Strzelczyk, P; Podbielski, M; Marks, M; Niewiadomska-Szynkiewicz, E
The paper addresses issues associated with distributed computing systems and the application of mixed GPU&CPU technology to the password recovery and the data encryption and decryption algorithms. We describe our heterogenous cluster formed by two types of nodes: Intel processor with NVIDIA graphics processing unit and AMD processor with AMD graphics processing unit (formerly ATI) which is used to solve different cryptography and cryptanalysis problems. The main goal of our research was to assess how different hardware architectures, namely GPUs and CPUs coming from three different vendors: Intel, NVIDIA and AMD, which form our heterogenous cluster can be exploited in some cryptography and cryptanalysis tasks. Different hardware parallelization concepts (SIMD, MIMD, SIMT), different processor architectures (VLIW, superscalar) and different memory access patterns, and as a result also different code optimization techniques make some of the platforms more suitable in some applications while at the same time less suitable in others. The knowledge about how different platforms cope with specific tasks can be used to plan the work distribution among the cluster nodes and to optimize the task scheduling process. The numerical results for some of our parallel implementations of cryptography techniques are presented in the final part to illustrate the posed research.