Porting and optimization of solidification application for CPU–MIC hybrid platforms

Abstract

Modern heterogeneous computing platforms have become powerful HPC solutions, which could be applied to a wide range of real-life applications. In particular, the hybrid platforms equipped with Intel Xeon Phi coprocessors offer the advantages of massively parallel computing, while supporting practically the same parallel programming model as conventional homogeneous solutions. However, there is still an open issue as to how scientific applications can efficiently utilize hybrid platforms with Intel MIC coprocessors. In this article, we propose an approach for porting a real-life scientific application to such hybrid platforms, assuming no significant modifications of the application code. It allows us to take advantage of all the computing components, including two CPUs and two coprocessors, for the parallel execution of computational workloads. In this study, we focus on the parallel implementation of a numerical model of the dendritic solidification process in isothermal conditions. We develop a sequence of steps that are necessary for the porting and optimization of the solidification application to hybrid platforms with Intel coprocessors. The main challenges include not only overlapping data movements with computations, but also ensuring adequate utilization of cores/threads and vector units of processors, as well as coprocessors. To reach this aim, we propose an efficient and flexible method for the workload distribution between heterogeneous computing components. For implementing the potential benefits of the proposed approach, we choose a heterogeneous programming model based on a combination of the offload mode for Intel MIC and OpenMP programming standard. The developed approach allows us to execute the whole application up to 9.33 3 faster than the original parallel version that uses two CPUs. Furthermore, the CPU-MIC hybrid platforms enable achieving the speedup of about 1.9 3 that of the CPU platform with 24 cores based on the Ivy Bridge architecture, and about 1.5 3 that of the Haswell-based CPU platform with 36 cores.

Strategy for data-flow synchronizations in stencil parallel computations on multi-/manycore systems

“…Another subject of our research is adaptation of the data-flow synchronization strategy to the numerical modeling of 2D solidification problems [12]. The resulting computing scheme belongs to the group of forward-in-time, iterative algorithms.…”

“…In consequence, the five synchronization points are required per every time step. In our previous works [5,12,13], we developed a methodology for porting the parallel code of this application to hybrid platforms with CPUs and Intel Xeon Phi coprocessors, assuming no significant modifications of the code. In particular, we evaluated different setups for the OpenMP parallelization.…”

“…The huge capacity of modern HPC platforms allows complex problems, previously thought impossible, to be solved [12]. Reaching this goal requires to develop more effi-cient programming abstractions enabling better utilization of computing resources [7].…”

“…Another application area tackled in the work refers to the phase-field method, which is a powerful tool for solving interfacial problems in materials science [11]. In this paper, we focus on the parallel implementation of a numerical model of the dendritic solidification process in the isothermal conditions [12]. These two applications belong to the class of the forward-in-time algorithms, which assume the iterative execution of multiple time steps.…”

Performance portable parallel programming of heterogeneous stencils across shared-memory platforms with modern Intel processors

“…With the quick development of computing platforms and software environments, application developers are forced to contend with a variety of parallel architectures. Since applications far outlive any computer system, porting application codes has become a difficult, time-consuming, and significant challenge for scientific and commercial environments (Szustak et al, 2016; Unat et al, 2014).…”