| Project: | ISO JTC1/SC22/WG21: Programming Language C++ |
|---|---|
| Number: | D0020r3 |
| Date: | 2016-06-21 |
| Reply-to: | hcedwar@sandia.gov |
| Author: | H. Carter Edwards |
| Contact: | hcedwar@sandia.gov |
| Author: | Hans Boehm |
| Contact: | hboehm@google.com |
| Author: | Olivier Giroux |
| Contact: | ogiroux@nvidia.com |
| Author: | JF Bastien |
| Contact: | jfb@google.com |
| Author: | James Reus |
| Contact: | reus1@llnl.gov |
| Audience: | SG1 Concurrency, Library Evolution |
| URL: | https://github.com/kokkos/ISO-CPP-Papers/blob/master/P0020.rst |
| Revision History | |
| D0020r3 | Update in conformance with N4594. Remove reference to atomic-view paper. |
This paper proposes an extension to the atomic operations library [atomics] for atomic addition on an object conforming to the atomic<T> where T is a floating type (N4594 3.9.1p8). This capability is critical for high performance computing (HPC) applications.
Insert into N4594 29.2 Header <atomic> synopsis
// 29.5, generic typestemplate<> struct atomic< floating >;// 29.6.3, arithmetic operations on atomic types// In the following declarations, atomic-floating is atomic<T>.floating atomic_fetch_add(volatile atomic-floating * , floating) noexcept ;floating atomic_fetch_add( atomic-floating * , floating) noexcept ;floating atomic_fetch_add_explicit(volatile atomic-floating * , floating , memory_order ) noexcept ;floating atomic_fetch_add_explicit( atomic-floating * , floating , memory_order ) noexcept ;floating atomic_fetch_sub(volatile atomic-floating * , floating) noexcept ;floating atomic_fetch_sub( atomic-floating * , floating) noexcept ;floating atomic_fetch_sub_explicit(volatile atomic-floating * , floating , memory_order ) noexcept ;floating atomic_fetch_sub_explicit( atomic-floating * , floating , memory_order ) noexcept ;
Add to N4594 29.5, Atomic Types, before 29.5p1.
template<> struct atomic< floating > {bool is_lock_free() const volatile noexceptbool is_lock_free() const noexceptstatic constexpr bool is_always_lock_free = implementation-defined ;void store( floating , memory_order = memory_order_seq_cst ) volatile noexcept;void store( floating , memory_order = memory_order_seq_cst ) noexcept;operator floating () const volatile noexcept;operator floating () const noexcept;floating exchange( floating , memory_order = memory_order_seq_cst ) volatile noexcept;floating exchange( floating , memory_order = memory_order_seq_cst ) noexcept;bool compare_exchange_weak( floating & , floating , memory_order , memory_order ) volatile noexcept;bool compare_exchange_weak( floating & , floating , memory_order , memory_order ) noexcept;bool compare_exchange_strong( floating & , floating , memory_order , memory_order ) volatile noexcept;bool compare_exchange_strong( floating & , floating , memory_order , memory_order ) noexcept;bool compare_exchange_weak( floating & , floating , memory_order order = memory_order_seq_cst ) volatile noexcept;bool compare_exchange_weak( floating & , floating , memory_order order = memory_order_seq_cst ) noexcept;bool compare_exchange_strong( floating & , floating , memory_order order = memory_order_seq_cst ) volatile noexcept;bool compare_exchange_strong( floating & , floating , memory_order order = memory_order_seq_cst ) noexcept;floating fetch_add( floating , memory_order = memory_order_seq_cst ) volatile noexcept;floating fetch_add( floating , memory_order = memory_order_seq_cst ) noexcept;floating fetch_sub( floating , memory_order = memory_order_seq_cst ) volatile noexcept;floating fetch_sub( floating , memory_order = memory_order_seq_cst ) noexcept;atomic() noexcept = default;constexpr atomic( floating ) noexcept;atomic( const atomic & ) = delete ;atomic & operator=( const atomic & ) = delete ;atomic & operator=( const atomic & ) volatile = delete ;floating operator=( floating ) volatile noexcept;floating operator=( floating ) noexcept;floating operator+=( floating ) volatile noexcept;floating operator+=( floating ) noexcept;floating operator-=( floating operand ) volatile noexcept;floating operator-=( floating operand ) noexcept;};
Insert N4594 29.6.3p3:
In the declarations of these functions and function template specializations, the name floating refers to a floating type and the name atomic-floating refers to atomic< floating >.
Append to Remark paragraph 29.6.5p31.
For floating types atomic arithmatic operations should conform to std::numeric_limits< floating > (18.3.2) traits associated with the floating point type. The floating point environment (26.3) for floating operations may be different than the calling thread's floating point environment.