Reservoir Labs, Inc. — Department of Energy SBIR Phase I: 02b

Reservoir Labs, Inc. — SBIR Phase I award from Department of Energy.

Amount
$229,993
Agency
Department of Energy
Program / Phase
SBIR · Phase I
Topic
02b
Solicitation
DE-FOA-0001618
NAICS
Place of performance
NY
Period
2017-02-21 → 2017-11-20

Description

The need to port highly-optimized existing codes to newer machines and programming frameworks is a longstanding and ongoing issue faced by scientists and application developers. This need is generally addressed by long and tedious manual rewriting of large segments of programs in order to, first, de-optimize the existing implementation and, second, re-optimize it for the new target. The porting task is then generally a source of major costs, productivity losses, bugs and delays. How the Problem is Being Addressed In this project, we will develop a new automatic tool to assist the application developers in porting existing, legacy programs on new platforms. The new tool, Portacy, will automatically analyze previously optimized programs, perform some advanced optimizations such as data layout transformations or automatic communication generation, and finally rewrite the program for another architecture. Portacy will result from the hardening of an existing polyhedral source-to-source code transformation infrastructure. Phase II development will extend the set of optimizations that can be analyzed in the original legacy program to further increase the productivity gain provided by the project. Phase I Project In Phase I, we will focus on hardening the compiler to support a broader range of input programs and implement a new code modernization scheme able to exploit the existing optimizations in legacy programs. Support for programming languages commonly used in scientific computing will be increased by connecting a new compiler front-end to the existing optimization infrastructure. Additional compiler passes will also be implemented in order to automatically analyze, exploit, and revert the optimizations found in the targeted legacy program before re-optimizing the program for a new target platform. All the hardening and improvements will significantly increase user productivity and reduce the level of expertise required to use the automatic tool. Commercial Applications and Other Benefits The developments performed in the project will be integrated in an existing commercial compiler as part of our commercialization plan. The need for a robust code modernization assistant is significant so the market is expected to be large, encompassing very diverse industrial sectors. The developed tool will also be available to the scientific community and to the government, reducing the cost of future code modernization projects.