In Search for a Program Generator to Implement Generic Transformations for High-performance Computing

International audienceThe quality of compiler-optimized code for high-performance applications lags way behind what optimization and domain expertscan achieve by hand. This paper explores in-between solutions,besides fully automatic and fully-manual code optimization. Thiswork discusses how generative approaches can help the design andoptimization of supercomputing applications. It outlines early results and research directions, using MetaOCaml for the design ofa generative tool-box to design portable optimized code. We alsoidentify some limitations of the MetaOCaml system. We finallyadvocate for an offshoring approach to bring high-level and safemetaprogramming to imperative languages