The atomic functions are shown below. Click an item in the table below for details about that function.
atomic_add | atomic_sub | atomic_xchg | atomic_inc |
atomic_dec | atomic_cmpxchg | atomic_min | atomic_max |
atomic_and | atomic_or | atomic_xor |
The cl_khr_int64_base_atomics
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 :
enable
in the OpenCL program source.
The cl_khr_int64_extended_atomics
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
: enable
in the OpenCL program source.
These functions provide atomic operations on 32-bit signed, unsigned
integers and single precision floating-point to locations in __global or __local 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 OpenCL Extension Specification 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.