Intel® OpenMP* Runtime Library
 All Classes Functions Variables Typedefs Enumerations Enumerator Groups Pages
kmp_global.c
1 /*
2  * kmp_global.c -- KPTS global variables for runtime support library
3  * $Revision: 42431 $
4  * $Date: 2013-06-11 12:28:25 -0500 (Tue, 11 Jun 2013) $
5  */
6 
7 /* <copyright>
8  Copyright (c) 1997-2013 Intel Corporation. All Rights Reserved.
9 
10  Redistribution and use in source and binary forms, with or without
11  modification, are permitted provided that the following conditions
12  are met:
13 
14  * Redistributions of source code must retain the above copyright
15  notice, this list of conditions and the following disclaimer.
16  * Redistributions in binary form must reproduce the above copyright
17  notice, this list of conditions and the following disclaimer in the
18  documentation and/or other materials provided with the distribution.
19  * Neither the name of Intel Corporation nor the names of its
20  contributors may be used to endorse or promote products derived
21  from this software without specific prior written permission.
22 
23  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
26  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
27  HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
28  SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
29  LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
30  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
31  THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
32  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
33  OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 
35 </copyright> */
36 
37 #include "kmp.h"
38 
39 #ifdef KMP_SETVERSION
40 char __kmp_setversion_string[] = VERSION_STRING;
41 #endif
42 
43 kmp_key_t __kmp_gtid_threadprivate_key;
44 
45 kmp_cpuinfo_t __kmp_cpuinfo = { 0 }; // Not initialized
46 kmp_uint64 __kmp_cpu_frequency = 0;
47 
48 
49 /* ----------------------------------------------------- */
50 /* INITIALIZATION VARIABLES */
51 /* they are syncronized to write during init, but read anytime */
52 volatile int __kmp_init_serial = FALSE;
53 volatile int __kmp_init_gtid = FALSE;
54 volatile int __kmp_init_common = FALSE;
55 volatile int __kmp_init_middle = FALSE;
56 volatile int __kmp_init_parallel = FALSE;
57 volatile int __kmp_init_monitor = 0; /* 1 - launched, 2 - actually started (Windows* OS only) */
58 volatile int __kmp_init_user_locks = FALSE;
59 
60 /* list of address of allocated caches for commons */
61 kmp_cached_addr_t *__kmp_threadpriv_cache_list = NULL;
62 
63 int __kmp_init_counter = 0;
64 int __kmp_root_counter = 0;
65 int __kmp_version = 0;
66 
67 volatile kmp_uint32 __kmp_team_counter = 0;
68 volatile kmp_uint32 __kmp_task_counter = 0;
69 
70 unsigned int __kmp_init_wait = KMP_DEFAULT_INIT_WAIT; /* initial number of spin-tests */
71 unsigned int __kmp_next_wait = KMP_DEFAULT_NEXT_WAIT; /* susequent number of spin-tests */
72 
73 size_t __kmp_stksize = KMP_DEFAULT_STKSIZE;
74 size_t __kmp_monitor_stksize = 0; // auto adjust
75 size_t __kmp_stkoffset = KMP_DEFAULT_STKOFFSET;
76 
77 size_t __kmp_malloc_pool_incr = KMP_DEFAULT_MALLOC_POOL_INCR;
78 
79 /* Barrier method defaults, settings, and strings */
80 /* branch factor = 2^branch_bits (only relevant for tree and hyper barrier types) */
81 #if KMP_ARCH_X86_64
82 kmp_uint32 __kmp_barrier_gather_bb_dflt = 2; /* branch_factor = 4 */ /* hyper2: C78980 */
83 kmp_uint32 __kmp_barrier_release_bb_dflt = 2; /* branch_factor = 4 */ /* hyper2: C78980 */
84 #else
85 kmp_uint32 __kmp_barrier_gather_bb_dflt = 2; /* branch_factor = 4 */ /* communication in core for MIC */
86 kmp_uint32 __kmp_barrier_release_bb_dflt = 2; /* branch_factor = 4 */ /* communication in core for MIC */
87 #endif // KMP_ARCH_X86_64
88 #if KMP_ARCH_X86_64
89 kmp_bar_pat_e __kmp_barrier_gather_pat_dflt = bp_hyper_bar; /* hyper2: C78980 */
90 kmp_bar_pat_e __kmp_barrier_release_pat_dflt = bp_hyper_bar; /* hyper2: C78980 */
91 #else
92 kmp_bar_pat_e __kmp_barrier_gather_pat_dflt = bp_linear_bar;
93 kmp_bar_pat_e __kmp_barrier_release_pat_dflt = bp_linear_bar;
94 #endif
95 kmp_uint32 __kmp_barrier_gather_branch_bits [ bs_last_barrier ] = { 0 };
96 kmp_uint32 __kmp_barrier_release_branch_bits [ bs_last_barrier ] = { 0 };
97 kmp_bar_pat_e __kmp_barrier_gather_pattern [ bs_last_barrier ] = { bp_linear_bar };
98 kmp_bar_pat_e __kmp_barrier_release_pattern [ bs_last_barrier ] = { bp_linear_bar };
99 char const *__kmp_barrier_branch_bit_env_name [ bs_last_barrier ] =
100  { "KMP_PLAIN_BARRIER", "KMP_FORKJOIN_BARRIER"
101  #if KMP_FAST_REDUCTION_BARRIER
102  , "KMP_REDUCTION_BARRIER"
103  #endif // KMP_FAST_REDUCTION_BARRIER
104  };
105 char const *__kmp_barrier_pattern_env_name [ bs_last_barrier ] =
106  { "KMP_PLAIN_BARRIER_PATTERN", "KMP_FORKJOIN_BARRIER_PATTERN"
107  #if KMP_FAST_REDUCTION_BARRIER
108  , "KMP_REDUCTION_BARRIER_PATTERN"
109  #endif // KMP_FAST_REDUCTION_BARRIER
110  };
111 char const *__kmp_barrier_type_name [ bs_last_barrier ] =
112  { "plain", "forkjoin"
113  #if KMP_FAST_REDUCTION_BARRIER
114  , "reduction"
115  #endif // KMP_FAST_REDUCTION_BARRIER
116  };
117 char const *__kmp_barrier_pattern_name [ bp_last_bar ] = { "linear", "tree", "hyper" };
118 
119 
120 int __kmp_allThreadsSpecified = 0;
121 size_t __kmp_align_alloc = CACHE_LINE;
122 
123 
124 int __kmp_generate_warnings = kmp_warnings_low;
125 int __kmp_reserve_warn = 0;
126 int __kmp_xproc = 0;
127 int __kmp_avail_proc = 0;
128 int __kmp_sys_min_stksize = KMP_MIN_STKSIZE;
129 int __kmp_sys_max_nth = KMP_MAX_NTH;
130 int __kmp_max_nth = 0;
131 int __kmp_threads_capacity = 0;
132 int __kmp_dflt_team_nth = 0;
133 int __kmp_dflt_team_nth_ub = 0;
134 int __kmp_tp_capacity = 0;
135 int __kmp_tp_cached = 0;
136 int __kmp_dflt_nested = FALSE;
137 #if OMP_30_ENABLED
138 int __kmp_dflt_max_active_levels = KMP_MAX_ACTIVE_LEVELS_LIMIT; /* max_active_levels limit */
139 #endif // OMP_30_ENABLED
140 enum library_type __kmp_library = library_none;
141 enum sched_type __kmp_sched = kmp_sch_default; /* scheduling method for runtime scheduling */
142 enum sched_type __kmp_static = kmp_sch_static_greedy; /* default static scheduling method */
143 enum sched_type __kmp_guided = kmp_sch_guided_iterative_chunked; /* default guided scheduling method */
144 #if OMP_30_ENABLED
145 enum sched_type __kmp_auto = kmp_sch_guided_analytical_chunked; /* default auto scheduling method */
146 #endif // OMP_30_ENABLED
147 int __kmp_dflt_blocktime = KMP_DEFAULT_BLOCKTIME;
148 int __kmp_monitor_wakeups = KMP_MIN_MONITOR_WAKEUPS;
149 int __kmp_bt_intervals = KMP_INTERVALS_FROM_BLOCKTIME( KMP_DEFAULT_BLOCKTIME, KMP_MIN_MONITOR_WAKEUPS );
150 #ifdef KMP_ADJUST_BLOCKTIME
151 int __kmp_zero_bt = FALSE;
152 #endif /* KMP_ADJUST_BLOCKTIME */
153 int __kmp_ht_capable = FALSE;
154 int __kmp_ht_enabled = FALSE;
155 int __kmp_ht_log_per_phy = 1;
156 int __kmp_ncores = 0;
157 int __kmp_chunk = 0;
158 int __kmp_abort_delay = 0;
159 #if KMP_OS_LINUX && defined(KMP_TDATA_GTID)
160 int __kmp_gtid_mode = 3; /* use __declspec(thread) TLS to store gtid */
161 int __kmp_adjust_gtid_mode = FALSE;
162 #elif KMP_OS_WINDOWS
163 int __kmp_gtid_mode = 2; /* use TLS functions to store gtid */
164 int __kmp_adjust_gtid_mode = FALSE;
165 #else
166 int __kmp_gtid_mode = 0; /* select method to get gtid based on #threads */
167 int __kmp_adjust_gtid_mode = TRUE;
168 #endif /* KMP_OS_LINUX && defined(KMP_TDATA_GTID) */
169 #ifdef KMP_TDATA_GTID
170 #if KMP_OS_WINDOWS
171 __declspec(thread) int __kmp_gtid = KMP_GTID_DNE;
172 #else
173 __thread int __kmp_gtid = KMP_GTID_DNE;
174 #endif /* KMP_OS_WINDOWS - workaround because Intel(R) Many Integrated Core compiler 20110316 doesn't accept __declspec */
175 #endif /* KMP_TDATA_GTID */
176 int __kmp_tls_gtid_min = INT_MAX;
177 int __kmp_foreign_tp = TRUE;
178 #if KMP_ARCH_X86 || KMP_ARCH_X86_64
179 int __kmp_inherit_fp_control = TRUE;
180 kmp_int16 __kmp_init_x87_fpu_control_word = 0;
181 kmp_uint32 __kmp_init_mxcsr = 0;
182 #endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */
183 
184 #ifdef USE_LOAD_BALANCE
185 double __kmp_load_balance_interval = 1.0;
186 #endif /* USE_LOAD_BALANCE */
187 
188 kmp_nested_nthreads_t __kmp_nested_nth = { NULL, 0, 0 };
189 
190 #if KMP_USE_ADAPTIVE_LOCKS
191 
192 kmp_adaptive_backoff_params_t __kmp_adaptive_backoff_params = { 1, 1024 }; // TODO: tune it!
193 
194 #if KMP_DEBUG_ADAPTIVE_LOCKS
195 char * __kmp_speculative_statsfile = "-";
196 #endif
197 
198 #endif // KMP_USE_ADAPTIVE_LOCKS
199 
200 /* map OMP 3.0 schedule types with our internal schedule types */
201 enum sched_type __kmp_sch_map[ kmp_sched_upper - kmp_sched_lower_ext + kmp_sched_upper_std - kmp_sched_lower - 2 ] = {
202  kmp_sch_static_chunked, // ==> kmp_sched_static = 1
203  kmp_sch_dynamic_chunked, // ==> kmp_sched_dynamic = 2
204  kmp_sch_guided_chunked, // ==> kmp_sched_guided = 3
205  kmp_sch_auto, // ==> kmp_sched_auto = 4
206  kmp_sch_trapezoidal // ==> kmp_sched_trapezoidal = 101
207  // will likely not used, introduced here just to debug the code
208  // of public intel extension schedules
209 };
210 
211 #if KMP_OS_LINUX
212 enum clock_function_type __kmp_clock_function;
213 int __kmp_clock_function_param;
214 #endif /* KMP_OS_LINUX */
215 
216 #if KMP_OS_LINUX || KMP_OS_WINDOWS
217 
218 # if KMP_OS_WINDOWS && KMP_ARCH_X86_64
219 
220 int __kmp_num_proc_groups = 1;
221 
222 kmp_GetActiveProcessorCount_t __kmp_GetActiveProcessorCount = NULL;
223 kmp_GetActiveProcessorGroupCount_t __kmp_GetActiveProcessorGroupCount = NULL;
224 kmp_GetThreadGroupAffinity_t __kmp_GetThreadGroupAffinity = NULL;
225 kmp_SetThreadGroupAffinity_t __kmp_SetThreadGroupAffinity = NULL;
226 
227 # endif /* KMP_OS_WINDOWS && KMP_ARCH_X86_64 */
228 
229 size_t __kmp_affin_mask_size = 0;
230 enum affinity_type __kmp_affinity_type = affinity_default;
231 enum affinity_gran __kmp_affinity_gran = affinity_gran_default;
232 int __kmp_affinity_gran_levels = -1;
233 int __kmp_affinity_dups = TRUE;
234 enum affinity_top_method __kmp_affinity_top_method = affinity_top_method_default;
235 int __kmp_affinity_compact = 0;
236 int __kmp_affinity_offset = 0;
237 int __kmp_affinity_verbose = FALSE;
238 int __kmp_affinity_warnings = TRUE;
239 int __kmp_affinity_respect_mask = affinity_respect_mask_default;
240 char * __kmp_affinity_proclist = NULL;
241 kmp_affin_mask_t *__kmp_affinity_masks = NULL;
242 unsigned __kmp_affinity_num_masks = 0;
243 
244 char const * __kmp_cpuinfo_file = NULL;
245 
246 #endif /* KMP_OS_LINUX || KMP_OS_WINDOWS */
247 
248 #if OMP_40_ENABLED
249 kmp_nested_proc_bind_t __kmp_nested_proc_bind = { NULL, 0, 0 };
250 int __kmp_affinity_num_places = 0;
251 #endif
252 
253 #if KMP_MIC
254 unsigned int __kmp_place_num_cores = 0;
255 unsigned int __kmp_place_num_threads_per_core = 0;
256 unsigned int __kmp_place_core_offset = 0;
257 #endif
258 
259 #if OMP_30_ENABLED
260 kmp_tasking_mode_t __kmp_tasking_mode = tskm_task_teams;
261 
262 /* This check ensures that the compiler is passing the correct data type
263  * for the flags formal parameter of the function kmpc_omp_task_alloc().
264  * If the type is not a 4-byte type, then give an error message about
265  * a non-positive length array pointing here. If that happens, the
266  * kmp_tasking_flags_t structure must be redefined to have exactly 32 bits.
267  */
268 KMP_BUILD_ASSERT( sizeof(kmp_tasking_flags_t) == 4 );
269 
270 kmp_int32 __kmp_task_stealing_constraint = 1; /* Constrain task stealing by default */
271 
272 #endif /* OMP_30_ENABLED */
273 
274 #ifdef DEBUG_SUSPEND
275 int __kmp_suspend_count = 0;
276 #endif
277 
278 int __kmp_settings = FALSE;
279 int __kmp_duplicate_library_ok = 0;
280 int __kmp_forkjoin_frames = 1;
281 PACKED_REDUCTION_METHOD_T __kmp_force_reduction_method = reduction_method_not_defined;
282 int __kmp_determ_red = FALSE;
283 
284 #ifdef KMP_DEBUG
285 int kmp_a_debug = 0;
286 int kmp_b_debug = 0;
287 int kmp_c_debug = 0;
288 int kmp_d_debug = 0;
289 int kmp_e_debug = 0;
290 int kmp_f_debug = 0;
291 int kmp_diag = 0;
292 #endif
293 
294 /* For debug information logging using rotating buffer */
295 int __kmp_debug_buf = FALSE; /* TRUE means use buffer, FALSE means print to stderr */
296 int __kmp_debug_buf_lines = KMP_DEBUG_BUF_LINES_INIT; /* Lines of debug stored in buffer */
297 int __kmp_debug_buf_chars = KMP_DEBUG_BUF_CHARS_INIT; /* Characters allowed per line in buffer */
298 int __kmp_debug_buf_atomic = FALSE; /* TRUE means use atomic update of buffer entry pointer */
299 
300 char *__kmp_debug_buffer = NULL; /* Debug buffer itself */
301 int __kmp_debug_count = 0; /* Counter for number of lines printed in buffer so far */
302 int __kmp_debug_buf_warn_chars = 0; /* Keep track of char increase recommended in warnings */
303 /* end rotating debug buffer */
304 
305 #ifdef KMP_DEBUG
306 int __kmp_par_range; /* +1 => only go par for constructs in range */
307  /* -1 => only go par for constructs outside range */
308 char __kmp_par_range_routine[KMP_PAR_RANGE_ROUTINE_LEN] = { '\0' };
309 char __kmp_par_range_filename[KMP_PAR_RANGE_FILENAME_LEN] = { '\0' };
310 int __kmp_par_range_lb = 0;
311 int __kmp_par_range_ub = INT_MAX;
312 #endif /* KMP_DEBUG */
313 
314 /* For printing out dynamic storage map for threads and teams */
315 int __kmp_storage_map = FALSE; /* True means print storage map for threads and teams */
316 int __kmp_storage_map_verbose = FALSE; /* True means storage map includes placement info */
317 int __kmp_storage_map_verbose_specified = FALSE;
318 /* Initialize the library data structures when we fork a child process, defaults to TRUE */
319 int __kmp_need_register_atfork = TRUE; /* At initialization, call pthread_atfork to install fork handler */
320 int __kmp_need_register_atfork_specified = TRUE;
321 
322 
323 int __kmp_env_chunk = FALSE; /* KMP_CHUNK specified? */
324 int __kmp_env_stksize = FALSE; /* KMP_STACKSIZE specified? */
325 int __kmp_env_omp_stksize = FALSE; /* OMP_STACKSIZE specified? */
326 int __kmp_env_all_threads = FALSE;/* KMP_ALL_THREADS or KMP_MAX_THREADS specified? */
327 int __kmp_env_omp_all_threads = FALSE;/* OMP_THREAD_LIMIT specified? */
328 int __kmp_env_blocktime = FALSE; /* KMP_BLOCKTIME specified? */
329 int __kmp_env_checks = FALSE; /* KMP_CHECKS specified? */
330 int __kmp_env_consistency_check = FALSE; /* KMP_CONSISTENCY_CHECK specified? */
331 
332 kmp_uint32 __kmp_yield_init = KMP_INIT_WAIT;
333 kmp_uint32 __kmp_yield_next = KMP_NEXT_WAIT;
334 kmp_uint32 __kmp_yielding_on = 1;
335 kmp_uint32 __kmp_yield_cycle = 1; /* Yield-cycle is on by default */
336 kmp_int32 __kmp_yield_on_count = 10; /* By default, yielding is on for 10 monitor periods. */
337 kmp_int32 __kmp_yield_off_count = 1; /* By default, yielding is off for 1 monitor periods. */
338 /* ----------------------------------------------------- */
339 
340 
341 /* ------------------------------------------------------ */
342 /* STATE mostly syncronized with global lock */
343 /* data written to rarely by masters, read often by workers */
344 /*
345  * SHALL WE EDIT THE COMMENT BELOW IN SOME WAY?
346  * TODO: None of this global padding stuff works consistently because
347  * the order of declaration is not necessarily correlated to storage order.
348  * To fix this, all the important globals must be put in a big structure
349  * instead.
350  */
351 KMP_ALIGN_CACHE
352  kmp_info_t **__kmp_threads = NULL;
353  kmp_root_t **__kmp_root = NULL;
354 
355 /* data read/written to often by masters */
356 KMP_ALIGN_CACHE
357 volatile int __kmp_nth = 0;
358 volatile int __kmp_all_nth = 0;
359 int __kmp_thread_pool_nth = 0;
360 volatile kmp_info_t *__kmp_thread_pool = NULL;
361 volatile kmp_team_t *__kmp_team_pool = NULL;
362 
363 KMP_ALIGN_CACHE
364 volatile int __kmp_thread_pool_active_nth = 0;
365 
366 /* -------------------------------------------------
367  * GLOBAL/ROOT STATE */
368 KMP_ALIGN_CACHE
369 kmp_global_t __kmp_global = {{ 0 }};
370 
371 /* ----------------------------------------------- */
372 /* GLOBAL SYNCRONIZATION LOCKS */
373 /* TODO verify the need for these locks and if they need to be global */
374 KMP_ALIGN_CACHE
375 
376 kmp_bootstrap_lock_t __kmp_initz_lock = KMP_BOOTSTRAP_LOCK_INITIALIZER( __kmp_initz_lock ); /* Control initializations */
377 kmp_bootstrap_lock_t __kmp_forkjoin_lock; /* control fork/join access */
378 kmp_bootstrap_lock_t __kmp_exit_lock; /* exit() is not always thread-safe */
379 kmp_bootstrap_lock_t __kmp_monitor_lock; /* control monitor thread creation */
380 kmp_bootstrap_lock_t __kmp_tp_cached_lock; /* used for the hack to allow threadprivate cache and __kmp_threads expansion to co-exist */
381 
382 KMP_ALIGN(128)
383 kmp_lock_t __kmp_global_lock; /* Control OS/global access */
384 KMP_ALIGN(128)
385 kmp_queuing_lock_t __kmp_dispatch_lock; /* Control dispatch access */
386 KMP_ALIGN(128)
387 kmp_lock_t __kmp_debug_lock; /* Control I/O access for KMP_DEBUG */
388 
389 /* ----------------------------------------------- */
390 
391 #if KMP_HANDLE_SIGNALS
392  /*
393  Signal handling is disabled by default, because it confuses users: In case of sigsegv
394  (or other trouble) in user code signal handler catches the signal, which then "appears" in
395  the monitor thread (when the monitor executes raise() function). Users see signal in the
396  monitor thread and blame OpenMP RTL.
397 
398  Grant said signal handling required on some older OSes (Irix?) supported by KAI, because
399  bad applications hung but not aborted. Currently it is not a problem for Linux* OS, OS X* and
400  Windows* OS.
401 
402  Grant: Found new hangs for EL4, EL5, and a Fedora Core machine. So I'm putting
403  the default back for now to see if that fixes hangs on those machines.
404 
405  2010-04013 Lev: It was a bug in Fortran RTL. Fortran RTL prints a kind of stack backtrace
406  when program is aborting, but the code is not signal-safe. When multiple signals raised at
407  the same time (which occurs in dynamic negative tests because all the worker threads detects
408  the same error), Fortran RTL may hang. The bug finally fixed in Fortran RTL library provided
409  by Steve R., and will be available soon.
410  */
411  int __kmp_handle_signals = FALSE;
412 #endif
413 
414 /* ----------------------------------------------- */
415 #ifdef BUILD_TV
416 kmp_key_t __kmp_tv_key = 0;
417 #endif
418 
419 /* ------------------------------------------------------------------------ */
420 /* ------------------------------------------------------------------------ */
421 
422 #ifdef DEBUG_SUSPEND
423 int
424 get_suspend_count_( void ) {
425  int count = __kmp_suspend_count;
426  __kmp_suspend_count = 0;
427  return count;
428 }
429 void
430 set_suspend_count_( int * value ) {
431  __kmp_suspend_count = *value;
432 }
433 #endif
434 
435 // Symbols for MS mutual detection.
436 int _You_must_link_with_exactly_one_OpenMP_library = 1;
437 int _You_must_link_with_Intel_OpenMP_library = 1;
438 #if KMP_OS_WINDOWS && ( KMP_VERSION_MAJOR > 4 )
439  int _You_must_link_with_Microsoft_OpenMP_library = 1;
440 #endif
441 
442 // end of file //
Generated on Mon Jul 29 2013 13:01:21 for Intel® OpenMP* Runtime Library by   doxygen 1.8.4