Turning control flow graphs into function calls: Code generation for heterogeneous architectures

Abstract

Heterogeneous machines are gaining momentum in the High Performance Computing field, due to the theoretical speedups and power consumption. In practice, while some applications meet the performance expectations, heterogeneous architectures still require a tremendous effort from the application developers. This work presents a code generation method to port codes into heterogeneous platforms, based on transformations of the control flow into function calls. The results show that the cost of the function-call mechanism is affordable for the tested HPC kernels. The complete toolchain, based on the LLVM compiler infrastructure, is fully automated once the sequential specification is provided.

Performance Evaluation of a 3D-Stencil Library for Distributed Memory Array Accelerators

“…From the view of the former especially in the HPC field, computer systems have employed high-performance processors equipped with SIMD units, such as SPARC64 with HPC-ACE [7], Intel with SSE/AVX extensions [3], and ARM with Neon extensions [4], or such general-purpose accelerators as GPGPU [6], [9] and Xeon Phi [8]. To im-prove the execution efficiency of these computer systems, new algorithms, efficient compilers, and libraries have been proposed to make the best use of each CPU architecture [10], [15], [16].…”

Aplicaciones del cómputo científico: mantenimiento del software heredado

“…Fortran es unos de los lenguajes de programación de alto nivel más antiguos que ha sido utilizado por los científicos para producir software desde sus orígenes [157,13,151]. Podemos decir que a través del tiempo, éste se ha convertido en el lenguaje de programación científica "De Facto" o por lo menos en el más utilizado [214,17,120,158,44,21,189,19]. La primera versión del manual del usuario de este lenguaje de programación fue publicada el 15 de octubre de 1956, según [13] por el equipo de IBM liderado por J.…”