The Ideal HPC Programming Language
The Ideal HPC Programming Language
Ebony Prioleau
Strayer University
CIS110 Computer Program Design
Douglas Reed
November 30, 2014
“This article describes programmability studies undertaken by Sun Microsystems in its HPCS participation.” FORTRAN has been the preferred language among the High Performance Computing (HPC) community, it is less likely that every programmer would have ended up preferring FORTRAN and more likely that the problems with the original source codes have more to do with reasons than the limitations of existing programming languages. The need for HPC benchmarks in the new languages could reduce source-code volume substantially, but rewriting these benchmarks even in FORTRAN could achieve similar source-code reductions and corresponding improvements in expressivity. This article mentions that the programmer rewrote many of HPC benchmarks and applications using modern FORTRAN taking into account the human costs of software development: programmability and associated characteristics such as readability, verifiability, and maintainability. “These are important considerations; although copy-and-paste is a fast way of writing lines of code, it degrades readability and increases maintenance costs.”
This article focuses a lot on the output of the rewriting activity, inspecting the rewritten HPC programs and causes of source-code bloat. “A key metric was the number of SLOC (source lines of code).” The use of source code, uncovers various indications that the rewritten programs had fewer lines of code, and also that they were easier to read, verify, and modify. The new code is clear, concise, and easy to read. Readability and maintainability of the source code benefited greatly from condensing these transformations into a few functions, but the performance suffered from the extra procedure calls and loss of many specializations and optimizations of the transformations. The battle to deliver good performance on expressive HPC source code must still
Ebony Prioleau
Strayer University
CIS110 Computer Program Design
Douglas Reed
November 30, 2014
“This article describes programmability studies undertaken by Sun Microsystems in its HPCS participation.” FORTRAN has been the preferred language among the High Performance Computing (HPC) community, it is less likely that every programmer would have ended up preferring FORTRAN and more likely that the problems with the original source codes have more to do with reasons than the limitations of existing programming languages. The need for HPC benchmarks in the new languages could reduce source-code volume substantially, but rewriting these benchmarks even in FORTRAN could achieve similar source-code reductions and corresponding improvements in expressivity. This article mentions that the programmer rewrote many of HPC benchmarks and applications using modern FORTRAN taking into account the human costs of software development: programmability and associated characteristics such as readability, verifiability, and maintainability. “These are important considerations; although copy-and-paste is a fast way of writing lines of code, it degrades readability and increases maintenance costs.”
This article focuses a lot on the output of the rewriting activity, inspecting the rewritten HPC programs and causes of source-code bloat. “A key metric was the number of SLOC (source lines of code).” The use of source code, uncovers various indications that the rewritten programs had fewer lines of code, and also that they were easier to read, verify, and modify. The new code is clear, concise, and easy to read. Readability and maintainability of the source code benefited greatly from condensing these transformations into a few functions, but the performance suffered from the extra procedure calls and loss of many specializations and optimizations of the transformations. The battle to deliver good performance on expressive HPC source code must still
References: Loh, Eugene. (2010). The Ideal HPC Programming Language