OpenCL performance portability for general-purpose computation on graphics processor units: an exploration on cryptographic primitives

摘要

The modern trend toward heterogeneous many-core architectures has led to high architectural diversity inrnboth high performance and high-end embedded systems. To effectively exploit the computational resourcesrnof such a wide range of architectures, programming languages and APIs such as OpenCL have becomernincreasingly popular. Although OpenCL provides functional code portability and the ability to fine tunernthe application to the target hardware, providing performance portability is still an open problem. Thus,rnmany research works have investigated the optimization of specific combinations of application and targetrnplatform. In this paper, we aim at leveraging the experience obtained in the implementation of algorithmsrnfrom the cryptography domain to provide a set of guidelines for modern many-core heterogeneous architecturernperformance portability and to establish a base on which domain-specific languages and compilerrntransformations could be built in the near future. We study algorithmic choices and the effect of compilerrntransformations on three representative applications in the chosen domain on a set of seven target platforms.rnTo estimate how well the application fits the architecture, we define a metric of computational intensity bothrnfor the architecture and the application implementation. Besides being useful to compare either differentrnimplementation or algorithmic choices and their fitness to a specific architecture, it can also be useful to therncompiler to guide the code optimization process.