A comparative study of automatic vectorizing compilers

David Levine, David P. Callahan, Jack J. Dongarra · 1993

. We compare the capabilities of several commercially available, vectorizing Fortran compilers using a test suite of Fortran loops. We present the results of compiling and executing these loops on a variety of supercomputers, mini-supercomputers, and mainframes. 1 Introduction This paper describes the use of a collection of Fortran loops to test the analysis capabilities of automatic vectorizing compilers. An automatic vectorizing compiler is one that takes code written in a serial language (usually Fortran) and translates it into vector instructions. The vector instructions may be machine specific or in a source form such as the proposed Fortran 90 array extensions or as subroutine calls to a vector library. Most of the loops in the test suite were written by people involved in the development of vectorizing compilers, although several we wrote ourselves. All of the loops test a compiler for a specific feature. These loops reflect constructs whose vectorization ranges from easy to ch...

Read the paper · More papers on PaperTik