ForOpenCL: Transformations Exploiting Array Syntax in Fortran for Accelerator Programming

摘要

Emerging GPU architectures for high performance computing are well suited toa data-parallel programming model. This paper presents preliminary workexamining a programming methodology that provides Fortran programmers withaccess to these emerging systems. We use array constructs in Fortran to showhow this infrequently exploited, standardized language feature is easilytransformed to lower-level accelerator code. The transformations in ForOpenCLare based on a simple mapping from Fortran to OpenCL. We demonstrate, using astencil code solving the shallow-water fluid equations, that the performance ofthe ForOpenCL compiler-generated transformations is comparable with that ofhand-optimized OpenCL code.