Every feature in OpenRDM is designed to provide the user with a convenient way to reach its goal. Not surprisingly, there is a cost to be paid for such convenience. As such, before using OpenRDM, we invite the user to spend some time to take a glance at some important points and technical details about the OpenRDM library to become familiar with pros and cons of each existing feature.

OpenMP

The OpenMP application programming interface (API) has armed us with an efficient tool to keep the sequential and parallelized versions of the code within the same source code file. Such unification strategy presents a choice to the user between the parallel or sequential implementations of the code. Due to the presence of many nested loops of various depths, all efforts have been directed towards minimizing the number of OpenMP pragma directives and enlarging the parallel regions while minimizing the number of implied and explicit barriers as well as the usage of single and critical constructions. This conservative approach will not only reduces the overhead pertinent to the OpenMP constructs, it bestows readability and performance upon our implementation. Nevertheless, many important factors are required to be considered before making any judgements about the performance of the code. First, apart from the author of the code, compiler also plays a crucial role to the optimization of the code. Different compilers with various levels of optimizations create an extra hidden layer behind the complexity of the sequential or parallel code.