The atomic functions are shown below. click an item in the table below for details about that function.
| atomic_add(3clc) | atomic_sub(3clc) | atomic_xchg(3clc) | atomic_inc(3clc) |
|---|---|---|---|
| atomic_dec(3clc) | atomic_cmpxchg(3clc) | atomic_min(3clc) | atomic_max(3clc) |
| atomic_and(3clc) | atomic_or(3clc) | atomic_xor(3clc) |
The cl_khr_int64_base_atomics(3clc) extension enables additional atomic functions using types long and unsigned long. An application that wants to use this extension will need to include the #pragma OPENCL EXTENSION cl_khr_int64_base_atomics(3clc) : enable in the OpenCL program source.
| atom_add(3clc) | atom_sub(3clc) | atom_xchg(3clc) | atom_inc(3clc) |
|---|---|---|---|
| atom_dec(3clc) | atom_cmpxchg(3clc) |
The cl_khr_int64_extended_atomics(3clc) extension enables additional atomic functions using types long and unsigned long. An application that wants to use this extension will need to include the #pragma OPENCL EXTENSION cl_khr_int64_extended_atomics(3clc) : enable in the OpenCL program source.
| atom_min(3clc) | atom_max(3clc) | atom_xor(3clc) |
|---|---|---|
| atom_and(3clc) | atom_or(3clc) |
These functions provide atomic operations on 32-bit signed, unsigned integers and single precision floating-point to locations in global(3clc) or local(3clc) memory. Only the atomic_xchg operation is supported for single precision floating-point data type.
The atomic built-in functions that use the atom_ prefix and are described in the \m[blue]OpenCL Extension Specification\m[]\s-2\u[1]\d\s+2 and are enabled by: cl_khr_global_int32_base_atomics, cl_khr_global_int32_extended_atomics, cl_khr_local_int32_base_atomics, and cl_khr_local_int32_extended_atomics in sections 9.5 and 9.6 of the OpenCL 1.0 specification are also supported. .
The 64-bit transactions are atomic for the device executing these atomic functions. There is no guarantee of atomicity if the atomic operations to the same memory location are being performed by kernels executing on multiple devices.
\m[blue]OpenCL Specification\m[]\s-2\u[2]\d\s+2
The Khronos Group
Copyright © 2007-2011 The Khronos Group Inc.
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and/or associated documentation files (the "Materials"), to deal in the Materials without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Materials, and to permit persons to whom the Materials are furnished to do so, subject to the condition that this copyright notice and permission notice shall be included in all copies or substantial portions of the Materials.
OpenCL Extension Specification
page 10, section 9.3 - 64-bit Atomics
OpenCL Specification
page 279, section 6.12.11 - Atomic Functions for 32-bit integers