| Doc. no. | P0024 |
| Date: | 2015-09-25 |
| Project: | Programming Language C++ |
| Reply to: | Jared Hoberock <jhoberock@nvidia.com> |
We survey implementation experience with the recently published C++ Technical Specification for Extensions for Parallelism and conclude that ample experience with its functionality exists to justify standardization in C++17. This paper describes various existing and pre-existing implementations of the TS's content and describes the additions to be made to the current C++ working paper (N4527) to integrate execution policies and parallel algorithms into the C++ Standard Library.
The technical content of the Parallelism TS was developed by domain experts in parallelism over the course of a few years. In 2012, representatives from NVIDIA (N3408) as well as representiatives from Microsoft and Intel (N3429) independently proposed library approaches to parallelism within the C++ Standard Library. At the suggestion of SG1 the authors of these proposals submitted a design in a joint proposal (N3554) to parallelize the existing standard algorithms library. This proposal was refined into the Parallelism TS over the course of two years. During that refinement process, the authors of the Parallelism TS incorporated feedback from experimental implementations into the final design which was published in 2015. In total, the C++ Standardization Committee has three years of experience with the TS's design.
Several different implementations of the Parallelism TS emerged during its preparation. We are aware of the following publically documented implementations.
These implementations implement the functionality of the Parallelism TS to
varying degrees and in different ways. For example, Microsoft's
implementation appears complete and is implemented via Windows-specific
tasking facilities. Thibaut Lutz' version also appears complete and is
implemented by manipulating std::thread in a standard way.
NVIDIA's implementation is partial and is implemented as a thin wrapper
around Thrust, a pre-existing library similar in content to the Parallelism
TS. This variety of implementation approaches exists by design: the
abstractions of the Parallelism TS are intended to maximize flexibility of
implementation.
The design of the Parallelism TS's functionality was inspired by several pre-existing libraries. Each of the following parallel algorithms libraries expose an iterator-based algorithm interface based on the conventions of the original Standard Template Library. We believe these libraries are a reasonable proxy for the content of the Parallelism TS.
These libraries have existed for several years, and some are widely deployed in production. Accordingly, we believe the features of the Parallelism TS are proven abstractions that represent standard practice and solve real challenges faced by real C++ programmers. These challenges exist because parallel architectures are so pervasive, and programming them correctly with existing low-level standard components is difficult. As a remedy, we believe that the high-level abstractions of the Parallelism TS must be standardized as soon as possible. C++ programmers should not have to wait beyond 2017 for standard parallel algorithms.
The parallel algorithms and execution policies of the Parallelism TS are only a starting point. Already we anticipate opportunities for extending the Parallelism TS's functionality to increase programmer flexibility and expressivity. A fully-realized executors feature (N4414, N4406) will yield new, flexible ways of creating execution, including the execution of parallel algorithms. For example, executors will provide a programmatic means of specifying where execution is allowed to occur during parallel algorithm execution and will open the door for user-defined execution policies in addition to the Parallelism TS's closed set of standard policies. If the first version of the Parallelism TS is standardized in 2017, such additional features for parallelism will be well-positioned for 2020.
We propose to standardize the functionality of the Parallelism TS as specified. In summary:
exception_list as a new subclause to Clause 19.Add the following entry to Table 44:
| 20.15 | Execution policies | <execution_policy> |
Add a new subclause to Clause 20:
This subclause describes classes that are execution policy types. An object of an execution policy type indicates the kinds of parallelism allowed in the execution of an algorithm and expresses the consequent requirements on the element access functions.
std::vector<int> v = ...
// standard sequential sort
std::sort(v.begin(), v.end());
using namespace std::experimental::parallel;
// explicitly sequential sort
sort(seq, v.begin(), v.end());
// permitting parallel execution
sort(par, v.begin(), v.end());
// permitting vectorization as well
sort(par_vec, v.begin(), v.end());
// sort with dynamically-selected execution
size_t threshold = ...
execution_policy exec = seq;
if (v.size() > threshold)
{
exec = par;
}
sort(exec, v.begin(), v.end());<execution_policy> synopsisnamespace std {
// 20.15.3, Execution policy type trait
template<class T> struct is_execution_policy;
template<class T> constexpr bool is_execution_policy_v = is_execution_policy<T>::value;
// 20.15.4, Sequential execution policy
class sequential_execution_policy;
// 20.15.5, Parallel execution policy
class parallel_execution_policy;
// 20.15.6, Parallel+Vector execution policy
class parallel_vector_execution_policy;
// 20.15.7, Dynamic execution policy
class execution_policy;
}template<class T> struct is_execution_policy { see below };
is_execution_policy
can be used to detect parallel execution policies for the purpose of
excluding function signatures from otherwise ambiguous overload
resolution participation.
is_execution_policy<T> shall be a UnaryTypeTrait with a BaseCharacteristic of true_type if T is the type of a standard or implementation-defined execution policy, otherwise false_type.
The behavior of a program that adds specializations for is_execution_policy is undefined.
class sequential_execution_policy{ unspecified };
The class sequential_execution_policy
is an execution policy type used as a unique type to disambiguate
parallel algorithm overloading and require that a parallel algorithm's
execution may not be parallelized.
class parallel_execution_policy{ unspecified };
The class parallel_execution_policy
is an execution policy type used as a unique type to disambiguate
parallel algorithm overloading and indicate that a parallel algorithm's
execution may be parallelized.
class parallel_vector_execution_policy{ unspecified };
The class class parallel_vector_execution_policy
is an execution policy type used as a unique type to disambiguate
parallel algorithm overloading and indicate that a parallel algorithm's
execution may be vectorized and parallelized.
class execution_policy
{
public:
// 20.15.7.1, execution_policy construct/assign
template<class T> execution_policy(const T& exec);
template<class T> execution_policy& operator=(const T& exec);
// 20.15.7.2, execution_policy object access
const type_info& type() const noexcept;
template<class T> T* get() noexcept;
template<class T> const T* get() const noexcept;
};
The class execution_policy is a container for execution policy objects.
execution_policy allows dynamic control over standard algorithm execution.
std::vector<float> sort_me = ...
using namespace std::experimental::parallel;
execution_policy exec = seq;
if(sort_me.size() > threshold)
{
exec = std::par;
}
std::sort(exec, std::begin(sort_me), std::end(sort_me));
— end example ]
Objects of type execution_policy shall be constructible and assignable from objects of
type T for which is_execution_policy<T>::value is true.
execution_policy construct/assigntemplate<class T> execution_policy(const T& exec); execution_policy object with a copy of exec's state.is_execution_policy<T>::value is true.
template<class T> execution_policy& operator=(const T& exec); exec's state to *this.*this.
execution_policy object accessconst type_info& type() const noexcept; typeid(T), such that T is the type of the execution policy object contained by *this.template<class T> T* get() noexcept; template<class T> const T* get() const noexcept; target_type() == typeid(T), a pointer to the stored execution policy object; otherwise a null pointer.is_execution_policy<T>::value is true.
constexpr sequential_execution_policy seq{};
constexpr parallel_execution_policy par{};
constexpr parallel_vector_execution_policy par_vec{};
The header <experimental/execution_policy> declares a global object associated with each type of execution policy defined by this Technical Specification.
exception_listAdd the following entry to Table 41:
| 19.6 | Exception list | <exception_list> |
Add a new subclause to Clause 19:
exception_listnamespace std {
namespace experimental {
namespace parallel {
inline namespace v1 {
class exception_list : public exception
{
public:
typedef unspecified iterator;
size_t size() const noexcept;
iterator begin() const noexcept;
iterator end() const noexcept;
const char* what() const noexcept override;
};
}
}
}
}
The class exception_list owns a sequence of exception_ptr objects. The parallel
algorithms may use the exception_list to communicate uncaught exceptions encountered during parallel execution to the
caller of the algorithm.
The type exception_list::iterator shall fulfill the requirements of
ForwardIterator.
size_t size() const noexcept; exception_ptr objects contained within the exception_list.
iterator begin() const noexcept; exception_ptr object contained within the exception_list.
iterator end() const noexcept; const char* what() const noexcept override; Add the following subclause to Clause 25:
A parallel algorithm is a function template described by this Technical Specification declared in namespace std::experimental::parallel::v1 with a formal template parameter named ExecutionPolicy.
Parallel algorithms access objects indirectly accessible via their arguments by invoking the following functions:
sort function may invoke the following element access functions:
RandomAccessIterator.
swap function on the elements of the sequence (as per 25.4.1.1 [sort]/2).
Compare function object.
Function objects passed into parallel algorithms as objects of type BinaryPredicate,
Compare, and BinaryOperation shall not directly or indirectly modify
objects via their arguments.
Parallel algorithms have template parameters named ExecutionPolicy which describe
the manner in which the execution of these algorithms may be parallelized and the manner in
which they apply the element access functions.
The invocations of element access functions in parallel algorithms invoked with an execution
policy object of type sequential_execution_policy execute in sequential order in
the calling thread.
The invocations of element access functions in parallel algorithms invoked with an execution
policy object of type parallel_execution_policy are permitted to execute in an
unordered fashion in either the invoking thread or in a thread implicitly created by the library
to support parallel algorithm execution. Any such invocations executing in the same thread are
indeterminately sequenced with respect to each other.
using namespace std::experimental::parallel;
int a[] = {0,1};
std::vector<int> v;
for_each(par, std::begin(a), std::end(a), [&](int i) {
v.push_back(i*2+1);
});
The program above has a data race because of the unsynchronized access to the container
v.
— end example ]
using namespace std::experimental::parallel;
std::atomic<int> x = 0;
int a[] = {1,2};
for_each(par, std::begin(a), std::end(a), [&](int n) {
x.fetch_add(1, std::memory_order_relaxed);
// spin wait for another iteration to change the value of x
while (x.load(std::memory_order_relaxed) == 1) { }
});
The above example depends on the order of execution of the iterations, and is therefore
undefined (may deadlock).
— end example ]
using namespace std::experimental::parallel;
int x=0;
std::mutex m;
int a[] = {1,2};
for_each(par, std::begin(a), std::end(a), [&](int) {
m.lock();
++x;
m.unlock();
});
The above example synchronizes access to object x ensuring that it is
incremented correctly.
— end example ]
The invocations of element access functions in parallel algorithms invoked with an execution
policy of type parallel_vector_execution_policy
are permitted to execute in an unordered fashion in unspecified threads, and unsequenced
with respect to one another within each thread.
parallel_vector_execution_policy allows the execution of element access functions to be
interleaved on a single thread, synchronization, including the use of mutexes, risks deadlock. Thus the
synchronization with parallel_vector_execution_policy is restricted as follows:
A standard library function is vectorization-unsafe if it is specified to synchronize with
another function invocation, or another function invocation is specified to synchronize with it, and if
it is not a memory allocation or deallocation function. Vectorization-unsafe standard library functions
may not be invoked by user code called from parallel_vector_execution_policy algorithms.
using namespace std::experimental::parallel;
int x=0;
std::mutex m;
int a[] = {1,2};
for_each(par_vec, std::begin(a), std::end(a), [&](int) {
m.lock();
++x;
m.unlock();
});
The above program is invalid because the applications of the function object are not
guaranteed to run on different threads.
— end example ]
m.lock on the same thread, which may deadlock.
— end note ]
parallel_execution_policy or the
parallel_vector_execution_policy invocation allow the implementation to fall back to
sequential execution if the system cannot parallelize an algorithm invocation due to lack of
resources.
— end note ]
Algorithms invoked with an execution policy object of type execution_policy
execute internally as if invoked with the contained execution policy object.
The semantics of parallel algorithms invoked with an execution policy object of implementation-defined type are implementation-defined.
During the execution of a standard parallel algorithm,
if temporary memory resources are required and none are available,
the algorithm throws a std::bad_alloc exception.
During the execution of a standard parallel algorithm, if the invocation of an element access function exits via an uncaught exception, the behavior of the program is determined by the type of execution policy used to invoke the algorithm:
class parallel_vector_execution_policy,
std::terminate shall be called.
sequential_execution_policy or
parallel_execution_policy, the execution of the algorithm exits via an exception. The
exception shall be an exception_list containing all uncaught exceptions thrown during
the invocations of element access functions, or optionally the uncaught exception if there was only one.
for_each is executed sequentially, if an invocation of the user-provided
function object throws an exception, for_each can exit via the uncaught exception, or throw an
exception_list containing the original exception.
— end note ]
std::bad_alloc, all exceptions thrown during the execution of the algorithm are communicated to the
caller. It is unspecified whether an algorithm implementation will "forge ahead" after encountering and capturing
a user exception.
— end note ]
std::bad_alloc exception
even if one or more user-provided
function objects have exited via an exception. For
example, this can happen when an algorithm fails to allocate memory
while
creating or adding elements to the exception_list object.
— end note ]
ExecutionPolicy algorithm overloads
The Parallel Algorithms Library provides overloads for each of the algorithms named in
Table 1, corresponding to the algorithms with the same name in the C++ Standard Algorithms Library.
For each algorithm in ExecutionPolicy, which shall be the first template parameter.
In addition, each such overload shall have the new function parameter as the
first function parameter of type ExecutionPolicy&&.
Unless otherwise specified, the semantics of ExecutionPolicy algorithm overloads
are identical to their overloads without.
Parallel algorithms shall not participate in overload resolution unless
is_execution_policy<decay_t<ExecutionPolicy>>::value is true.
adjacent_difference |
adjacent_find |
all_of |
any_of |
copy |
copy_if |
copy_n |
count |
count_if |
equal |
exclusive_scan |
fill |
fill_n |
find |
find_end |
find_first_of |
find_if |
find_if_not |
for_each |
for_each_n |
generate |
generate_n |
includes |
inclusive_scan |
inner_product |
inplace_merge |
is_heap |
is_heap_until |
is_partitioned |
is_sorted |
is_sorted_until |
lexicographical_compare |
max_element |
merge |
min_element |
minmax_element |
mismatch |
move |
none_of |
nth_element |
partial_sort |
partial_sort_copy |
partition |
partition_copy |
reduce |
remove |
remove_copy |
remove_copy_if |
remove_if |
replace |
replace_copy |
replace_copy_if |
replace_if |
reverse |
reverse_copy |
rotate |
rotate_copy |
search |
search_n |
set_difference |
set_intersection |
set_symmetric_difference |
set_union |
sort |
stable_partition |
stable_sort |
swap_ranges |
transform |
transform_exclusive_scan |
transform_inclusive_scan |
transform_reduce |
uninitialized_copy |
uninitialized_copy_n |
uninitialized_fill |
uninitialized_fill_n |
unique |
unique_copy |
for_each with ExecutionPolicy, sequential for_each_n, and for_each_n with ExecutionPolicy to subclause 4.3.2:
template<class ExecutionPolicy,
class InputIterator, class Function>
void for_each(ExecutionPolicy&& exec,
InputIterator first, InputIterator last,
Function f); f to the result of dereferencing every iterator in the range [first,last).
first satisfies the requirements of a mutable iterator, f may
apply nonconstant functions through the dereferenced iterator.
— end note ]
f exactly last - first times.
f returns a result, the result is ignored.
for_each does not return a copy of
its Function parameter, since parallelization may not permit efficient state
accumulation.
for_each requires
Function to meet the requirements of CopyConstructible.
template<class InputIterator, class Size, class Function>
InputIterator for_each_n(InputIterator first, Size n,
Function f); Function shall meet the requirements of MoveConstructible
Function need not meet the requirements of CopyConstructible.
— end note ]
f to the result of dereferencing every iterator in the range
[first,first + n), starting from first and proceeding to first + n - 1.
first satisfies the requirements of a mutable iterator,
f may apply nonconstant functions through the dereferenced iterator.
— end note ]
first + n for non-negative values of n and first for negative values.
f returns a result, the result is ignored.
template<class ExecutionPolicy,
class InputIterator, class Size, class Function>
InputIterator for_each_n(ExecutionPolicy && exec,
InputIterator first, Size n,
Function f); f to the result of dereferencing every iterator in the range
[first,first + n), starting from first and proceeding to first + n - 1.
first satisfies the requirements of a mutable iterator,
f may apply nonconstant functions through the dereferenced iterator.
— end note ]
first + n for non-negative values of n and first for negative values.
f returns a result, the result is ignored.
for_each_n requires
Function to meet the requirements of CopyConstructible.
Insert the following entry to Table 113:
| 26.2 | Definitions |
Insert the following subclause to Clause 26:
Define GENERALIZED_SUM(op, a1, ..., aN) as follows:
a1 when N is 1op(GENERALIZED_SUM(op, b1, ..., bK), GENERALIZED_SUM(op, bM, ..., bN)) where
b1, ..., bN may be any permutation of a1, ..., aN and1 < K+1 = M ≤ N.
Define GENERALIZED_NONCOMMUTATIVE_SUM(op, a1, ..., aN) as follows:
a1 when N is 1op(GENERALIZED_NONCOMMUTATIVE_SUM(op, a1, ..., aK), GENERALIZED_NONCOMMUTATIVE_SUM(op, aM, ..., aN) where 1 < K+1 = M ≤ N.
reduce, exclusive_scan, inclusive_scan, transform_reduce, transform_exclusive_scan, and transform_inclusive_scan to Clause 26.7.
The innermost section number is given as NaN to allow editorial discretion.
template<class InputIterator>
typename iterator_traits<InputIterator>::value_type
reduce(InputIterator first, InputIterator last); reduce(first, last, typename iterator_traits<InputIterator>::value_type{}).
template<class InputIterator, class T>
T reduce(InputIterator first, InputIterator last, T init); reduce(first, last, init, plus<>()).
template<class InputIterator, class T, class BinaryOperation>
T reduce(InputIterator first, InputIterator last, T init,
BinaryOperation binary_op); GENERALIZED_SUM(binary_op, init, *first, ..., *(first + (last - first) - 1)).
binary_op shall not invalidate iterators or subranges, nor modify elements in the
range [first,last).
last - first) applications of binary_op.
reduce and accumulate is that the behavior
of reduce may be non-deterministic for non-associative or non-commutative binary_op.
template<class InputIterator, class OutputIterator, class T>
OutputIterator exclusive_scan(InputIterator first, InputIterator last,
OutputIterator result,
T init); exclusive_scan(first, last, result, init, plus<>()).
template<class InputIterator, class OutputIterator, class T, class BinaryOperation>
OutputIterator exclusive_scan(InputIterator first, InputIterator last,
OutputIterator result,
T init, BinaryOperation binary_op); i in [result,result + (last - first)) the
value of GENERALIZED_NONCOMMUTATIVE_SUM(binary_op, init, *first, ..., *(first + (i - result) - 1)).
result.
binary_op shall not invalidate iterators or subranges, nor modify elements in the
ranges [first,last) or [result,result + (last - first)).
last - first) applications of binary_op.
exclusive_scan and inclusive_scan is that
exclusive_scan excludes the ith input element from the ith
sum. If binary_op is not mathematically associative, the behavior of
exclusive_scan may be non-deterministic.
template<class InputIterator, class OutputIterator>
OutputIterator inclusive_scan(InputIterator first, InputIterator last,
OutputIterator result); inclusive_scan(first, last, result, plus<>()).
template<class InputIterator, class OutputIterator, class BinaryOperation>
OutputIterator inclusive_scan(InputIterator first, InputIterator last,
OutputIterator result,
BinaryOperation binary_op); template<class InputIterator, class OutputIterator, class BinaryOperation>
OutputIterator inclusive_scan(InputIterator first, InputIterator last,
OutputIterator result,
BinaryOperation binary_op, T init); i in [result,result + (last - first)) the value of
GENERALIZED_NONCOMMUTATIVE_SUM(binary_op, *first, ..., *(first + (i - result))) or
GENERALIZED_NONCOMMUTATIVE_SUM(binary_op, init, *first, ..., *(first + (i - result)))
if init is provided.
result.
binary_op shall not invalidate iterators or subranges, nor modify elements in the
ranges [first,last) or [result,result + (last - first)).
last - first) applications of binary_op.
exclusive_scan and inclusive_scan is that
inclusive_scan includes the ith input element in the ith sum.
If binary_op is not mathematically associative, the behavior of
inclusive_scan may be non-deterministic.
template<class InputIterator, class UnaryFunction, class T, class BinaryOperation>
T transform_reduce(InputIterator first, InputIterator last,
UnaryOperation unary_op, T init, BinaryOperation binary_op); GENERALIZED_SUM(binary_op, init, unary_op(*first), ..., unary_op(*(first + (last - first) -1))).
unary_op nor binary_op shall invalidate subranges, or modify elements in the range [first,last)
last - first) applications each of unary_op and binary_op.
transform_reduce does not apply unary_op to init.
template<class InputIterator, class OutputIterator,
class UnaryOperation,
class T, class BinaryOperation>
OutputIterator transform_exclusive_scan(InputIterator first, InputIterator last,
OutputIterator result,
UnaryOperation unary_op,
T init, BinaryOperation binary_op); i in [result,result + (last - first)) the value of
GENERALIZED_NONCOMMUTATIVE_SUM(binary_op, init, unary_op(*first), ..., unary_op(*(first + (i- result) - 1))).
result.
unary_op nor binary_op shall invalidate iterators or subranges, or modify elements in the
ranges [first,last) or [result,result + (last - first)).
last - first) applications each of unary_op and binary_op.
transform_exclusive_scan and transform_inclusive_scan is that transform_exclusive_scan
excludes the ith input element from the ith sum. If binary_op is not mathematically associative, the behavior of
transform_exclusive_scan may be non-deterministic. transform_exclusive_scan does not apply unary_op to init.
template<class InputIterator, class OutputIterator,
class UnaryOperation,
class BinaryOperation>
OutputIterator transform_inclusive_scan(InputIterator first, InputIterator last,
OutputIterator result,
UnaryOperation unary_op,
BinaryOperation binary_op); template<class InputIterator, class OutputIterator,
class UnaryOperation,
class BinaryOperation, class T>
OutputIterator transform_inclusive_scan(InputIterator first, InputIterator last,
OutputIterator result,
UnaryOperation unary_op,
BinaryOperation binary_op, T init); i in [result,result + (last - first)) the value of
GENERALIZED_NONCOMMUTATIVE_SUM(binary_op, unary_op(*first), ..., unary_op(*(first + (i -result)))) or
GENERALIZED_NONCOMMUTATIVE_SUM(binary_op, init, unary_op(*first), ..., unary_op(*(first + (i- result))))
if init is provided.
result.
unary_op nor binary_op shall invalidate iterators or subranges, or modify elements in the ranges [first,last)
or [result,result + (last - first)).
last - first) applications each of unary_op and binary_op.
transform_exclusive_scan and transform_inclusive_scan is that transform_inclusive_scan
includes the ith input element from the ith sum. If binary_op is not mathematically associative, the behavior of
transform_inclusive_scan may be non-deterministic. transform_inclusive_scan does not apply unary_op to init.