT void __kmpc_proxy_task_completed(kmp_int32 gtid, kmp_task_t *ptask)
KMP_EXPORT void __kmpc_omp_wait_deps(ident_t *loc_ref, kmp_int32 gtid, kmp_int32 ndeps, kmp_depend_info_t *dep_list, kmp_int32 ndeps_noalias, kmp_depend_info_t *noalias_dep_list)
KMP_EXPORT void * __kmpc_taskred_init(int gtid, int num_data, void *data)
KMP_EXPORT void ** __kmpc_omp_get_target_async_handle_ptr(kmp_int32 gtid)
void(* kmpc_dtor)(void *)
Definition kmp.h:1769
void *(* kmpc_cctor)(void *, void *)
Definition kmp.h:1776
KMP_EXPORT void __kmpc_threadprivate_register(ident_t *, void *data, kmpc_ctor ctor, kmpc_cctor cctor, kmpc_dtor dtor)
KMP_EXPORT void __kmpc_copyprivate(ident_t *loc, kmp_int32 global_tid, size_t cpy_size, void *cpy_data, void(*cpy_func)(void *, void *), kmp_int32 didit)
void *(* kmpc_cctor_vec)(void *, void *, size_t)
Definition kmp.h:1798
void *(* kmpc_ctor)(void *)
Definition kmp.h:1763
KMP_EXPORT void * __kmpc_copyprivate_light(ident_t *loc, kmp_int32 gtid, void *cpy_data)
void *(* kmpc_ctor_vec)(void *, size_t)
Definition kmp.h:1786
KMP_EXPORT void * __kmpc_threadprivate_cached(ident_t *loc, kmp_int32 global_tid, void *data, size_t size, void ***cache)
void(* kmpc_dtor_vec)(void *, size_t)
Definition kmp.h:1792
KMP_EXPORT void __kmpc_threadprivate_register_vec(ident_t *, void *data, kmpc_ctor_vec ctor, kmpc_cctor_vec cctor, kmpc_dtor_vec dtor, size_t vector_length)
KMP_EXPORT kmp_int32 __kmpc_global_num_threads(ident_t *)
KMP_EXPORT kmp_int32 __kmpc_global_thread_num(ident_t *)
KMP_EXPORT kmp_int32 __kmpc_in_parallel(ident_t *loc)
KMP_EXPORT kmp_int32 __kmpc_bound_thread_num(ident_t *)
KMP_EXPORT kmp_int32 __kmpc_bound_num_threads(ident_t *)
KMP_EXPORT void __kmpc_end_ordered(ident_t *, kmp_int32 global_tid)
KMP_EXPORT void __kmpc_end_critical(ident_t *, kmp_int32 global_tid, kmp_critical_name *)
KMP_EXPORT void __kmpc_for_static_fini(ident_t *loc, kmp_int32 global_tid)
sched_type
Definition kmp.h:370
KMP_EXPORT void __kmpc_end_masked(ident_t *, kmp_int32 global_tid)
KMP_EXPORT kmp_int32 __kmpc_master(ident_t *, kmp_int32 global_tid)
KMP_EXPORT kmp_int32 __kmpc_single(ident_t *, kmp_int32 global_tid)
KMP_EXPORT kmp_int32 __kmpc_next_section(ident_t *loc, kmp_int32 global_tid, kmp_int32 numberOfSections)
KMP_EXPORT void __kmpc_end_master(ident_t *, kmp_int32 global_tid)
KMP_EXPORT void __kmpc_end_sections(ident_t *loc, kmp_int32 global_tid)
KMP_EXPORT void __kmpc_end_single(ident_t *, kmp_int32 global_tid)
KMP_EXPORT kmp_int32 __kmpc_sections_init(ident_t *loc, kmp_int32 global_tid)
KMP_EXPORT void __kmpc_ordered(ident_t *, kmp_int32 global_tid)
KMP_EXPORT kmp_int32 __kmpc_masked(ident_t *, kmp_int32 global_tid, kmp_int32 filter)
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)
KMP_EXPORT void __kmpc_critical(ident_t *, kmp_int32 global_tid, kmp_critical_name *)
@ kmp_nm_guided_chunked
Definition kmp.h:421
@ kmp_sch_runtime_simd
Definition kmp.h:392
@ kmp_nm_ord_auto
Definition kmp.h:440
@ kmp_sch_auto
Definition kmp.h:377
@ kmp_nm_auto
Definition kmp.h:423
@ kmp_distribute_static_chunked
Definition kmp.h:408
@ kmp_sch_static
Definition kmp.h:373
@ kmp_sch_guided_simd
Definition kmp.h:391
@ kmp_sch_modifier_monotonic
Definition kmp.h:458
@ kmp_sch_default
Definition kmp.h:478
@ kmp_sch_modifier_nonmonotonic
Definition kmp.h:460
@ kmp_nm_ord_static
Definition kmp.h:436
@ kmp_distribute_static
Definition kmp.h:409
@ kmp_sch_guided_chunked
Definition kmp.h:375
@ kmp_nm_static
Definition kmp.h:419
@ kmp_sch_lower
Definition kmp.h:371
@ kmp_nm_upper
Definition kmp.h:442
@ kmp_ord_lower
Definition kmp.h:397
@ kmp_ord_static
Definition kmp.h:399
@ kmp_sch_upper
Definition kmp.h:395
@ kmp_ord_upper
Definition kmp.h:405
@ kmp_nm_lower
Definition kmp.h:415
@ kmp_ord_auto
Definition kmp.h:403
Definition kmp.h:247
kmp_int32 reserved_1
Definition kmp.h:248
char const * psource
Definition kmp.h:257
kmp_int32 reserved_2
Definition kmp.h:251
kmp_int32 reserved_3
Definition kmp.h:256
kmp_int32 flags
Definition kmp.h:249