### Overview Changes in OpenCL 1L VN Square Grid CAAPI implementation version 110 **(2015 April)** ###

 1. Major features:

  - Added the possibility to use native math function in kernels

  - Alarms functions isActivate and areAllActivate have changed name
    to isActivated and areAllActivated

  - Added caActivateAlarmMask function in caapi

  - Increase the number of arguments to 17 for CA function

 2. Minor items:

  - This version is derived from the internal revision 971

  - Updated the code to compile using clang 3.5 and gcc 4.8

  - Added the possibility to dump Table object

  - Changed the way OpenCL events are managed to reduce memory leaks.

  - It is possible to choose between OpenCL constant or global memory
    for the Table object

  - Improved device fission between OpenCL 1.1 and 1.2 versions.
  
 3. Known missing features & problems:

  - 

### Overview Changes in OpenCL 1L VN Square Grid CAAPI implementation Version 100 **(2014 February)** ###

 1. Major features:

  - OpenCL implementation of the CAAPI that uses a square grid with
    one single level of von Neumann neighbourhood that can run on
    multi-core CPU or GPU.

  - Implement the version 100 of the CAAPI

  - It is possible to choose which OpenCL platform and devices to run
    the execution by using the implementation option config-file. This
    file and other options can be used to set up various other OpenCL
    parameters.

  - There are two example files (config_INTEL_CPU.csv and
    config_NVIDIA_GPU.csv) that are used to run an execution into an
    INTEL CPU and an NVIDIA GPU.

  - The implementation contains two extra executable (convertCATOHPP
    and clinfo) which respectively convert a CA function into a header
    with OpenCL code and return various information about the OpenCL
    platform and devices available.

 2. Minor items:

  - This version is derived from the internal revision 801

 3. Known missing features & problems:

  - This implementation has only been tested on NVIDA GPUs and INTEL
    CPUs. No test was performed on AMD CPU and GPUs. 

  - It need the FULL_PROFILE implementation of OpenCL. It has not been
    tested on an embadded profile.
