(CLOSED) DE-FOA-0002460 - X-Stack: Programming Environments for Scientific Computing

The DOE SC program in Advanced Scientific Computing Research (ASCR) hereby announces its interest in basic research in computer science exploring innovative approaches to creating, verifying, validating, optimizing, maintaining, and executing scientific software targeting distributed, heterogeneous, high-performance computing platforms.

Institutional Limit: No more than two pre-applications or applications as the lead institution in a multi-institutional team.

Research Areas

Each pre-application and application must address at least one of two research areas:
1. Automated test synthesis: Innovative methods for automatically generating tests for scientific applications to ensure that, within scientifically-relevant regimes, their behavior meets user-defined requirements and they function with the required level of equivalence across different distributed, heterogeneous, HPC platforms. All aspects of program behavior can be considered, including performance. Additional considerations may include, but are not limited to, the following objectives and the tradeoffs between them:

• How to best interact with human programmers to obtain otherwise-unavailable information and/or present results.
• How to minimize the number, size, and/or complexity of the generated tests.
• How to minimize the execution time and/or usage of expensive/large-scale resources for testing.
• How to maximize the likelihood of discovering defects and/or the determinism of the testing process.
• How to provide information helpful to isolating the causes of failures.

2. Parallel-programming-model translation: Innovative methods enabling the transformation of scientific applications that make use of one parallel-programming model into applications that target heterogeneous systems using a different parallel programming model. Additional considerations may include, but are not limited to, the following objectives and the tradeoffs between them:

• How, and to what extent, can behavioral equivalence be preserved by the translation process.
• How to minimize the size and/or complexity of the generated code.
• How to maximize the performance, readability, and/or naturalness (e.g., similarity to idiomatic usage of language constructs) of the generated code.
• How to enable the continued maintenance of the transformed application.

These methods should use any techniques that might be necessary to accomplish their goals, including, but not limited to, machine learning, natural-language processing, formal reasoning, instrumentation, and compiler technology. The sustainability and explainability of scientific software is critically important to the scientific process, and as a result, particular consideration should be given to maximizing the extent to which human programmers understand the outputs of these methods.