CCCL (CUDA C++ Core Libraries) provides: - CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk) - Thrust: high-level parallel algorithms (transform_reduce, sort, scan) - libcudacxx: CUDA C++ standard library (atomics, barriers, memory) - cudax: experimental features (memory resources, allocators) - Tuning policies: per-SM hardware-specific algorithm parameters Competition optimization vectors mapped to CCCL: - Output TPS (83% weight): warp_reduce, block_reduce, device_topk - Input TPS (14% weight): device_scan, block_load, prefetch - Cache TPS (3% weight): prefix caching strategy patterns - Memory (0.9 util): pooled/cached/buddy allocators Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only) License: Apache-2.0
226 lines
7.7 KiB
C++
226 lines
7.7 KiB
C++
// SPDX-FileCopyrightText: Copyright (c) 2008-2013, NVIDIA Corporation. All rights reserved.
|
|
// SPDX-License-Identifier: Apache-2.0
|
|
|
|
/*! \file discard_block_engine.h
|
|
* \brief A random number engine which adapts a base engine and produces
|
|
* numbers by discarding all but a contiguous blocks of its values.
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include <thrust/detail/config.h>
|
|
|
|
#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
|
|
# pragma GCC system_header
|
|
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
|
|
# pragma clang system_header
|
|
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
|
|
# pragma system_header
|
|
#endif // no system header
|
|
|
|
#include <thrust/random/detail/random_core_access.h>
|
|
|
|
#include <cuda/std/__host_stdlib/istream>
|
|
#include <cuda/std/__host_stdlib/ostream>
|
|
#include <cuda/std/cstdint>
|
|
|
|
THRUST_NAMESPACE_BEGIN
|
|
|
|
namespace random
|
|
{
|
|
/*! \addtogroup random_number_engine_adaptors Random Number Engine Adaptor Class Templates
|
|
* \ingroup random
|
|
* \{
|
|
*/
|
|
|
|
/*! \class discard_block_engine
|
|
* \brief A \p discard_block_engine adapts an existing base random number engine and produces
|
|
* random values by discarding some of the values returned by its base engine.
|
|
* Each cycle of the compound engine begins by returning \c r values successively produced
|
|
* by the base engine and ends by discarding <tt>p-r</tt> such values. The engine's state
|
|
* is the state of its base engine followed by the number of calls to <tt>operator()</tt>
|
|
* that have occurred since the beginning of the current cycle.
|
|
*
|
|
* \tparam Engine The type of the base random number engine to adapt.
|
|
* \tparam p The discard cycle length.
|
|
* \tparam r The number of values to return of the base engine. Because <tt>p-r</tt> will be
|
|
* discarded, <tt>r <= p</tt>.
|
|
*
|
|
* The following code snippet shows an example of using a \p discard_block_engine instance:
|
|
*
|
|
* \code
|
|
* #include <thrust/random/linear_congruential_engine.h>
|
|
* #include <thrust/random/discard_block_engine.h>
|
|
* #include <iostream>
|
|
*
|
|
* int main()
|
|
* {
|
|
* // create a discard_block_engine from minstd_rand, with a cycle length of 13
|
|
* // keep every first 10 values, and discard the next 3
|
|
* thrust::discard_block_engine<thrust::minstd_rand, 13, 10> rng;
|
|
*
|
|
* // print a random number to standard output
|
|
* std::cout << rng() << '\n';
|
|
*
|
|
* return 0;
|
|
* }
|
|
* \endcode
|
|
*/
|
|
template <typename Engine, size_t p, size_t r>
|
|
class discard_block_engine
|
|
{
|
|
public:
|
|
// types
|
|
|
|
/*! \typedef base_type
|
|
* \brief The type of the adapted base random number engine.
|
|
*/
|
|
using base_type = Engine;
|
|
|
|
/*! \typedef result_type
|
|
* \brief The type of the unsigned integer produced by this \p linear_congruential_engine.
|
|
*/
|
|
using result_type = typename base_type::result_type;
|
|
|
|
// engine characteristics
|
|
|
|
/*! The length of the production cycle.
|
|
*/
|
|
static const size_t block_size = p;
|
|
|
|
/*! The number of used numbers per production cycle.
|
|
*/
|
|
static const size_t used_block = r;
|
|
|
|
/*! The smallest value this \p discard_block_engine may potentially produce.
|
|
*/
|
|
static const result_type min = base_type::min;
|
|
|
|
/*! The largest value this \p discard_block_engine may potentially produce.
|
|
*/
|
|
static const result_type max = base_type::max;
|
|
|
|
// constructors and seeding functions
|
|
|
|
/*! This constructor constructs a new \p discard_block_engine and constructs
|
|
* its \p base_type engine using its null constructor.
|
|
*/
|
|
_CCCL_HOST_DEVICE discard_block_engine();
|
|
|
|
/*! This constructor constructs a new \p discard_block_engine using
|
|
* a given \p base_type engine to initialize its adapted base engine.
|
|
*
|
|
* \param urng A \p base_type to use to initialize this \p discard_block_engine's
|
|
* adapted base engine.
|
|
*/
|
|
_CCCL_HOST_DEVICE explicit discard_block_engine(const base_type& urng);
|
|
|
|
/*! This constructor initializes a new \p discard_block_engine with a given seed.
|
|
*
|
|
* \param s The seed used to initialize this \p discard_block_engine's adapted base engine.
|
|
*/
|
|
_CCCL_HOST_DEVICE explicit discard_block_engine(result_type s);
|
|
|
|
/*! This method initializes the state of this \p discard_block_engine's adapted base engine
|
|
* by using its \p default_seed value.
|
|
*/
|
|
_CCCL_HOST_DEVICE void seed();
|
|
|
|
/*! This method initializes the state of this \p discard_block_engine's adapted base engine
|
|
* by using the given seed.
|
|
*
|
|
* \param s The seed with which to initialize this \p discard_block_engine's adapted base engine.
|
|
*/
|
|
_CCCL_HOST_DEVICE void seed(result_type s);
|
|
|
|
// generating functions
|
|
|
|
/*! This member function produces a new random value and updates this \p discard_block_engine's state.
|
|
* \return A new random number.
|
|
*/
|
|
_CCCL_HOST_DEVICE result_type operator()();
|
|
|
|
/*! This member function advances this \p discard_block_engine's state a given number of times
|
|
* and discards the results.
|
|
*
|
|
* \param z The number of random values to discard.
|
|
* \note This function is provided because an implementation may be able to accelerate it.
|
|
*/
|
|
_CCCL_HOST_DEVICE void discard(unsigned long long z);
|
|
|
|
// property functions
|
|
|
|
/*! This member function returns a const reference to this \p discard_block_engine's
|
|
* adapted base engine.
|
|
*
|
|
* \return A const reference to the base engine this \p discard_block_engine adapts.
|
|
*/
|
|
_CCCL_HOST_DEVICE const base_type& base() const;
|
|
|
|
/*! \cond
|
|
*/
|
|
|
|
private:
|
|
base_type m_e;
|
|
unsigned int m_n;
|
|
|
|
friend struct thrust::random::detail::random_core_access;
|
|
|
|
_CCCL_HOST_DEVICE bool equal(const discard_block_engine& rhs) const;
|
|
|
|
template <typename CharT, typename Traits>
|
|
std::basic_ostream<CharT, Traits>& stream_out(std::basic_ostream<CharT, Traits>& os) const;
|
|
|
|
template <typename CharT, typename Traits>
|
|
std::basic_istream<CharT, Traits>& stream_in(std::basic_istream<CharT, Traits>& is);
|
|
/*! \endcond
|
|
*/
|
|
}; // end discard_block_engine
|
|
|
|
/*! This function checks two \p discard_block_engines for equality.
|
|
* \param lhs The first \p discard_block_engine to test.
|
|
* \param rhs The second \p discard_block_engine to test.
|
|
* \return \c true if \p lhs is equal to \p rhs; \c false, otherwise.
|
|
*/
|
|
template <typename Engine, size_t p, size_t r>
|
|
_CCCL_HOST_DEVICE bool
|
|
operator==(const discard_block_engine<Engine, p, r>& lhs, const discard_block_engine<Engine, p, r>& rhs);
|
|
|
|
/*! This function checks two \p discard_block_engines for inequality.
|
|
* \param lhs The first \p discard_block_engine to test.
|
|
* \param rhs The second \p discard_block_engine to test.
|
|
* \return \c true if \p lhs is not equal to \p rhs; \c false, otherwise.
|
|
*/
|
|
template <typename Engine, size_t p, size_t r>
|
|
_CCCL_HOST_DEVICE bool
|
|
operator!=(const discard_block_engine<Engine, p, r>& lhs, const discard_block_engine<Engine, p, r>& rhs);
|
|
|
|
/*! This function streams a discard_block_engine to a \p std::basic_ostream.
|
|
* \param os The \p basic_ostream to stream out to.
|
|
* \param e The \p discard_block_engine to stream out.
|
|
* \return \p os
|
|
*/
|
|
template <typename Engine, size_t p, size_t r, typename CharT, typename Traits>
|
|
std::basic_ostream<CharT, Traits>&
|
|
operator<<(std::basic_ostream<CharT, Traits>& os, const discard_block_engine<Engine, p, r>& e);
|
|
|
|
/*! This function streams a discard_block_engine in from a std::basic_istream.
|
|
* \param is The \p basic_istream to stream from.
|
|
* \param e The \p discard_block_engine to stream in.
|
|
* \return \p is
|
|
*/
|
|
template <typename Engine, size_t p, size_t r, typename CharT, typename Traits>
|
|
std::basic_istream<CharT, Traits>&
|
|
operator>>(std::basic_istream<CharT, Traits>& is, discard_block_engine<Engine, p, r>& e);
|
|
|
|
/*! \} // end random_number_engine_adaptors
|
|
*/
|
|
} // namespace random
|
|
|
|
// import names into thrust::
|
|
using random::discard_block_engine;
|
|
|
|
THRUST_NAMESPACE_END
|
|
|
|
#include <thrust/random/detail/discard_block_engine.inl>
|