text stringlengths 494 1.02k |
|---|
cancellation is disabled and the cancel construct and cancellation points are effectively ignored.
11
The behavior of the program is implementation defined if OMP_CANCELLATION is set to neither
12
true nor false.
13
Cross References
14
• cancel directive, see Section 16.1
15
• cancel-var ICV, see Table 2.1
16
21.2.7 OMP_DEFAULT_DEVICE
17
The OMP_DEFAULT_DEVICE environment variable sets the device number to use in device
18
constructs by setting the initial value of the default-device-var ICV. The value of this environment
19
variable must be a non-negative integer value.
20
Cross References
21
• Device Directives and Clauses, see Chapter 13
22
• default-device-var ICV, see Table 2.1
23
21.2.8 OMP_TARGET_OFFLOAD
24
The OMP_TARGET_OFFLOAD environment variable sets the initial value of the target-offload-var
25
ICV. Its value must be one of the following:
26
mandatory | disabled | default
27
The mandatory value specifies that the effect of any device construct or device memory routine
28
that uses a device that is un |
g, see
25
Section 1.3 and Chapter 12.
26
In parallel regions, references by the primary thread are to the copy of the variable in the thread
27
that encountered the parallel region.
28
During a sequential part, references are to the initial thread’s copy of the variable. The values of
29
data in the initial thread’s copy of a threadprivate variable are guaranteed to persist between any
30
CHAPTER 5. DATA ENVIRONMENT
101
two consecutive references to the variable in the program, provided that no teams construct that is
1
not nested inside of a target construct is encountered between the references and that the initial
2
thread is not executing code inside of a teams region. For initial threads that are executing code
3
inside of a teams region, the values of data in the copies of a threadprivate variable of those initial
4
threads are guaranteed to persist between any two consecutive references to the variable inside that
5
teams region.
6
The values of data in the threadprivate variables of threads that are n |
task_reduction clause or a reduction clause with task as
1
the reduction-modifier that is specified on a construct that corresponds to a region in which the
2
region of the participating task is closely nested must match each list item. The construct that
3
corresponds to the innermost enclosing region that meets this condition must specify the same
4
reduction-identifier for the matching list item as the in_reduction clause.
5
Cross References
6
• target directive, see Section 13.8
7
• task directive, see Section 12.5
8
• taskloop directive, see Section 12.6
9
5.5.11 declare reduction Directive
10
Name: declare reduction
Association: none
Category: declarative
Properties: pure
11
Arguments
12
declare reduction(reduction-specifier)
13
Name
Type
Properties
reduction-specifier
OpenMP reduction specifier
default
14
Clauses
15
initializer
16
Semantics
17
The declare reduction directive declares a reduction-identifier that can be used in a
18
reduction clause as a user-defined reduction. The directive argument reduction- |
atisfied if the operators in expr have precedence greater than binop, or by using parentheses
9
around expr or subexpressions of expr.
10
• The expression expr binop x must be numerically equivalent to (expr) binop x. This requirement
11
is satisfied if the operators in expr have precedence equal to or greater than binop, or by using
12
parentheses around expr or subexpressions of expr.
13
• The expression x ordop expr must be numerically equivalent to x ordop (expr). This requirement
14
is satisfied if the operators in expr have precedence greater than ordop, or by using parentheses
15
around expr or subexpressions of expr.
16
• The expression expr ordop x must be numerically equivalent to (expr) ordop x. This requirement
17
is satisfied if the operators in expr have precedence equal to or greater than ordop, or by using
18
parentheses around expr or subexpressions of expr.
19
• The expression x == e must be numerically equivalent to x == (e). This requirement is satisfied
20
if the operators in e have precedence |
base language.
26
program order
An ordering of operations performed by the same thread as determined by the
27
execution sequence of operations specified by the base language.
28
2
OpenMP API – Version 5.2 November 2021
COMMENT: For versions of C and C++ that include base language
1
support for threading, program order corresponds to the sequenced before
2
relation between operations performed by the same thread.
3
structured block
For C/C++, an executable statement, possibly compound, with a single entry at the
4
top and a single exit at the bottom, or an OpenMP construct.
5
For Fortran, a strictly structured block or a loosely structured block.
6
structured block
sequence
For C/C++, a sequence of zero or more executable statements (including OpenMP
7
constructs) that together have a single entry at the top and a single exit at the bottom.
8
For Fortran, a block of zero or more executable constructs (including OpenMP
9
constructs) with a single entry at the top and a single exit at the bottom.
10
strictly str |
cifiers (see Section 3.1).
3
• The omp_all_memory reserved locator was added (see Section 3.1), and the depend clause
4
was extended to allow its use (see Section 15.9.5).
5
• Support for private and firstprivate as an argument to the default clause in C and
6
C++ was added (see Section 5.4.1).
7
• Support was added so that iterators may be defined and used in a map clause (see Section 5.8.3)
8
or in data-motion clause on a target update directive (see Section 13.9).
9
• The present argument was added to the defaultmap clause (see Section 5.8.7).
10
• Support for the align clause on the allocate directive and allocator and align
11
modifiers on the allocate clause was added (see Chapter 6).
12
• The target_device trait set was added to the OpenMP context (see Section 7.1), and the
13
target_device selector set was added to context selectors (see Section 7.2).
14
• For C/C++, the declare variant directive was extended to support elision of preprocessed code
15
and to allow enclosed function definitions to be inter |
near-modifier(list) was deprecated and the step modifier
16
was added for specifying the linear step (see Section 5.4.6).
17
• The minus (-) operator for reductions was deprecated (see Section 5.5.5).
18
• The syntax of modifiers without comma separators in the map clause was deprecated (see
19
Section 5.8.3).
20
• To support the complete range of user-defined mappers and to improve consistency of map
21
clause usage, the declare mapper directive was extended to accept iterator-modifier and the
22
present map-type-modifier (see Section 5.8.3 and Section 5.8.8).
23
• If a matching mapped list item is not found in the data environment, the pointer retains its
24
original value as per the firstprivate semantics (see Section 5.8.6).
25
• The enter clause was added as a synonym for the to clause on the declare target directive, and
26
the corresponding to clause was deprecated to reduce parsing ambiguity (see Section 5.8.4 and
27
Section 7.8).
28
Fortran
• Metadirectives (see Section 7.4), assumption directives (see Sect |
se flush, and it will load any value of a shared variable propagated by a
11
release flush that synchronizes with it into its temporary view so that it may be subsequently read.
12
Therefore, release and acquire flushes may also be used to guarantee that a value written to a
13
variable by one thread may be read by a second thread. To accomplish this, the programmer must
14
ensure that the second thread has not written to the variable since its last acquire flush, and that the
15
following sequence of events happen in this specific order:
16
1. The value is written to the variable by the first thread;
17
2. The first thread performs a release flush;
18
3. The second thread performs an acquire flush; and
19
4. The value is read from the variable by the second thread.
20
21
Note – OpenMP synchronization operations, described in Chapter 15 and in Section 18.9, are
22
recommended for enforcing this order. Synchronization through variables is possible but is not
23
recommended because the proper timing of flushes is difficult |
ialized with the value ompt_data_none.
3
This runtime entry point is async signal safe.
4
Between a parallel-begin event and an implicit-task-begin event, a call to
5
ompt_get_parallel_info(0,...) may return information about the outer parallel team or
6
the new parallel team.
7
If a thread is in the state ompt_state_wait_barrier_implicit_parallel then a call to
8
ompt_get_parallel_info may return a pointer to a copy of the specified parallel region’s
9
parallel_data rather than a pointer to the data word for the region itself. This convention enables
10
the primary thread for a parallel region to free storage for the region immediately after the region
11
ends, yet avoid having some other thread in the team that is executing the region potentially
12
reference the parallel_data object for the region after it has been freed.
13
Description of Arguments
14
The ancestor_level argument specifies the parallel region of interest by its ancestor level. Ancestor
15
level 0 refers to the innermost parallel region; info |
ved.
17
• The definition of the nest-var, dyn-var, nthreads-var and run-sched-var internal control variables
18
(ICVs) were modified to provide one copy of these ICVs per task instead of one copy for the
19
whole program (see Chapter 2). The omp_set_num_threads, omp_set_nested, and
20
omp_set_dynamic runtime library routines were specified to support their use from inside a
21
parallel region (see Section 18.2.1, Section 18.2.6 and Section 18.2.9).
22
• The thread-limit-var ICV, the omp_get_thread_limit runtime library routine and the
23
OMP_THREAD_LIMIT environment variable were added to support control of the maximum
24
number of threads (see Section 2.1, Section 18.2.13 and Section 21.1.3).
25
• The max-active-levels-var ICV, omp_set_max_active_levels and
26
omp_get_max_active_levels runtime library routines, and
27
OMP_MAX_ACTIVE_LEVELS environment variable were added to support control of the
28
number of nested active parallel regions (see Section 2.1, Section 18.2.15, Section 18.2.16
29
and Section 21.1.4 |
e variable passed
32
by reference in callback and ompt_get_callback returns 1; otherwise, it returns 0. If
33
ompt_get_callback returns 0, the value of the variable passed by reference as callback is
34
undefined.
35
514
OpenMP API – Version 5.2 November 2021
Description of Arguments
1
The event argument indicates the event for which the callback would be invoked.
2
The callback argument returns a pointer to the callback associated with event.
3
Constraints on Arguments
4
The callback argument cannot be NULL and must point to valid storage.
5
Cross References
6
• Callbacks, see Section 19.4.2
7
• ompt_callback_t, see Section 19.4.4.1
8
• ompt_set_callback_t, see Section 19.6.1.3
9
19.6.1.5 ompt_get_thread_data_t
10
Summary
11
The ompt_get_thread_data_t type is the type signature of the
12
ompt_get_thread_data runtime entry point, which returns the address of the thread data
13
object for the current thread.
14
Format
15
C / C++
typedef ompt_data_t *(*ompt_get_thread_data_t) (void);
16
C / C++
Semantics
17
Each |
eger.
24
The stride must evaluate to a positive integer.
25
When the size of the array dimension is not known, the length must be specified explicitly.
26
When the stride is absent it defaults to 1.
27
When the length is absent it defaults to ⌈⌈(size − lower-bound)/stride⌉⌉, where size is the size of the
28
array dimension.
29
When the lower-bound is absent it defaults to 0.
30
64
OpenMP API – Version 5.2 November 2021
C/C++ (cont.)
The precedence of a subscript operator that uses the array section syntax is the same as the
1
precedence of a subscript operator that does not use the array section syntax.
2
3
Note – The following are examples of array sections:
4
a[0:6]
5
a[0:6:1]
6
a[1:10]
7
a[1:]
8
a[:10:2]
9
b[10][:][:]
10
b[10][:][:0]
11
c[42][0:6][:]
12
c[42][0:6:2][:]
13
c[1:10][42][0:6]
14
S.c[:100]
15
p->y[:10]
16
this->a[:N]
17
(p+10)[:N]
18
Assume a is declared to be a 1-dimensional array with dimension size 11. The first two examples
19
are equivalent, and the third and fourth examples are equivalent. |
17
range-decl of each associated loop has the value that it would have if the set of the associated loops
18
was executed sequentially.
19
When an if clause is present and the if clause expression evaluates to false, undeferred tasks are
20
generated. The use of a variable in an if clause expression causes an implicit reference to the
21
variable in all enclosing constructs.
22
C++
For firstprivate variables of class type, the number of invocations of copy constructors that
23
perform the initialization is implementation defined.
24
C++
25
Note – When storage is shared by a taskloop region, the programmer must ensure, by adding
26
proper synchronization, that the storage does not reach the end of its lifetime before the taskloop
27
region and its descendent tasks complete their execution.
28
29
Execution Model Events
30
The taskloop-begin event occurs upon entering the taskloop region. A taskloop-begin will
31
precede any task-create events for the generated tasks. The taskloop-end event occurs upon
32
comple |
Clauses
23
Clause groups
24
Properties: unique, inarguable
Members: simd, threads
25
Directives
26
ordered
27
Semantics
28
The parallelization-level clause grouping defines a set of clauses that indicate the level of
29
parallelization with which to associate a construct.
30
Cross References
31
• ordered directive, see Section 15.10.2
32
CHAPTER 15. SYNCHRONIZATION CONSTRUCTS AND CLAUSES
331
16 Cancellation Constructs
1
This chapter defines constructs related to cancellation of OpenMP regions.
2
16.1 cancel Construct
3
Name: cancel
Association: none
Category: executable
Properties: default
4
Clauses
5
if, do, for, parallel, sections, taskgroup
6
Additional information
7
The cancel-directive-name clause set consists of the directive-name of each directive that has the
8
cancellable property (i.e., directive-name for the worksharing-loop construct, parallel,
9
sections and taskgroup). This clause set has the required, unique and exclusive properties.
10
Binding
11
The binding thread set of the cancel region is th |
indirect Clause
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
209
8
Informational and Utility Directives
210
8.1
at Clause . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
210
8.2
requires Directive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
210
8.2.1
requirement Clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
212
8.3
Assumption Directives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
213
8.3.1
assumption Clauses
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
213
8.3.2
assumes Directive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
214
8.3.3
assume Directive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
215
8.3.4
begin assumes Directive . . . . . . . . . . . . . . . . . . . . . . . . . .
215
8.4
nothing Directive
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
216
8.5
error Directive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
216
8.5.1
sev |
Fortran
Variables with predetermined data-sharing attributes may not be listed in data-sharing attribute
15
clauses, except for the cases listed below. For these exceptions only, listing a predetermined
16
variable in a data-sharing attribute clause is allowed and overrides the variable’s predetermined
17
data-sharing attributes.
18
• The loop iteration variable in any associated loop of a loop-associated construct may be listed in
19
a private or lastprivate clause.
20
• If a simd construct has just one associated loop then its loop iteration variable may be listed in a
21
linear clause with a linear-step that is the increment of the associated loop.
22
C / C++
• Variables with const-qualified type with no mutable members may be listed in a
23
firstprivate clause, even if they are static data members.
24
• The __func__ variable and similar function-local predefined variables may be listed in a
25
shared or firstprivate clause.
26
C / C++
98
OpenMP API – Version 5.2 November 2021
Fortran
• Loop iteration variab |
lay_affinity, 370
OMP_DISPLAY_ENV, 615
omp_display_env, 438
OMP_DYNAMIC, 600
omp_free, 430
omp_fulfill_event, 414
omp_get_active_level, 362
omp_get_affinity_format, 369
omp_get_ancestor_thread_num, 360
omp_get_cancellation, 353
omp_get_default_allocator, 428
omp_get_default_device, 382
omp_get_device_num, 384
omp_get_dynamic, 352
omp_get_initial_device, 385
omp_get_interop_int, 417
omp_get_interop_name, 420
omp_get_interop_ptr, 418
omp_get_interop_rc_desc, 421
omp_get_interop_str, 419
omp_get_interop_type_desc, 421
omp_get_level, 360
omp_get_mapped_ptr, 402
omp_get_max_active_levels, 359
omp_get_max_task_priority, 377
omp_get_max_teams, 374
omp_get_max_threads, 350
omp_get_nested, 354
omp_get_num_devices, 383
omp_get_num_interop_properties,
417
omp_get_num_places, 364
omp_get_num_procs, 381
omp_get_num_teams, 372
omp_get_num_threads, 349
omp_get_partition_num_places,
367
omp_get_partition_place_nums,
368
omp_get_place_num, 366
omp_get_place_num_procs, 365
omp_get_place_proc_ids, 365
omp_get_proc_bind, 363
omp |
performed.
24
Fortran
The omp_target_memcpy_rect_async routine requires an explicit interface and so might
25
not be provided in omp_lib.h.
26
Fortran
Execution Model Events
27
The target-data-op-begin event occurs before a thread initiates a data transfer in the
28
omp_target_memcpy_rect_async region.
29
The target-data-op-end event occurs after a thread initiates a data transfer in the
30
omp_target_memcpy_rect_async region.
31
Tool Callbacks
32
A thread dispatches a registered ompt_callback_target_data_op_emi callback with
33
ompt_scope_begin as its endpoint argument for each occurrence of a target-data-op-begin
34
event in that thread. Similarly, a thread dispatches a registered
35
ompt_callback_target_data_op_emi callback with ompt_scope_end as its endpoint
36
argument for each occurrence of a target-data-op-end event in that thread. These callbacks have
37
type signature ompt_callback_target_data_op_emi_t.
38
398
OpenMP API – Version 5.2 November 2021
A thread dispatches a registered ompt_callback_targ |
pe signature ompt_callback_masked_t.
31
Cross References
32
• ompt_callback_masked_t, see Section 19.5.2.12
33
• ompt_scope_endpoint_t, see Section 19.4.4.11
34
• filter clause, see Section 10.5.1
35
238
OpenMP API – Version 5.2 November 2021
10.5.1 filter Clause
1
Name: filter
Properties: unique
2
Arguments
3
Name
Type
Properties
thread_num
expression of integer type
default
4
Directives
5
masked
6
Semantics
7
If thread_num specifies the thread number of the current thread in the current team then the
8
filter clause selects the current thread. If the filter clause is not specified, the effect is as if
9
the clause is specified with thread_num equal to zero, so that the filter clause selects the
10
primary thread. The use of a variable in a thread_num clause expression causes an implicit
11
reference to the variable in all enclosing constructs.
12
Cross References
13
• masked directive, see Section 10.5
14
CHAPTER 10. PARALLELISM GENERATION AND CONTROL
239
11 Work-Distribution Constructs
1
A work-distribution co |
t item type
default
11
Directives
12
scan
13
Semantics
14
The inclusive clause is used on a separating directive that separates a structured block into two
15
structured block sequences. The clause determines the association of the structured block sequence
16
that precedes the directive on which the clause appears to a phase of that directive.
17
The list items that appear in an inclusive clause may include array sections.
18
Cross References
19
• scan directive, see Section 5.6
20
5.6.2 exclusive Clause
21
Name: exclusive
Properties: unique
22
Arguments
23
Name
Type
Properties
list
list of variable list item type
default
24
Directives
25
scan
26
CHAPTER 5. DATA ENVIRONMENT
143
Semantics
1
The exclusive clause is used on a separating directive that separates a structured block into two
2
structured block sequences. The clause determines the association of the structured block sequence
3
that precedes the directive on which the clause appears to a phase of that directive.
4
The list items that appear in an ex |
on-identifier.
22
After the end of the region, the original list item is updated with the values of the private copies
23
using the combiner associated with the reduction-identifier.
24
If reduction-modifier is not present or the default reduction-modifier is present, the behavior is
25
as follows. For parallel and worksharing constructs, one or more private copies of each list
26
134
OpenMP API – Version 5.2 November 2021
item are created for each implicit task, as if the private clause had been used. For the simd
1
construct, one or more private copies of each list item are created for each SIMD lane, as if the
2
private clause had been used. For the taskloop construct, private copies are created
3
according to the rules of the reduction scoping clauses. For the teams construct, one or more
4
private copies of each list item are created for the initial task of each team in the league, as if the
5
private clause had been used. For the loop construct, private copies are created and used in the
6
construct accordi |
em that appears in a reduction clause of a teams construct must not appear in a
25
firstprivate clause on a distribute construct if any of the distribute regions that
26
arise from the distribute construct ever bind to any of the teams regions that arise from the
27
teams construct.
28
• A list item that appears in a reduction clause of a worksharing construct must not appear in a
29
firstprivate clause in a task construct encountered during execution of any of the
30
worksharing regions that arise from the worksharing construct.
31
CHAPTER 5. DATA ENVIRONMENT
113
C++
• A variable of class type (or array thereof) that appears in a firstprivate clause requires an
1
accessible, unambiguous copy constructor for the class type.
2
• If the original list item in a firstprivate clause on a work-distribution construct has a
3
reference type then it must bind to the same object for all threads in the binding thread set of the
4
work-distribution region.
5
C++
Fortran
• If the list item is a polymorphic variable with t |
n unit,
21
omp_alloc and omp_aligned_alloc invocations that appear in target regions must not
22
pass omp_null_allocator as the allocator argument, which must be a constant expression
23
that evaluates to one of the predefined memory allocator values. The alignment argument to
24
omp_aligned_alloc must be a power of two and the size argument must be a multiple of
25
alignment.
26
Binding
27
The binding task set for an omp_alloc or omp_aligned_alloc region is the generating task.
28
Effect
29
The omp_alloc and omp_aligned_alloc routines request a memory allocation of size bytes
30
from the specified memory allocator. If the allocator argument is omp_null_allocator the
31
memory allocator used by the routines will be the one specified by the def-allocator-var ICV of the
32
binding implicit task. Upon success they return a pointer to the allocated memory. Otherwise, the
33
behavior that the fallback trait of the allocator specifies will be followed. If size is 0,
34
omp_alloc and omp_aligned_alloc will return NULL.
|
sk exists that modifies it.
22
As another example, if a lock acquire and release happen in different parts of a task region, no
23
attempt should be made to acquire the same lock in any part of another task that the executing
24
thread may schedule. Otherwise, a deadlock is possible. A similar situation can occur when a
25
critical region spans multiple parts of a task and another schedulable task contains a
26
critical region with the same name.
27
The use of threadprivate variables and the use of locks or critical sections in an explicit task with an
28
if clause must take into account that when the if clause evaluates to false, the task is executed
29
immediately, without regard to Task Scheduling Constraint 2.
30
31
Execution Model Events
32
The task-schedule event occurs in a thread when the thread switches tasks at a task scheduling
33
point; no event occurs when switching to or from a merged task.
34
CHAPTER 12. TASKING CONSTRUCTS
273
Tool Callbacks
1
A thread dispatches a registered ompt_callback_task_s |
or routine creates a new allocator that is associated with the
25
memspace memory space and returns a handle to it. All allocations through the created allocator
26
will behave according to the allocator traits specified in the traits argument. The number of traits in
27
the traits argument is specified by the ntraits argument. Specifying the same allocator trait more
28
than once results in unspecified behavior. The routine returns a handle for the created allocator. If
29
the special omp_atv_default value is used for a given trait, then its value will be the default
30
value specified in Table 6.2 for that given trait.
31
If memspace is omp_default_mem_space and the traits argument is an empty set this routine
32
will always return a handle to an allocator. Otherwise if an allocator based on the requirements
33
cannot be created then the special omp_null_allocator handle is returned.
34
CHAPTER 18. RUNTIME LIBRARY ROUTINES
425
Restrictions
1
The restrictions to the omp_init_allocator routine are as follows:
2
• |
out
min
omp_priv = Maximal
representable number in the
reduction list item type
omp_out = omp_in < omp_out ?
omp_in : omp_out
C / C++
Fortran
Table 5.2 lists each reduction identifier that is implicitly declared for numeric and logical types and
1
its semantic initializer value. The actual initializer value is that value as expressed in the data type
2
of the reduction list item.
3
TABLE 5.2: Implicitly Declared Fortran Reduction Identifiers
Identifier
Initializer
Combiner
+
omp_priv = 0
omp_out = omp_in + omp_out
- (depre-
cated)
omp_priv = 0
omp_out = omp_in + omp_out
*
omp_priv = 1
omp_out = omp_in * omp_out
table continued on next page
CHAPTER 5. DATA ENVIRONMENT
129
table continued from previous page
Identifier
Initializer
Combiner
.and.
omp_priv = .true.
omp_out = omp_in .and. omp_out
.or.
omp_priv = .false.
omp_out = omp_in .or. omp_out
.eqv.
omp_priv = .true.
omp_out = omp_in .eqv. omp_out
.neqv.
omp_priv = .false.
omp_out = omp_in .neqv. omp_out
max
omp_priv = Minimal
representable number in the
reductio |
Section 6.5
5
Fortran
6.8 uses_allocators Clause
6
Name: uses_allocators
Properties: data-environment attribute, data-
sharing attribute
7
Arguments
8
Name
Type
Properties
allocator
expression of allocator_handle type
default
9
Modifiers
10
Name
Modifies
Type
Properties
mem-space
Generic
Complex, name: memspace
Arguments:
memspace-handle
expression of
memspace_handle type (de-
fault)
default
traits-array
Generic
Complex, name: traits
Arguments:
traits variable of alloctrait
array type (default)
default
11
Directives
12
target
13
Additional information
14
The comma-separated list syntax, in which each list item is a clause-argument-specification of the
15
form allocator[(traits)] may also be used for the uses_allocators clause arguments. With
16
this syntax, traits must be a constant array with constant values. This syntax has been deprecated.
17
CHAPTER 6. MEMORY MANAGEMENT
181
Semantics
1
The uses_allocators clause enables the use of the specified allocator in the region associated
2
with the directive on which |
callback for each
30
occurrence of a taskwait-complete event. This callback has the type signature
31
ompt_callback_task_schedule_t with ompt_taskwait_complete as its
32
prior_task_status argument.
33
Restrictions
34
Restrictions to the taskwait construct are as follows:
35
• The mutexinoutset dependence-type may not appear in a depend clause on a taskwait
36
construct.
37
• If the dependence-type of a depend clause is depobj then the dependence objects cannot
38
represent dependences of the mutexinoutset dependence type.
39
CHAPTER 15. SYNCHRONIZATION CONSTRUCTS AND CLAUSES
307
• The nowait clause may only appear on a taskwait directive if the depend clause is present.
1
Cross References
2
• ompt_callback_sync_region_t, see Section 19.5.2.13
3
• ompt_scope_endpoint_t, see Section 19.4.4.11
4
• ompt_sync_region_t, see Section 19.4.4.14
5
• depend clause, see Section 15.9.5
6
• nowait clause, see Section 15.6
7
• task directive, see Section 12.5
8
15.6 nowait Clause
9
Name: nowait
Properties: unique, end-claus |
h returns the place number of the place to which the current thread is
33
bound.
34
CHAPTER 19. OMPT INTERFACE
517
Format
1
C / C++
typedef int (*ompt_get_place_num_t) (void);
2
C / C++
Semantics
3
When the current thread is bound to a place, ompt_get_place_num returns the place number
4
associated with the thread. The returned value is between 0 and one less than the value returned by
5
ompt_get_num_places, inclusive. When the current thread is not bound to a place, the routine
6
returns -1. This runtime entry point is async signal safe.
7
19.6.1.10 ompt_get_partition_place_nums_t
8
Summary
9
The ompt_get_partition_place_nums_t type is the type signature of the
10
ompt_get_partition_place_nums runtime entry point, which returns a list of place
11
numbers that correspond to the places in the place-partition-var ICV of the innermost implicit task.
12
Format
13
C / C++
typedef int (*ompt_get_partition_place_nums_t) (
14
int place_nums_size,
15
int *place_nums
16
);
17
C / C++
Semantics
18
The ompt_get_partition |
hint_t omp_lock_hint_t;
9
C / C++
Fortran
integer, parameter :: omp_lock_hint_kind = omp_sync_hint_kind
10
11
integer (kind=omp_sync_hint_kind), &
12
parameter :: omp_sync_hint_none = &
13
int(Z’0’, kind=omp_sync_hint_kind)
14
integer (kind=omp_lock_hint_kind), &
15
parameter :: omp_lock_hint_none = omp_sync_hint_none
16
integer (kind=omp_sync_hint_kind), &
17
parameter :: omp_sync_hint_uncontended = &
18
int(Z’1’, kind=omp_sync_hint_kind)
19
integer (kind=omp_lock_hint_kind), &
20
parameter :: omp_lock_hint_uncontended = &
21
omp_sync_hint_uncontended
22
integer (kind=omp_sync_hint_kind), &
23
parameter :: omp_sync_hint_contended = &
24
int(Z’2’, kind=omp_sync_hint_kind)
25
integer (kind=omp_lock_hint_kind), &
26
parameter :: omp_lock_hint_contended = &
27
omp_sync_hint_contended
28
integer (kind=omp_sync_hint_kind), &
29
parameter :: omp_sync_hint_nonspeculative = &
30
int(Z’4’, kind=omp_sync_hint_kind)
31
integer (kind=omp_lock_hint_kind), &
32
parameter :: omp_lock_hint_nonspeculative = &
33
omp_sync_hint |
r ICV.
22
Cross References
23
• place-partition-var ICV, see Table 2.1
24
CHAPTER 18. RUNTIME LIBRARY ROUTINES
367
18.3.7 omp_get_partition_place_nums
1
Summary
2
The omp_get_partition_place_nums routine returns the list of place numbers
3
corresponding to the places in the place-partition-var ICV of the innermost implicit task.
4
Format
5
C / C++
void omp_get_partition_place_nums(int *place_nums);
6
C / C++
Fortran
subroutine omp_get_partition_place_nums(place_nums)
7
integer place_nums(*)
8
Fortran
Binding
9
The binding task set for an omp_get_partition_place_nums region is the encountering
10
implicit task.
11
Effect
12
The omp_get_partition_place_nums routine returns the list of place numbers that
13
correspond to the places in the place-partition-var ICV of the innermost implicit task. The array
14
must be sufficiently large to contain omp_get_partition_num_places() integers;
15
otherwise, the behavior is unspecified.
16
Cross References
17
• omp_get_partition_num_places, see Section 18.3.6
18
• place-parti |
the numerical identifiers of each processor
11
associated with the place numbered place_num. The numerical identifiers are non-negative and
12
their meaning is implementation defined. The numerical identifiers are returned in the array ids and
13
their order in the array is implementation defined. The array must be sufficiently large to contain
14
omp_get_place_num_procs(place_num) integers; otherwise, the behavior is unspecified.
15
The routine has no effect when place_num has a negative value or a value greater than or equal to
16
omp_get_num_places().
17
Cross References
18
• OMP_PLACES, see Section 21.1.6
19
• omp_get_num_places, see Section 18.3.2
20
• omp_get_place_num_procs, see Section 18.3.3
21
18.3.5 omp_get_place_num
22
Summary
23
The omp_get_place_num routine returns the place number of the place to which the
24
encountering thread is bound.
25
Format
26
C / C++
int omp_get_place_num(void);
27
C / C++
Fortran
integer function omp_get_place_num()
28
Fortran
366
OpenMP API – Version 5.2 November 2021
Binding |
cation and that includes a
14
442
OpenMP API – Version 5.2 November 2021
definition of ompt_start_tool.
1
If the value of tool-var is enabled, the OpenMP implementation must check if a tool has provided
2
an implementation of ompt_start_tool. The OpenMP implementation first checks if a
3
tool-provided implementation of ompt_start_tool is available in the address space, either
4
statically-linked into the application or in a dynamically-linked library loaded in the address space.
5
If multiple implementations of ompt_start_tool are available, the OpenMP implementation
6
will use the first tool-provided implementation of ompt_start_tool that it finds.
7
If the implementation does not find a tool-provided implementation of ompt_start_tool in the
8
address space, it consults the tool-libraries-var ICV, which contains a (possibly empty) list of
9
dynamically-linked libraries. As described in detail in Section 21.3.2, the libraries in
10
tool-libraries-var are then searched for the first usable implementation of ompt_sta |
ion 17.1 for the restrictions on nesting.
2
closely nested region
A region nested inside another region with no parallel region nested between
3
them.
4
strictly nested region
A region nested inside another region with no other explicit region nested between
5
them.
6
all threads
All OpenMP threads participating in the OpenMP program.
7
current team
All threads in the team executing the innermost enclosing parallel region.
8
encountering thread
For a given region, the thread that encounters the corresponding construct.
9
all tasks
All tasks participating in the OpenMP program.
10
current team tasks
All tasks encountered by the corresponding team. The implicit tasks constituting the
11
parallel region and any descendent tasks encountered during the execution of
12
these implicit tasks are included in this set of tasks.
13
generating task
For a given region, the task for which execution by a thread generated the region.
14
binding thread set
The set of threads that are affected by, or provide the context for, th |
alue of
5
max-active-levels-var is set to 1. The behavior of the program is implementation defined if the
6
value of OMP_NESTED is neither true nor false.
7
If both the OMP_NESTED and OMP_MAX_ACTIVE_LEVELS environment variables are set, the
8
value of OMP_NESTED is false, and the value of OMP_MAX_ACTIVE_LEVELS is greater than
9
1, then the behavior is implementation defined. Otherwise, if both environment variables are set
10
then the OMP_NESTED environment variable has no effect.
11
The OMP_NESTED environment variable has been deprecated.
12
Example:
13
setenv OMP_NESTED false
14
Cross References
15
• OMP_MAX_ACTIVE_LEVELS, see Section 21.1.4
16
• max-active-levels-var ICV, see Table 2.1
17
21.1.6 OMP_PLACES
18
The OMP_PLACES environment variable sets the initial value of the place-partition-var ICV. A list
19
of places can be specified in the OMP_PLACES environment variable. The value of OMP_PLACES
20
can be one of two types of values: either an abstract name that describes a set of places or an
21
explicit lis |
oyed with omp_destroy_allocator results in unspecified
12
behavior.
13
Cross References
14
• Memory Allocators, see Section 6.2
15
• omp_destroy_allocator, see Section 18.13.3
16
18.13.8 omp_calloc and omp_aligned_calloc
17
Summary
18
The omp_calloc and omp_aligned_calloc routines request a zero initialized memory
19
allocation from a memory allocator.
20
Format
21
C
void *omp_calloc(
22
size_t nmemb,
23
size_t size,
24
omp_allocator_handle_t allocator
25
);
26
void *omp_aligned_calloc(
27
size_t alignment,
28
size_t nmemb,
29
size_t size,
30
omp_allocator_handle_t allocator
31
);
32
C
CHAPTER 18. RUNTIME LIBRARY ROUTINES
431
C++
void *omp_calloc(
1
size_t nmemb,
2
size_t size,
3
omp_allocator_handle_t allocator=omp_null_allocator
4
);
5
void *omp_aligned_calloc(
6
size_t alignment,
7
size_t nmemb,
8
size_t size,
9
omp_allocator_handle_t allocator=omp_null_allocator
10
);
11
C++
Fortran
type(c_ptr) function omp_calloc(nmemb, size, allocator) bind(c)
12
use, intrinsic :: iso_c_binding, only : c_ptr, c_size_t
13 |
in
3
the program unit.
4
• If a threadprivate variable or a threadprivate common block is declared with the BIND attribute,
5
the corresponding C entities must also be specified in a threadprivate directive in the C
6
program.
7
• A variable may only appear as an argument in a threadprivate directive in the scope in
8
which it is declared. It must not be an element of a common block or appear in an
9
EQUIVALENCE statement.
10
• A variable that appears as an argument in a threadprivate directive must be declared in the
11
scope of a module or have the SAVE attribute, either explicitly or implicitly.
12
• The effect of an access to a threadprivate variable in a DO CONCURRENT construct is unspecified.
13
Fortran
Cross References
14
• Determining the Number of Threads for a parallel Region, see Section 10.1.1
15
• copyin clause, see Section 5.7.1
16
• dyn-var ICV, see Table 2.1
17
• order clause, see Section 10.3
18
5.3 List Item Privatization
19
Some data-sharing attribute clauses, including reduction clauses, spec |
ICV is a list. The runtime call omp_set_num_threads sets
4
the value of the first element of this list, and omp_get_max_threads retrieves the value of
5
the first element of this list.
6
• Detailed values in the place-partition-var ICV are retrieved using the listed runtime calls.
7
CHAPTER 2. INTERNAL CONTROL VARIABLES
43
• The thread_limit clause sets the thread-limit-var ICV for the region of the construct on
1
which it appears.
2
Cross References
3
• omp_get_active_level, see Section 18.2.20
4
• omp_get_affinity_format, see Section 18.3.9
5
• omp_get_cancellation, see Section 18.2.8
6
• omp_get_default_allocator, see Section 18.13.5
7
• omp_get_default_device, see Section 18.7.3
8
• omp_get_dynamic, see Section 18.2.7
9
• omp_get_level, see Section 18.2.17
10
• omp_get_max_active_levels, see Section 18.2.16
11
• omp_get_max_task_priority, see Section 18.5.1
12
• omp_get_max_teams, see Section 18.4.4
13
• omp_get_max_threads, see Section 18.2.3
14
• omp_get_num_procs, see Section 18.7.1
15
• omp_get_num_thre |
API – Version 5.2 November 2021
• nogroup clause, see Section 15.7
1
• num_tasks clause, see Section 12.6.2
2
• priority clause, see Section 12.4
3
• private clause, see Section 5.4.3
4
• reduction clause, see Section 5.5.8
5
• shared clause, see Section 5.4.2
6
• task directive, see Section 12.5
7
• taskgroup directive, see Section 15.4
8
• untied clause, see Section 12.1
9
12.6.1 grainsize Clause
10
Name: grainsize
Properties: unique
11
Arguments
12
Name
Type
Properties
grain-size
expression of integer type
positive
13
Modifiers
14
Name
Modifies
Type
Properties
prescriptiveness
grain-size
Keyword: strict
unique
15
Directives
16
taskloop
17
Semantics
18
The grainsize clause specifies the number of logical iterations, Lt, that are assigned to each
19
generated task t. If prescriptiveness is not specified as strict, other than possibly for the
20
generated task that contains the sequentially last iteration, Lt is greater than or equal to the
21
minimum of the value of the grain-size expression and the number of l |
5
7.6.2 nocontext Clause
16
Name: nocontext
Properties: unique
17
Arguments
18
Name
Type
Properties
do-not-update-context
expression of logical type
default
19
Directives
20
dispatch
21
Semantics
22
If do-not-update-context evaluates to true, the construct on which the nocontext clause appears
23
is not added to the construct set of the OpenMP context. The use of a variable in
24
do-not-update-context causes an implicit reference to the variable in all enclosing constructs.
25
do-not-update-context is evaluated in the enclosing context.
26
Cross References
27
• dispatch directive, see Section 7.6
28
CHAPTER 7. VARIANT DIRECTIVES
201
7.7 declare simd Directive
1
Name: declare simd
Association: declaration
Category: declarative
Properties: pure
2
Arguments
3
declare simd[(proc-name)]
4
Name
Type
Properties
proc-name
identifier of function type
optional
5
Clause groups
6
branch
7
Clauses
8
aligned, linear, simdlen, uniform
9
Semantics
10
The association of one or more declare simd directives with a function decla |
1
• ompd_rc_callback if a callback returned an unexpected error, which leads to a failure of the
1
query;
2
• ompd_rc_needs_state_tracking if the information cannot be provided while the
3
debug-var is disabled;
4
• ompd_rc_ok on success; or
5
• ompd_rc_error for any other error.
6
20.5.1 Per OMPD Library Initialization and Finalization
7
The OMPD library must be initialized exactly once after it is loaded, and finalized exactly once
8
before it is unloaded. Per OpenMP process or core file initialization and finalization are also
9
required. Once loaded, the tool can determine the version of the OMPD API that the library
10
supports by calling ompd_get_api_version (see Section 20.5.1.2). If the tool supports the
11
version that ompd_get_api_version returns, the tool starts the initialization by calling
12
ompd_initialize (see Section 20.5.1.1) using the version of the OMPD API that the library
13
supports. If the tool does not support the version that ompd_get_api_version returns, it may
14
attempt to call ompd_ |
evenly, the exact number of threads in a particular
6
subpartition is implementation defined.
7
The determination of whether the affinity request can be fulfilled is implementation defined. If the
8
affinity request cannot be fulfilled, then the affinity of threads in the team is implementation defined.
9
10
Note – Wrap around is needed if the end of a place partition is reached before all thread
11
assignments are done. For example, wrap around may be needed in the case of close and T ≤ P,
12
if the primary thread is assigned to a place other than the first place in the place partition. In this
13
case, thread 1 is assigned to the place after the place of the primary thread, thread 2 is assigned to
14
the place after that, and so on. The end of the place partition may be reached before all threads are
15
assigned. In this case, assignment of threads is resumed with the first place in the place partition.
16
17
Cross References
18
• bind-var ICV, see Table 2.1
19
• parallel directive, see Section 10.1
20
• place-partitio |
ent loop schedules, 95
construct syntax, 48
constructs
allocators, 180
atomic, 311
barrier, 301
cancel, 332
cancellation constructs, 332
cancellation point, 336
combined constructs, 343
composite constructs, 343
642
OpenMP API – Version 5.2 November 2021
critical, 299
depobj, 322
device constructs, 275
dispatch, 200
distribute, 254
do, 251
flush, 315
for, 250
interop, 291
loop, 257
masked, 238
ordered, 328–330
parallel, 223
scope, 242
sections, 243
simd, 235
single, 240
target, 283
target data, 279
target enter data, 280
target exit data, 282
target update, 289
task, 262
taskgroup, 304
tasking constructs, 260
taskloop, 266
taskwait, 306
taskyield, 270
teams, 230
tile, 219
unroll, 220
work-distribution, 240
workshare, 245
worksharing, 240
worksharing-loop construct, 247
controlling OpenMP thread affinity, 228
copyin, 144
copyprivate, 146
critical, 299
D
data copying clauses, 144
data environment, 96
data terminology, 14
data-mapping control, 147
data-motion clauses, 165
data-sharing attribute clauses, 108
data-s |
n is
16
implementation defined (see Section 5.2).
17
Fortran
• is_device_ptr clause: Support for pointers created outside of the OpenMP device data
18
management routines is implementation defined (see Section 5.4.7).
19
Chapter 6:
20
• Memory spaces: The actual storage resources that each memory space defined in Table 6.1
21
represents are implementation defined. The mechanism that provides the constant value of the
22
variables allocated in the omp_const_mem_space memory space is implementation defined
23
(see Section 6.1).
24
• Memory allocators: The minimum size for partitioning allocated memory over storage
25
resources is implementation defined. The default value for the pool_size allocator trait (see
26
Table 6.2) is implementation defined. The memory spaces associated with the predefined
27
omp_cgroup_mem_alloc, omp_pteam_mem_alloc and omp_thread_mem_alloc
28
allocators (see Table 6.3) are implementation defined (see Section 6.2).
29
• aligned clause: If the alignment modifier is not specified, the default align |
11
Table 2.4 shows the override relationships among construct clauses and ICVs. The table only lists
12
ICVs that can be overridden by a clause.
13
TABLE 2.4: ICV Override Relationships
ICV
construct clause, if used
bind-var
proc_bind
def-allocator-var
allocate, allocator
nteams-var
num_teams
nthreads-var
num_threads
run-sched-var
schedule
teams-thread-limit-var
thread_limit
Semantics
14
• The num_threads clause overrides the value of the first element of the nthreads-var ICV.
15
• If a schedule clause specifies a modifier then that modifier overrides any modifier that is
16
specified in the run-sched-var ICV.
17
• If bind-var is not set to false then the proc_bind clause overrides the value of the first element
18
of the bind-var ICV; otherwise, the proc_bind clause has no effect.
19
Cross References
20
• allocate clause, see Section 6.6
21
• allocator clause, see Section 6.4
22
• num_teams clause, see Section 10.2.1
23
46
OpenMP API – Version 5.2 November 2021
• num_threads clause, see Section 10.1.2
1
• proc_bind |
ression evaluated to false, and the previous constraints are already met, the task is executed
10
immediately after generation of the task.
11
A program that relies on any other assumption about task scheduling is non-conforming.
12
13
Note – Task scheduling points dynamically divide task regions into parts. Each part is executed
14
uninterrupted from start to end. Different parts of the same task region are executed in the order in
15
which they are encountered. In the absence of task synchronization constructs, the order in which a
16
thread executes parts of different schedulable tasks is unspecified.
17
A program must behave correctly and consistently with all conceivable scheduling sequences that
18
are compatible with the rules above.
19
For example, if threadprivate storage is accessed (explicitly in the source code or implicitly
20
in calls to library routines) in one part of a task region, its value cannot be assumed to be preserved
21
into the next part of the same task region if another schedulable ta |
opriate,
16
codeptr_ra may be NULL.
17
Cross References
18
• ompt_data_t, see Section 19.4.4.4
19
• flush directive, see Section 15.8.5
20
19.5.2.18 ompt_callback_cancel_t
21
Summary
22
The ompt_callback_cancel_t type is used for callbacks that are dispatched for cancellation,
23
cancel and discarded-task events.
24
Format
25
C / C++
typedef void (*ompt_callback_cancel_t) (
26
ompt_data_t *task_data,
27
int flags,
28
const void *codeptr_ra
29
);
30
C / C++
CHAPTER 19. OMPT INTERFACE
493
Trace Record
1
C / C++
typedef struct ompt_record_cancel_t {
2
ompt_id_t task_id;
3
int flags;
4
const void *codeptr_ra;
5
} ompt_record_cancel_t;
6
C / C++
Description of Arguments
7
The binding of the task_data argument is the task that encounters a cancel construct, a
8
cancellation point construct, or a construct defined as having an implicit cancellation
9
point.
10
The flags argument, defined by the ompt_cancel_flag_t enumeration type, indicates whether
11
cancellation is activated by the current task or detected as being act |
cs
13
The full clause specifies that the associated loop is fully unrolled. The construct is replaced by a
14
structured block that only contains n instances of its loop body, one for each of the n logical
15
iterations of the associated loop and in their logical iteration order.
16
Restrictions
17
Restrictions to the full clause are as follows:
18
• The iteration count of the associated loop must be a compile-time constant.
19
Cross References
20
• unroll directive, see Section 9.2
21
9.2.2 partial Clause
22
Name: partial
Properties: unique
23
Arguments
24
Name
Type
Properties
unroll-factor
expression of integer type
optional, constant, posi-
tive
25
Directives
26
unroll
27
CHAPTER 9. LOOP TRANSFORMATION CONSTRUCTS
221
Semantics
1
The partial clause specifies that the associated loop is first tiled with a tile size of unroll-factor.
2
Then, the generated tile loop is fully unrolled. If the partial clause is used without an
3
unroll-factor argument then the unroll factor is a positive integer that is implementat |
. . . . .
540
20.2
Activating a Third-Party Tool
. . . . . . . . . . . . . . . . . . . . . . . . . .
540
20.2.1
Enabling Runtime Support for OMPD . . . . . . . . . . . . . . . . . . . . .
540
20.2.2
ompd_dll_locations
. . . . . . . . . . . . . . . . . . . . . . . . . .
540
20.2.3
ompd_dll_locations_valid . . . . . . . . . . . . . . . . . . . . . .
541
20.3
OMPD Data Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
542
20.3.1
Size Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
542
20.3.2
Wait ID Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
542
20.3.3
Basic Value Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
542
20.3.4
Address Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
543
20.3.5
Frame Information Type . . . . . . . . . . . . . . . . . . . . . . . . . . . .
543
20.3.6
System Device Identifiers . . . . . . . . . . . . . . . . . . . . . . . . . . .
544
20.3.7
Native Thread Identifiers |
.1
8
• teams directive, see Section 10.2
9
21.6.2 OMP_TEAMS_THREAD_LIMIT
10
The OMP_TEAMS_THREAD_LIMIT environment variable sets the maximum number of OpenMP
11
threads to use in each contention group created by a teams construct by setting the
12
teams-thread-limit-var ICV. The value of this environment variable must be a positive integer. The
13
behavior of the program is implementation defined if the requested value of
14
OMP_TEAMS_THREAD_LIMIT is greater than the number of threads that an implementation can
15
support, or if the value is not a positive integer.
16
Cross References
17
• teams directive, see Section 10.2
18
• teams-thread-limit-var ICV, see Table 2.1
19
21.7 OMP_DISPLAY_ENV
20
The OMP_DISPLAY_ENV environment variable instructs the runtime to display the information as
21
described in the omp_display_env routine section (Section 18.15). The value of the
22
OMP_DISPLAY_ENV environment variable may be set to one of these values:
23
true | false | verbose
24
If the environment variable is set to |
e current team. The binding region of the
12
cancel region is the innermost enclosing region of the type that corresponds to
13
cancel-directive-name.
14
Semantics
15
The cancel construct activates cancellation of the innermost enclosing region of the type
16
specified by cancel-directive-name, which must be the directive-name of a cancellable construct.
17
Cancellation of the binding region is activated only if the cancel-var ICV is true, in which case the
18
cancel construct causes the encountering task to continue execution at the end of the binding
19
region if cancel-directive-name is not taskgroup. If the cancel-var ICV is true and
20
cancel-directive-name is taskgroup, the encountering task continues execution at the end of the
21
current task region. If the cancel-var ICV is false, the cancel construct is ignored.
22
Threads check for active cancellation only at cancellation points that are implied at the following
23
locations:
24
• cancel regions;
25
• cancellation point regions;
26
• barrier regions |
fies a wait on completion of the taskgroup set associated with the
2
taskgroup region. When a thread encounters a taskgroup construct, it starts executing the
3
region.
4
An implicit task scheduling point occurs at the end of the taskgroup region. The current task is
5
suspended at the task scheduling point until all tasks in the taskgroup set complete execution.
6
Execution Model Events
7
The taskgroup-begin event occurs in each thread that encounters the taskgroup construct on
8
entry to the taskgroup region.
9
The taskgroup-wait-begin event occurs when a task begins an interval of active or passive waiting
10
in a taskgroup region.
11
The taskgroup-wait-end event occurs when a task ends an interval of active or passive waiting and
12
resumes execution in a taskgroup region.
13
The taskgroup-end event occurs in each thread that encounters the taskgroup construct after the
14
taskgroup synchronization on exit from the taskgroup region.
15
Tool Callbacks
16
A thread dispatches a registered ompt_callback_sync_r |
loop transformation construct that appears inside a loop nest is replaced according to its
22
semantics before any loop can be associated with a loop-associated directive that is applied to the
23
loop nest. The depth of the loop nest is determined according to the loops in the loop nest, after any
24
such replacements have taken place. A loop counts towards the depth of the loop nest if it is a base
25
language loop statement or generated loop and it matches loop-nest while applying the production
26
rules for canonical loop nest form to the loop nest.
27
The canonical loop nest form allows the iteration count of all associated loops to be computed
28
before executing the outermost loop.
29
For any associated loop, the iteration count is computed as follows:
30
CHAPTER 4. BASE LANGUAGE FORMATS AND RESTRICTIONS
91
C / C++
• If var has a signed integer type and the var operand of test-expr after usual arithmetic
1
conversions has an unsigned integer type then the loop iteration count is computed from lb,
2
te |
17
implementation-specific one. For the schedule kinds static, dynamic, and guided, the
18
chunk_size is set to the value of the second argument, or to the default chunk_size if the value of the
19
second argument is less than 1; for the schedule kind auto, the second argument has no meaning;
20
for implementation-specific schedule kinds, the values and associated meanings of the second
21
argument are implementation defined.
22
Each of the schedule kinds can be combined with the omp_sched_monotonic modifier by
23
using the + or | operators in C/C++ or the + operator in Fortran. If the schedule kind is combined
24
with the omp_sched_monotonic modifier, the schedule is modified as if the monotonic
25
schedule modifier was specified. Otherwise, the schedule modifier is nonmonotonic.
26
Cross References
27
• run-sched-var ICV, see Table 2.1
28
18.2.12 omp_get_schedule
29
Summary
30
The omp_get_schedule routine returns the schedule that is applied when the runtime schedule
31
is used.
32
Format
33
C / C++
void omp_get_sch |
askloop-chunk-begin event in that thread.
11
The callback binds to the explicit task executing the iterations. The callback has type signature
12
ompt_callback_dispatch_t.
13
Restrictions
14
Restrictions to the taskloop construct are as follows:
15
• The reduction-modifier must be default.
16
• The conditional lastprivate-modifier must not be specified.
17
Cross References
18
• Canonical Loop Nest Form, see Section 4.4.1
19
• ompt_callback_dispatch_t, see Section 19.5.2.6
20
• ompt_callback_work_t, see Section 19.5.2.5
21
• ompt_scope_endpoint_t, see Section 19.4.4.11
22
• ompt_work_t, see Section 19.4.4.16
23
• allocate clause, see Section 6.6
24
• collapse clause, see Section 4.4.3
25
• default clause, see Section 5.4.1
26
• final clause, see Section 12.3
27
• firstprivate clause, see Section 5.4.4
28
• grainsize clause, see Section 12.6.1
29
• if clause, see Section 3.4
30
• in_reduction clause, see Section 5.5.10
31
• lastprivate clause, see Section 5.4.5
32
• mergeable clause, see Section 12.2
33
268
OpenMP |
ection 19.6.2.6
11
• ompt_stop_trace_t, see Section 19.6.2.9
12
• ompt_translate_time_t, see Section 19.6.2.3
13
19.3 Finalizing a First-Party Tool
14
If the OMPT interface state is active, the tool finalizer, which has type signature
15
ompt_finalize_t and is specified by the finalize field in the
16
ompt_start_tool_result_t structure returned from the ompt_start_tool function, is
17
called when the OpenMP implementation shuts down.
18
Cross References
19
• ompt_finalize_t, see Section 19.5.1.2
20
19.4 OMPT Data Types
21
The C/C++ header file (omp-tools.h) provides the definitions of the types that are specified
22
throughout this subsection.
23
19.4.1 Tool Initialization and Finalization
24
Summary
25
A tool’s implementation of ompt_start_tool returns a pointer to an
26
ompt_start_tool_result_t structure, which contains pointers to the tool’s initialization
27
and finalization callbacks as well as an ompt_data_t object for use by the tool.
28
CHAPTER 19. OMPT INTERFACE
451
Format
1
C / C++
typedef struct ompt_start |
• use clause, see Section 14.1.3
35
292
OpenMP API – Version 5.2 November 2021
14.1.1 OpenMP Foreign Runtime Identifiers
1
An OpenMP foreign runtime identifier, foreign-runtime-id, is a base language string literal or a
2
compile-time constant OpenMP integer expression. Allowed values for foreign-runtime-id include
3
the names (as string literals) and integer values that the OpenMP Additional Definitions document
4
specifies and the corresponding omp_ifr_name constants of OpenMP interop_fr type.
5
Implementation-defined values for foreign-runtime-id may also be supported.
6
14.1.2 init Clause
7
Name: init
Properties: default
8
Arguments
9
Name
Type
Properties
interop-var
variable of omp_interop_t type
default
10
Modifiers
11
Name
Modifies
Type
Properties
interop-preference
Generic
Complex, name:
prefer_type Arguments:
preference_list OpenMP
foreign runtime preference
list (default)
complex, unique
interop-type
Generic
Keyword: target,
targetsync
repeatable, re-
quired
12
Directives
13
interop
14
Semantics
15
The in |
ked region.
7
• A barrier region may not be closely nested inside a worksharing, task, taskloop,
8
critical, ordered, atomic, or masked region.
9
• A masked region may not be closely nested inside a worksharing, atomic, task, or
10
taskloop region.
11
• An ordered region that corresponds to an ordered construct without any clause or with the
12
threads or depend clause may not be closely nested inside a critical, ordered, loop,
13
atomic, task, or taskloop region.
14
• An ordered region that corresponds to an ordered construct without the simd clause
15
specified must be closely nested inside a worksharing-loop region.
16
• An ordered region that corresponds to an ordered construct with the simd clause specified
17
must be closely nested inside a simd or worksharing-loop SIMD region.
18
• An ordered region that corresponds to an ordered construct with both the simd and
19
threads clauses must be closely nested inside a worksharing-loop SIMD region or closely
20
nested inside a worksharing-loop and simd region.
|
e equivalent (the first line represents the position of the first 9 columns):
15
c23456789
16
!$ 10 iam = omp_get_thread_num() +
17
!$
&
index
18
19
#ifdef _OPENMP
20
10 iam = omp_get_thread_num() +
21
&
index
22
#endif
23
24
25
Fortran
70
OpenMP API – Version 5.2 November 2021
Fortran
3.3.2 Free Source Form Conditional Compilation Sentinel
1
The following conditional compilation sentinel is recognized in free form source files:
2
!$
3
To enable conditional compilation, a line with a conditional compilation sentinel must satisfy the
4
following criteria:
5
• The sentinel can appear in any column but must be preceded only by white space;
6
• The sentinel must appear as a single word with no intervening white space;
7
• Initial lines must have a blank character after the sentinel; and
8
• Continued lines must have an ampersand as the last non-blank character on the line, prior to any
9
comment appearing on the conditionally compiled line.
10
Continuation lines can have an ampersand after the sentinel, with optiona |
nt and the if clause expression
16
evaluates to false.
17
The teams construct creates a league of teams, where each team is an initial team that comprises
18
an initial thread that executes the teams region. Each initial thread executes sequentially, as if the
19
code encountered is part of an initial task region that is generated by an implicit parallel region
20
associated with each team. Whether the initial threads concurrently execute the teams region is
21
unspecified, and a program that relies on their concurrent execution for the purposes of
22
synchronization may deadlock.
23
If a construct creates a data environment, the data environment is created at the time the construct is
24
encountered. The description of a construct defines whether it creates a data environment.
25
When any thread encounters a parallel construct, the thread creates a team of itself and zero or
26
more additional threads and becomes the primary thread of the new team. A set of implicit tasks,
27
one per thread, is generated. The |
a reduction-scoping or lastprivate clause, the final values of
1
the list items that appeared in those clauses are undefined.
2
When an if clause is present on a cancel construct and the if expression evaluates to false, the
3
cancel construct does not activate cancellation. The cancellation point associated with the
4
cancel construct is always encountered regardless of the value of the if expression.
5
6
Note – The programmer is responsible for releasing locks and other synchronization data
7
structures that might cause a deadlock when a cancel construct is encountered and blocked
8
threads cannot be canceled. The programmer is also responsible for ensuring proper
9
synchronizations to avoid deadlocks that might arise from cancellation of OpenMP regions that
10
contain OpenMP synchronization constructs.
11
12
Execution Model Events
13
If a task encounters a cancel construct that will activate cancellation then a cancel event occurs.
14
A discarded-task event occurs for any discarded tasks.
15
Tool Callbacks
|
thread as its parent
5
thread with respect to any resources associated with an OpenMP thread.
6
child thread
When a thread encounters a parallel construct, each of the threads in the
7
generated parallel region’s team are child threads of the encountering thread.
8
The target or teams region’s initial thread is not a child thread of the thread that
9
encountered the target or teams construct.
10
ancestor thread
For a given thread, its parent thread or one of its parent thread’s ancestor threads.
11
descendent thread
For a given thread, one of its child threads or one of its child threads’ descendent
12
threads.
13
team
A set of one or more threads participating in the execution of a parallel region.
14
COMMENTS:
15
For an active parallel region, the team comprises the primary thread and
16
at least one additional thread.
17
For an inactive parallel region, the team comprises only the primary
18
thread.
19
league
The set of teams created by a teams construct.
20
contention group
An initial thread and its desc |
T: An array is a containing array of itself. For the array section
14
(*p0).x0[k1].p1->p2[k2].x1[k3].x2[4][0:n], where identifiers pi have a
15
pointer type declaration and identifiers xi have an array type declaration,
16
the containing arrays are: (*p0).x0[k1].p1->p2[k2].x1 and
17
(*p0).x0[k1].p1->p2[k2].x1[k3].x2.
18
containing structure
For C/C++, a structure to which a series of zero or more . (dot) operators and/or
19
array subscript operators are applied to yield a given lvalue expression or array
20
section for which storage is contained by the structure.
21
For Fortran, a structure to which a series of zero or more component selectors and/or
22
array subscript selectors are applied to yield a given variable or array section for
23
which storage is contained by the structure.
24
COMMENT: A structure is a containing structure of itself. For C/C++, a
25
structure pointer p to which the -> operator applies is equivalent to the
26
application of a . (dot) operator to (*p) for the purposes of determining
27
|
or more blanks or horizontal tabs are optional to separate adjacent keywords in
1
directive-names unless otherwise specified.
2
3
Note – In the following example the three formats for specifying the directive are equivalent (the
4
first line represents the position of the first 9 columns):
5
!23456789
6
!$omp parallel do &
7
!$omp shared(a,b,c)
8
9
!$omp parallel &
10
!$omp&do shared(a,b,c)
11
12
!$omp paralleldo shared(a,b,c)
13
14
Fortran
3.2 Clause Format
15
This section defines the format and categories of OpenMP clauses. OpenMP clauses are specified
16
as part of a directive-specification. Clauses are optional and, thus, may be omitted from a
17
directive-specification unless otherwise specified. The order in which clauses appear on directives
18
is not significant unless otherwise specified. A clause-specification specifies each OpenMP clause
19
in a directive-specification where clause-specification for inarguable clauses is simply:
20
clause-name
21
Inarguable clauses often form natural groupings that have similar |
OpenMP
Application Programming
Interface
Version 5.2 November 2021
Copyright c⃝1997-2021 OpenMP Architecture Review Board.
Permission to copy without fee all or part of this material is granted, provided the OpenMP
Architecture Review Board copyright notice and the title of this document appear. Notice is
given that copying is by permission of the OpenMP Architecture Review Board.
This page intentionally left blank in published version.
Contents
1
Overview of the OpenMP API
1
1.1
Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1.2
Glossary
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2
1.2.1
Threading Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2
1.2.2
OpenMP Language Terminology
. . . . . . . . . . . . . . . . . . . . . . .
2
1.2.3
Loop Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9
1.2.4
Synchronization Terminology . . . . . . . . . . . . . . . . . . . . . . . . .
10
1.2.5
Tasking Termino |
. . . . . .
301
15.3.1
barrier Construct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
301
15.3.2
Implicit Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
303
15.3.3
Implementation-Specific Barriers . . . . . . . . . . . . . . . . . . . . . . .
304
15.4
taskgroup Construct
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
304
15.5
taskwait Construct
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
306
15.6
nowait Clause
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
308
15.7
nogroup Clause . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
309
15.8
OpenMP Memory Ordering . . . . . . . . . . . . . . . . . . . . . . . . . . .
309
15.8.1
memory-order Clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
309
15.8.2
atomic Clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
310
15.8.3
extended-atomic Clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . .
310
15.8.4
atomic Construct |
d ICV of the device with the corresponding device
21
number. If the corresponding environment variable that includes the _DEV suffix but no device
22
number is set, then the setting of that environment variable is used to set the value of the associated
23
ICV of any non-host device for which the device-number-specific corresponding environment
24
variable is not set. In all cases the setting of an environment variable for which a device number is
25
specified takes precedence.
26
Restrictions
27
Restrictions to device-specific environment variables are as follows:
28
• Device-specific environment variables must not correspond to environment variables that
29
initialize ICVs with global scope.
30
599
21.1 Parallel Region Environment Variables
1
This section defines environment variables that affect the operation of parallel regions.
2
21.1.1 OMP_DYNAMIC
3
The OMP_DYNAMIC environment variable controls dynamic adjustment of the number of threads
4
to use for executing parallel regions by setting the initial value of th |
11
omp_ifr_name, where name is the name of the foreign runtime environment. Available names
12
include those that are listed in the OpenMP Additional Definitions document;
13
implementation-defined names may also be supported. The value of omp_ifr_last is defined as
14
one greater than the value of the highest supported foreign-runtime-id value that is listed in the
15
aforementioned document.
16
Cross References
17
• Interoperability Routines, see Section 18.12
18
14.1 interop Construct
19
Name: interop
Association: none
Category: executable
Properties: device
20
Clauses
21
depend, destroy, device, init, nowait, use
22
Clause set
action-clause
23
Properties: required
Members: destroy, init, use
24
CHAPTER 14. INTEROPERABILITY
291
Binding
1
The binding task set for an interop region is the generating task. The interop region binds to
2
the region of the generating task.
3
Semantics
4
The interop construct retrieves interoperability properties from the OpenMP implementation to
5
enable interoperability with fore |
etadirective.
21
Restrictions
22
Restrictions to metadirectives are as follows:
23
• Replacement of the metadirective with the directive variant associated with any of the dynamic
24
replacement candidates must result in a conforming OpenMP program.
25
• Insertion of user code at the location of a metadirective must be allowed if the first dynamic
26
replacement candidate does not have a static context selector.
27
• All items must be executable directives if the first dynamic replacement candidate does not have
28
a static context selector.
29
Fortran
• A metadirective that appears in the specification part of a subprogram must follow all
30
variant-generating declarative directives that appear in the same specification part.
31
• All directive variants of a metadirective must be pure otherwise the metadirective is not pure.
32
Fortran
CHAPTER 7. VARIANT DIRECTIVES
189
7.4.1 when Clause
1
Name: when
Properties: default
2
Arguments
3
Name
Type
Properties
directive-variant
directive-specification
optional, unique
4 |
es that a directive accepts may form sets. These sets may imply restrictions on their use
22
on that directive or may otherwise capture properties for the clauses on the directive. While specific
23
properties may be defined for a clause set on a particular directive, the following clause-set
24
properties have general meanings and implications as indicated by the restrictions below: required,
25
unique, and exclusive.
26
58
OpenMP API – Version 5.2 November 2021
All clauses that are specified as a clause grouping form a clause set for which properties are
1
specified with the specification of the grouping. Some directives accept a clause grouping for which
2
each member is a directive-name of a directive that has a specific property. These groupings are
3
required, unique and exclusive unless otherwise specified.
4
Restrictions
5
Restrictions to clauses and clause sets are as follows:
6
• A required clause for a directive must appear on the directive.
7
• A unique clause for a directive may appear at most once on t |
ch callback for each occurrence of a
19
distribute-chunk-begin event in that thread. The callback occurs in the context of the initial task.
20
The callback has type signature ompt_callback_dispatch_t.
21
Restrictions
22
Restrictions to the distribute construct are as follows:
23
• The logical iteration space of the loops associated with the distribute construct must be the
24
same for all teams in the league.
25
• The region that corresponds to the distribute construct must be strictly nested inside a
26
teams region.
27
• A list item may appear in a firstprivate or lastprivate clause, but not in both.
28
• The conditional lastprivate-modifier must not be specified.
29
Cross References
30
• Consistent Loop Schedules, see Section 4.4.5
31
• ompt_callback_work_t, see Section 19.5.2.5
32
• ompt_work_t, see Section 19.4.4.16
33
• allocate clause, see Section 6.6
34
• collapse clause, see Section 4.4.3
35
• dist_schedule clause, see Section 11.6.1
36
CHAPTER 11. WORK-DISTRIBUTION CONSTRUCTS
255
• firstprivate claus |
. . . . . . . . . . . .
607
21.2.5
OMP_AFFINITY_FORMAT . . . . . . . . . . . . . . . . . . . . . . . . . .
608
21.2.6
OMP_CANCELLATION . . . . . . . . . . . . . . . . . . . . . . . . . . . .
610
21.2.7
OMP_DEFAULT_DEVICE
. . . . . . . . . . . . . . . . . . . . . . . . . .
610
21.2.8
OMP_TARGET_OFFLOAD
. . . . . . . . . . . . . . . . . . . . . . . . . .
610
xiv
OpenMP API – Version 5.2 November 2021
21.2.9
OMP_MAX_TASK_PRIORITY . . . . . . . . . . . . . . . . . . . . . . . .
611
21.3
OMPT Environment Variables . . . . . . . . . . . . . . . . . . . . . . . . . .
611
21.3.1
OMP_TOOL
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
611
21.3.2
OMP_TOOL_LIBRARIES
. . . . . . . . . . . . . . . . . . . . . . . . . .
612
21.3.3
OMP_TOOL_VERBOSE_INIT . . . . . . . . . . . . . . . . . . . . . . . .
612
21.4
OMPD Environment Variables . . . . . . . . . . . . . . . . . . . . . . . . . .
613
21.4.1
OMP_DEBUG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
613
21.5
Memory Allocation En |
is set to true, the thread affinity policy is implementation defined but
26
must conform to the previous paragraph. The behavior of the program is implementation defined if
27
the value in the OMP_PROC_BIND environment variable is not true, false, or a comma
28
604
OpenMP API – Version 5.2 November 2021
separated list of primary, master (master has been deprecated), close, or spread. The
1
behavior is also implementation defined if an initial thread cannot be bound to the first place in the
2
place-partition-var ICV.
3
The OMP_PROC_BIND environment variable sets the max-active-levels-var ICV to the number of
4
active levels of parallelism that the implementation supports if the OMP_PROC_BIND environment
5
variable is set to a comma-separated list of more than one element. The value of the
6
max-active-level-var ICV may be overridden by setting OMP_MAX_ACTIVE_LEVELS or
7
OMP_NESTED. See Section 21.1.4 and Section 21.1.5 for details.
8
Examples:
9
setenv OMP_PROC_BIND false
10
setenv OMP_PROC_BIND "spread, spread, cl |
allback_nest_lock_t, see Section 19.5.2.16
18
18.10 Timing Routines
19
This section describes routines that support a portable wall clock timer.
20
18.10.1 omp_get_wtime
21
Summary
22
The omp_get_wtime routine returns elapsed wall clock time in seconds.
23
Format
24
C / C++
double omp_get_wtime(void);
25
C / C++
Fortran
double precision function omp_get_wtime()
26
Fortran
Binding
27
The binding thread set for an omp_get_wtime region is the encountering thread. The routine’s
28
return value is not guaranteed to be consistent across any set of threads.
29
CHAPTER 18. RUNTIME LIBRARY ROUTINES
413
Effect
1
The omp_get_wtime routine returns a value equal to the elapsed wall clock time in seconds
2
since some time-in-the-past. The actual time-in-the-past is arbitrary, but it is guaranteed not to
3
change during the execution of the application program. The time returned is a per-thread time, so
4
it is not required to be globally consistent across all threads that participate in an application.
5
18.10.2 omp_get_wt |
for fixed source form Fortran to accommodate character
4
position requirements (see Section 3.1.1). Reserved clause names that begin with the ompx_
5
prefix for implementation-defined clauses on OpenMP directives (see Section 3.2). Reserved
6
names in the base language that start with the omp_ and ompx_ prefix and reserved the omp and
7
ompx namespaces (see Chapter 4) for the OpenMP runtime API and for implementation-defined
8
extensions to that API (see Chapter 18).
9
• Allowed any clause that can be specified on a paired end directive to be specified on the
10
directive (see Section 3.1), including the copyprivate clause (see Section 5.7.2) and the
11
nowait clause in Fortran (see Section 15.6).
12
• For consistency with the syntax of other definitions of the clause, the syntax of the destroy
13
clause on the depobj construct with no argument was deprecated (see Section 3.5).
14
• For consistency with the syntax of other clauses, the syntax of the linear clause that specifies
15
its argument and linear-modifier as li |
30
• Cross-iteration dependences across different logical iterations must not exist, except for
31
dependences for the list items specified in an inclusive or exclusive clause.
32
• Intra-iteration dependences from a statement in the structured block sequence that precede a
33
scan directive to a statement in the structured block sequence that follows a scan directive
34
must not exist, except for dependences for the list items specified in an inclusive or
35
exclusive clause.
36
• The private copy of list items that appear in the inclusive or exclusive clause must not be
37
modified in the scan phase.
38
142
OpenMP API – Version 5.2 November 2021
Cross References
1
• do directive, see Section 11.5.2
2
• exclusive clause, see Section 5.6.2
3
• for directive, see Section 11.5.1
4
• inclusive clause, see Section 5.6.1
5
• reduction clause, see Section 5.5.8
6
• simd directive, see Section 10.4
7
5.6.1 inclusive Clause
8
Name: inclusive
Properties: unique
9
Arguments
10
Name
Type
Properties
list
list of variable lis |
ffect of the map clauses on a construct would assign the value of a corresponding list item to
12
an original list item more than once, then an implementation is allowed to ignore additional
13
assignments of the same value to the original list item.
14
In all cases on exit from the region, concurrent reads or updates of any part of the original list item
15
must be synchronized with any update of the original list item that occurs as a result of the map
16
clause to avoid data races.
17
If a single contiguous part of the original storage of a list item with an implicit data-mapping
18
attribute has corresponding storage in the device data environment prior to a task encountering the
19
construct on which the map clause appears, only that part of the original storage will have
20
corresponding storage in the device data environment as a result of the map clause.
21
If a list item with an implicit data-mapping attribute does not have any corresponding storage in the
22
device data environment prior to a task en |
not be encountered during execution of a target region.
23
• The result of an omp_set_default_device, omp_get_default_device, or
24
omp_get_num_devices routine called within a target region is unspecified.
25
• The effect of an access to a threadprivate variable in a target region is unspecified.
26
• If a list item in a map clause is a structure element, any other element of that structure that is
27
referenced in the target construct must also appear as a list item in a map clause.
28
• A list item in a data-sharing attribute clause that is specified on a target construct must not
29
have the same base variable as a list item in a map clause on the construct.
30
• A variable referenced in a target region but not the target construct that is not declared in
31
the target region must appear in a declare target directive.
32
• A map-type in a map clause must be to, from, tofrom or alloc.
33
• If a device clause is specified with the ancestor device-modifier, only the device,
34
firstprivate, private, defaultmap, and |
tine are as follows.
2
• Freeing the storage returned by omp_target_alloc with any routine other than
3
omp_target_free results in unspecified behavior.
4
• When called from within a target region the effect is unspecified.
5
C / C++
• Unless the unified_address clause appears on a requires directive in the compilation
6
unit, pointer arithmetic is not supported on the device pointer returned by
7
omp_target_alloc.
8
C / C++
Cross References
9
• omp_target_free, see Section 18.8.2
10
• ompt_callback_target_data_op_emi_t and
11
ompt_callback_target_data_op_t, see Section 19.5.2.25
12
• is_device_ptr clause, see Section 5.4.7
13
• target directive, see Section 13.8
14
18.8.2 omp_target_free
15
Summary
16
The omp_target_free routine frees the device memory allocated by the
17
omp_target_alloc routine.
18
Format
19
C / C++
void omp_target_free(void *device_ptr, int device_num);
20
C / C++
Fortran
subroutine omp_target_free(device_ptr, device_num) bind(c)
21
use, intrinsic :: iso_c_binding, only : c_ptr, c_int
22
typ |
n, OpenMP mapper identifiers specify the name of a user-defined mapper, and
14
OpenMP foreign runtime identifiers specify the name of a foreign runtime.
15
Generic OpenMP types specify the type of expression or variable that is used in OpenMP contexts
16
regardless of the base language. These types support the definition of many important OpenMP
17
concepts independently of the base language in which they are used.
18
The assignable OpenMP type instance is defined to facilitate base language neutrality. An
19
assignable OpenMP type instance can be used as an argument of an OpenMP construct in order for
20
the implementation to modify the value of that instance.
21
C / C++
An assignable OpenMP type instance is an lvalue expression of that OpenMP type.
22
C / C++
Fortran
An assignable OpenMP type instance is a variable of that OpenMP type.
23
Fortran
74
OpenMP API – Version 5.2 November 2021
The OpenMP logical type supports logical variables and expressions in any base language.
1
C / C++
Any OpenMP logical expressi |
_callback_thread_begin callback for the
13
initial-thread-begin event in an initial thread. The callback occurs in the context of the initial
14
thread. The callback has type signature ompt_callback_thread_begin_t. The callback
15
receives ompt_thread_initial as its thread_type argument.
16
A thread dispatches a registered ompt_callback_implicit_task callback with
17
ompt_scope_begin as its endpoint argument for each occurrence of an initial-task-begin event
18
in that thread. Similarly, a thread dispatches a registered ompt_callback_implicit_task
19
callback with ompt_scope_end as its endpoint argument for each occurrence of an
20
initial-task-end event in that thread. The callbacks occur in the context of the initial task and have
21
type signature ompt_callback_implicit_task_t. In the dispatched callback,
22
(flag & ompt_task_initial) always evaluates to true.
23
A thread dispatches a registered ompt_callback_thread_end callback for the
24
initial-thread-end event in that thread. The callback occurs in the |
a corresponding list of their respective sizes.
7
Format
8
C
ompd_rc_t ompd_get_device_thread_id_kinds(
9
ompd_address_space_handle_t *device_handle,
10
ompd_thread_id_t **kinds,
11
ompd_size_t **thread_id_sizes,
12
int *count
13
);
14
C
Semantics
15
The ompd_get_device_thread_id_kinds function returns an array of supported native
16
thread identifier kinds and a corresponding array of their respective sizes for a given device. The
17
OMPD library allocates storage for the arrays with the memory allocation callback that the tool
18
provides. Each supported native thread identifier kind is guaranteed to be recognizable by the
19
OMPD library and may be mapped to and from any OpenMP thread that executes on the device.
20
The third-party tool owns the storage for the array of kinds and the array of sizes that is returned via
21
the kinds and thread_id_sizes arguments, and it is responsible for freeing that storage.
22
Description of Arguments
23
The device_handle argument is a pointer to an opaque address space h |
hes a registered ompt_callback_implicit_task callback with
1
ompt_scope_begin as its endpoint argument for each occurrence of an implicit-task-begin
2
event in that thread. Similarly, a thread dispatches a registered
3
ompt_callback_implicit_task callback with ompt_scope_end as its endpoint
4
argument for each occurrence of an implicit-task-end event in that thread. The callbacks occur in
5
the context of the implicit task and have type signature ompt_callback_implicit_task_t.
6
In the dispatched callback, (flags & ompt_task_implicit) evaluates to true.
7
A thread dispatches a registered ompt_callback_parallel_end callback for each
8
occurrence of a parallel-end event in that thread. The callback occurs in the task that encounters
9
the parallel construct. This callback has the type signature
10
ompt_callback_parallel_end_t.
11
A thread dispatches a registered ompt_callback_thread_begin callback for the
12
native-thread-begin event in that thread. The callback occurs in the context of the thread. The
13
callba |
emi_t and
21
ompt_callback_target_data_op_t
22
Summary
23
The ompt_callback_target_data_op_emi_t and
24
ompt_callback_target_data_op_t types are used for callbacks that are dispatched when
25
a thread maps data to a device.
26
Format
27
C / C++
typedef void (*ompt_callback_target_data_op_emi_t) (
28
ompt_scope_endpoint_t endpoint,
29
ompt_data_t *target_task_data,
30
ompt_data_t *target_data,
31
ompt_id_t *host_op_id,
32
ompt_target_data_op_t optype,
33
void *src_addr,
34
int src_device_num,
35
CHAPTER 19. OMPT INTERFACE
499
void *dest_addr,
1
int dest_device_num,
2
size_t bytes,
3
const void *codeptr_ra
4
);
5
typedef void (*ompt_callback_target_data_op_t) (
6
ompt_id_t target_id,
7
ompt_id_t host_op_id,
8
ompt_target_data_op_t optype,
9
void *src_addr,
10
int src_device_num,
11
void *dest_addr,
12
int dest_device_num,
13
size_t bytes,
14
const void *codeptr_ra
15
);
16
C / C++
Trace Record
17
C / C++
typedef struct ompt_record_target_data_op_t {
18
ompt_id_t host_op_id;
19
ompt_target_data_op_t optype;
20
v |
format that is used to
12
summarize native device trace records.
13
Format
14
C / C++
typedef struct ompt_record_abstract_t {
15
ompt_record_native_t rclass;
16
const char *type;
17
ompt_device_time_t start_time;
18
ompt_device_time_t end_time;
19
ompt_hwid_t hwid;
20
} ompt_record_abstract_t;
21
C / C++
Semantics
22
An ompt_record_abstract_t record contains information that a tool can use to process a
23
native record that it may not fully understand. The rclass field indicates that the record is
24
informational or that it represents an event; this information can help a tool determine how to
25
present the record. The record type field points to a statically-allocated, immutable character string
26
that provides a meaningful name that a tool can use to describe the event to a user. The start_time
27
and end_time fields are used to place an event in time. The times are relative to the device clock. If
28
an event does not have an associated start_time (end_time), the value of the start_time (end_time)
29
field |
;
27
332
• at the end of a worksharing-loop construct with a nowait clause and for which the same list
1
item appears in both firstprivate and lastprivate clauses; and
2
• implicit barrier regions.
3
When a thread reaches one of the above cancellation points and if the cancel-var ICV is true, then:
4
• If the thread is at a cancel or cancellation point region and cancel-directive-name is
5
not taskgroup, the thread continues execution at the end of the canceled region if cancellation
6
has been activated for the innermost enclosing region of the type specified.
7
• If the thread is at a cancel or cancellation point region and cancel-directive-name is
8
taskgroup, the encountering task checks for active cancellation of all of the taskgroup sets to
9
which the encountering task belongs, and continues execution at the end of the current task
10
region if cancellation has been activated for any of the taskgroup sets.
11
• If the encountering task is at a barrier region or at the end of a worksharing-loop construct |
ive.
28
198
OpenMP API – Version 5.2 November 2021
If a declare variant directive appears between a begin declare variant directive and its
1
paired end directive, the effective context selectors of the outer directive are appended to the
2
context selector of the inner directive to form the effective context selector of the inner directive. If
3
a trait-set-selector is present on both directives, the trait-selector list of the outer directive is
4
appended to the trait-selector list of the inner directive after equivalent trait-selectors have been
5
removed from the outer list. Restrictions that apply to explicitly specified context selectors also
6
apply to effective context selectors constructed through this process.
7
The symbol name of a function definition that appears between a begin declare variant
8
directive and its paired end directive is determined through the base language rules after the name
9
of the function has been augmented with a string that is determined according to the effective
10
context se |
ding to a parallel construct or a task region
26
corresponding to a task construct.
27
COMMENTS:
28
A sequential part is enclosed by an implicit parallel region.
29
Executable statements in called routines may be in both a sequential part
30
and any number of explicit parallel regions at different points in the
31
program execution.
32
primary thread
An OpenMP thread that has thread number 0. A primary thread may be an initial
33
thread or the thread that encounters a parallel construct, creates a team,
34
generates a set of implicit tasks, and then executes one of those tasks as thread
35
number 0.
36
CHAPTER 1. OVERVIEW OF THE OPENMP API
5
worker thread
An OpenMP thread that is not the primary thread of a team and that executes one of
1
the implicit tasks of a parallel region.
2
parent thread
The thread that encountered the parallel construct and generated a parallel
3
region is the parent thread of each of the threads in the team of that parallel
4
region. The primary thread of a parallel region is the same |
Arguments
10
Name
Type
Properties
list
list of variable list item type
default
11
Directives
12
dispatch, target
13
Semantics
14
The is_device_ptr clause indicates that its list items are device pointers. Support for device
15
pointers created outside of OpenMP, specifically outside of any OpenMP mechanism that returns a
16
device pointer, is implementation defined.
17
If the is_device_ptr clause is specified on a target construct, each list item privatized
18
inside the construct and the new list item is initialized to the device address to which the original
19
list item refers.
20
Fortran
If the is_device_ptr clause is specified on a target construct, if any list item is not of type
21
C_PTR, the behavior is as if the list item appeared in a has_device_addr clause. Support for
22
such list items in an is_device_ptr clause is deprecated.
23
Fortran
Restrictions
24
Restrictions to the is_device_ptr clause are as follows:
25
• Each list item must be a valid device pointer for the device data environment.
26
C
• |
er that uses OMPD should be able to
14
debug a 64-bit OpenMP program by loading a 32-bit OMPD implementation that can manage a
15
64-bit OpenMP runtime.
16
The ompd_dll_locations variable points to a NULL-terminated vector of zero or more
17
null-terminated pathname strings that do not have any filename conventions. This vector must be
18
fully initialized before ompd_dll_locations is set to a non-null value. Thus, if a third-party
19
tool, such as a debugger, stops execution of the OpenMP program at any point at which
20
ompd_dll_locations is non-null, the vector of strings to which it points shall be valid and
21
complete.
22
Cross References
23
• ompd_dll_locations_valid, see Section 20.2.3
24
20.2.3 ompd_dll_locations_valid
25
Summary
26
The OpenMP runtime notifies third-party tools that ompd_dll_locations is valid by allowing
27
execution to pass through a location that the symbol ompd_dll_locations_valid identifies.
28
Format
29
C
void ompd_dll_locations_valid(void);
30
C
Semantics
31
Since ompd_dll_locati |
t, a colon (:) or an
25
asterisk (*).
26
• If a type with deferred or assumed length parameter is specified in a declare reduction
27
directive, no other declare reduction directive with the same type, the same kind
28
parameters and the same reduction-identifier is allowed in the same scope.
29
Fortran
140
OpenMP API – Version 5.2 November 2021
Cross References
1
• OpenMP Combiner Expressions, see Section 5.5.2.1
2
• OpenMP Initializer Expressions, see Section 5.5.2.2
3
• OpenMP Reduction Identifiers, see Section 5.5.1
4
• initializer clause, see Section 5.5.4
5
5.6 scan Directive
6
Name: scan
Association: separating
Category: subsidiary
Properties: default
7
Separated directives
8
do, for, simd
9
Clauses
10
exclusive, inclusive
11
Clause set
12
Properties: unique, required, exclusive
Members: exclusive, inclusive
13
Semantics
14
The scan directive separates the final-loop-body of an enclosing simd construct or
15
worksharing-loop construct (or a composite construct that combines them) into a structured block
16 |
vars argument is the vector of display control variables to be released.
6
Description of Return Codes
7
This routine must return any of the general return codes listed at the beginning of Section 20.5.
8
Cross References
9
• Return Code Types, see Section 20.3.12
10
• ompd_get_display_control_vars, see Section 20.5.9.1
11
20.5.10 Accessing Scope-Specific Information
12
20.5.10.1 ompd_enumerate_icvs
13
Summary
14
The ompd_enumerate_icvs function enumerates ICVs.
15
Format
16
C
ompd_rc_t ompd_enumerate_icvs (
17
ompd_address_space_handle_t *handle,
18
ompd_icv_id_t current,
19
ompd_icv_id_t *next_id,
20
const char **next_icv_name,
21
ompd_scope_t *next_scope,
22
int *more
23
);
24
C
Semantics
25
An OpenMP implementation must support all ICVs listed in Section 2.1. An OpenMP
26
implementation may support additional implementation-specific variables. An implementation may
27
store ICVs in a different scope than Table 2.1 indicates. The ompd_enumerate_icvs function
28
enables a tool to enumerate the ICVs that an Ope |
libraries, separated by an implementation specific, platform typical separator. Whether the value of
5
this environment variable is case sensitive is implementation defined.
6
If the tool-var ICV is not enabled, the value of tool-libraries-var is ignored. Otherwise, if
7
ompt_start_tool is not visible in the address space on a device where OpenMP is being
8
initialized or if ompt_start_tool returns NULL, an OpenMP implementation will consider
9
libraries in the tool-libraries-var list in a left-to-right order. The OpenMP implementation will
10
search the list for a library that meets two criteria: it can be dynamically loaded on the current
11
device and it defines the symbol ompt_start_tool. If an OpenMP implementation finds a
12
suitable library, no further libraries in the list will be considered.
13
Example:
14
% setenv OMP_TOOL_LIBRARIES libtoolXY64.so:/usr/local/lib/
15
libtoolXY32.so
16
Cross References
17
• OMPT Interface, see Chapter 19
18
• ompt_start_tool, see Section 19.2.1
19
• tool-libraries-var ICV |
evices or both the host device and non-host devices. If
11
host is specified then only a host device version of the procedure or variable is made available. If
12
any is specified then both host device and non-host device versions of the procedure or variable are
13
made available. If nohost is specified for a procedure then only non-host device versions of the
14
procedure are made available. If nohost is specified for a variable then that variable is not
15
available on the host device. If the device_type clause is not specified, the behavior is as if the
16
device_type clause appears with any specified.
17
Cross References
18
• begin declare target directive, see Section 7.8.2
19
• declare target directive, see Section 7.8.1
20
275
13.2 device Clause
1
Name: device
Properties: unique
2
Arguments
3
Name
Type
Properties
device-description
expression of integer type
default
4
Modifiers
5
Name
Modifies
Type
Properties
device-modifier
device-description
Keyword: ancestor,
device_num
default
6
Directives
7
dispatch, inte |
containing structures.
28
For the array section (*p0).x0[k1].p1->p2[k2].x1[k3].x2[4][0:n], where
29
identifiers pi have a pointer type declaration and identifiers xi have an
30
array type declaration, the containing structures are: *(*p0).x0[k1].p1,
31
(*(*p0).x0[k1].p1).p2[k2] and (*(*p0).x0[k1].p1).p2[k2].x1[k3]
32
base array
For C/C++, a containing array of a given lvalue expression or array section that does
33
not appear in the expression of any of its other containing arrays.
34
For Fortran, a containing array of a given variable or array section that does not
35
appear in the designator of any of its other containing arrays.
36
COMMENT: For the array section
37
(*p0).x0[k1].p1->p2[k2].x1[k3].x2[4][0:n], where identifiers pi have a
38
CHAPTER 1. OVERVIEW OF THE OPENMP API
15
pointer type declaration and identifiers xi have an array type declaration,
1
the base array is: (*p0).x0[k1].p1->p2[k2].x1[k3].x2.
2
named array
For C/C++, a containing array of a given lvalue expression or array section, or a
3
contai |
ime API in its corresponding region, referred to as
12
intervening code. If intervening code is present, then a loop at the same depth within
13
the loop nest is not a perfectly nested loop.
14
C / C++
It must not contain iteration statements, continue statements or break statements
15
that apply to the enclosing loop.
16
C / C++
Fortran
It must not contain loops, array expressions, CYCLE statements or EXIT statements.
17
Fortran
final-loop-body
A structured block that terminates the scope of loops in the loop nest. If the loop nest
18
is associated with a loop-associated directive, loops in this structured block cannot be
19
associated with that directive.
20
C / C++
init-expr
One of the following:
21
var = lb
22
integer-type var = lb
23
CHAPTER 4. BASE LANGUAGE FORMATS AND RESTRICTIONS
87
C
pointer-type var = lb
1
C
C++
random-access-iterator-type var = lb
2
C++
test-expr
One of the following:
3
var relational-op ub
4
ub relational-op var
5
relational-op
One of the following:
6
<
7
<=
8
>
9
>=
10
!=
11
incr- |
ration or definition
11
enables the creation of corresponding SIMD versions of the associated function that can be used to
12
process multiple arguments from a single invocation in a SIMD loop concurrently.
13
If a SIMD version is created and the simdlen clause is not specified, the number of concurrent
14
arguments for the function is implementation defined.
15
For purposes of the linear clause, any integer-typed parameter that is specified in a uniform
16
clause on the directive is considered to be constant and so may be used in linear-step.
17
C / C++
The expressions that appear in the clauses of each directive are evaluated in the scope of the
18
arguments of the function declaration or definition.
19
C / C++
C++
The special this pointer can be used as if it was one of the arguments to the function in any of the
20
linear, aligned, or uniform clauses.
21
C++
Restrictions
22
Restrictions to the declare simd directive are as follows:
23
• The function or subroutine body must be a structured block.
24
• The execu |
edule(omp_sched_t *kind, int *chunk_size);
34
C / C++
356
OpenMP API – Version 5.2 November 2021
Fortran
subroutine omp_get_schedule(kind, chunk_size)
1
integer (kind=omp_sched_kind) kind
2
integer chunk_size
3
Fortran
Binding
4
The binding task set for an omp_get_schedule region is the generating task.
5
Effect
6
This routine returns the run-sched-var ICV in the task to which the routine binds. The first
7
argument kind returns the schedule to be used. It can be any of the standard schedule kinds as
8
defined in Section 18.2.11, or any implementation-specific schedule kind. If the returned schedule
9
kind is static, dynamic, or guided, the second argument chunk_size returns the chunk size to
10
be used, or a value less than 1 if the default chunk size is to be used. The value returned by the
11
second argument is implementation defined for any other schedule kinds.
12
Cross References
13
• run-sched-var ICV, see Table 2.1
14
18.2.13 omp_get_thread_limit
15
Summary
16
The omp_get_thread_limit routine returns the |
ment is relaxed. For
31
any other effective memory ordering, the default argument is equal to that effective memory
32
ordering. The weak clause specifies that the comparison performed by a conditional atomic update
33
may spuriously fail, evaluating to not equal even when the values are equal.
34
310
OpenMP API – Version 5.2 November 2021
1
Note – Allowing for spurious failure by specifying a weak clause can result in performance gains
2
on some systems when using compare-and-swap in a loop. For cases where a single
3
compare-and-swap would otherwise be sufficient, using a loop over a weak compare-and-swap is
4
unlikely to improve performance.
5
6
Restrictions
7
Restrictions to the atomic construct are as follows:
8
• acq_rel and release cannot be specified as arguments to the fail clause.
9
Cross References
10
• atomic Clauses, see Section 15.8.2
11
• atomic directive, see Section 15.8.4
12
• memory-order Clauses, see Section 15.8.1
13
15.8.4 atomic Construct
14
Name: atomic
Association: block (atomic structured |
t are dispatched when masked
32
regions start and end.
33
486
OpenMP API – Version 5.2 November 2021
Format
1
C / C++
typedef void (*ompt_callback_masked_t) (
2
ompt_scope_endpoint_t endpoint,
3
ompt_data_t *parallel_data,
4
ompt_data_t *task_data,
5
const void *codeptr_ra
6
);
7
C / C++
Trace Record
8
C / C++
typedef struct ompt_record_masked_t {
9
ompt_scope_endpoint_t endpoint;
10
ompt_id_t parallel_id;
11
ompt_id_t task_id;
12
const void *codeptr_ra;
13
} ompt_record_masked_t;
14
C / C++
Description of Arguments
15
The endpoint argument indicates that the callback signals the beginning of a scope or the end of a
16
scope.
17
The binding of the parallel_data argument is the current parallel region.
18
The binding of the task_data argument is the encountering task.
19
The codeptr_ra argument relates the implementation of an OpenMP region to its source code. If a
20
runtime routine implements the region associated with a callback that has type signature
21
ompt_callback_masked_t then codeptr_ra contains the |
End of preview. Expand
in Data Studio
YAML Metadata Warning: empty or missing yaml metadata in repo card
Check out the documentation for more information.
OpenMP Spec. Version 5.2. Not cleaned for line numbers or anything.
dataset_info: features: - name: text dtype: string splits: - name: train num_bytes: 1177273 num_examples: 1131 - name: test num_bytes: 131158 num_examples: 126 download_size: 620663 dataset_size: 1308431 configs: - config_name: default data_files: - split: train path: data/train-* - split: test path: data/test-*
- Downloads last month
- 5