**1. Introduction**

There are various heterogeneous phenomena in reactor core, which are related to multi-physics and multi-scales modeling and simulation [1]. The core program or software is the important approximation for the reactor core; therefore, mathematics models, numerical algorithms, preprocessing, post-processing, and visualization are sometimes managed in the similar software structure [2]. Based on the existing conditions of calculation, high-performance-computing technology is an important mean to accomplish the simulation of core programs by the cluster computer. For instance, there are several supercomputers (cluster) in ANL and ORNL that provide high-capacity resources for reactor engineering calculation and nuclear-related simulation [3]. Once the numerical algorithm is fixed, the time-to-solution becomes the necessary consideration during the research, and the approach pattern can be summed up in three steps [4] in this field:


According to the modeling and simulation for reactor and the features of multiphysics and multi-scale models, the reactor calculation is classified into core calculation and out-core calculation [5]. The core calculation covers considerable subjects, for instance, reactor physics, thermal-hydraulics, dynamics, and fuel performance, for example. For the purpose of research and engineering, different discrete systems and solvers are implemented so as to reveal different phenomena in the core [6–8]. As an example, the DENOVO transport software of ORNL utilizes 160,000 processors to run the benchmark core example in Jaguar supercomputer; up to now the computer has been upgraded to TITAN [9].

On the one hand, the simulation could be accelerated through effective parallelization in different models [10]. On the other hand, the parallel computing makes researchers focus on more research fields except mathematics and physics [11], which constructs one view from nuclear data to top-level application. With the attempt of the M&S content, the core software could be ascribed as the expression and translation of the set of state variables and data structure [12]. Thus, the conjunction of the core problem and parallel computing is from the new requirements of application; furthermore it is driven by the underlying hardware innovation. The Berkeley Parallel Computing Laboratory reviews parallel computing and points out that the parallel elements of different algorithms are the methodology of application driven and reuse. Since various features exist in different layers of software, the parallel algorithm should be designed specifically so as to adapt the parallel hardware [13]. The chapter describes and explains the software structure, underlying nuclear data, and parallel algorithm from a systematic view in the domain of reactor core calculation. Section 2 abstracts the software structure for core calculation. Then, Section 3 explains the basic elements of nuclear data and its association according to the software structure. After that, Section 4 lists the concept and steps of parallel algorithm analysis for core problem in practice. Sample programs are analyzed and discussed concisely in Section 4, and conclusions are given in Section 5.

## **2. Software structure**

As the complex system, the reactor core could be understood by various parts, such as concept models, numerical methods, model uncertainty, and model reuse. Algorithms and software are the center of digital simulation so that the calculation rule is regulated by the program technologies. Sometimes the fixed software structure could be abstracted, and common algorithms could be extracted to be close to the trend of reactor core simulation, which forms one united manner (also known as framework method [14]).

According to the description [12], one general software structure could be divided into seven layers, which covers all aspects from the application to the computer operating system by **Table 1**.

Then, the numerical calculation system is abstracted into multiple layers, which are implemented through controls, modularity, and dynamic interaction. So participants from different professional backgrounds pay more attention to their own business [15]. The core calculation learns from the above contents, and the software structure could be simplified in **Table 2**.

It is effective to develop numerical software based on reusable components or framework, so the discrete model could be modeled through fixed process. On the one hand, the usage leads to reuse. On the other hand, new programming pattern could be merged. For instance, with the help of the underlying framework, MC neutron transport combines the discrete model and algorithm analysis process and


#### **Table 1.**

*General software structure.*


### **Table 2.**

*Software structure for core calculation.*

then uses numerical components and structure to accomplish the transport task easily in the reactor core [16]. Since this integrated software structure method restricts the programming manner of the core calculation and provides reuse mechanism, then in practical, the prototype software NAC4R is designed and constructed that contains basic data operation, basic algebra operation and linear system solver, and so on. As an example of the above data, NAC4R can be used for the parallel analysis process of sample programs that explains the effectiveness and efficiency of the corresponding software structure.
