API Layer

The API layer performs argument validation and error handling, split operations into simpler components, and finally dispatches the low-level Compute functions that perform the actual number crunching.

The API-level functions are called drivers for the low-level operations. In many cases the low-level interface provides small pieces that together make an API function. The rationale behind this is to separate the parts that are platform-dependent from the parts that are not. The things belonging to the API layer can be separated into three distinct categories:

  1. Argument validation and error handling.
  2. Identifying simple or trivial cases (argument reduction).
  3. Algorithmic parts of the operation, such as decomposition of morphological operations, block processing and alignment-dependent selection of low-level processing functions.

Everything that should be implemented differently on different platforms, i.e. SSE and non-SSE, belong to the low-level Compute domain. For more information about the RAPP API, see the RAPP User's Manual.

Next section: Compute Layer


Generated on 1 Jun 2016 for RAPP Compute by  doxygen 1.6.1