Provided by: opencl-1.2-man-doc_1.0~svn33624-1_all bug


       atom_sub - 64-bit atom_sub functions.

       long atom_sub(volatile global(3clc) long *p, long val);

       long atom_sub(volatile global(3clc) long *p, long val);

       ulong atom_sub(volatile local(3clc) ulong *p, ulong val);

       ulong atom_sub(volatile local(3clc) ulong *p, ulong val);


       Read the 64-bit value (referred to as old) stored at location pointed by p. Compute (old -
       val) and store result at location pointed by p. The function returns old.

       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.

       This function is performed in one atomic transaction. These 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.

       Atomic operations on 64-bit integers and 32-bit integers (and float) are also atomic with
       regard to each other.


       OpenCL Specification[1]


       atomicFunctions(3clc), atomic_sub(3clc), cl_khr_int64_base_atomics(3clc)


       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.


        1. OpenCL Specification
           page 279, section 6.12.11 - Atomic Functions for 32-bit integers