Partial Evaluation for Scienti c Computing : The Supercomputer Toolkit
نویسنده
چکیده
We describe the key role played by partial evaluation in the Supercomputer Toolkit, a parallel computing system for scientiic applications that eeectively exploits the vast amount of parallelism exposed by partial evaluation. The Supercomputer Toolkit parallel processor and its associated partial evaluation-based compiler have been used extensively by scientists at M.I.T., and have made possible recent results in astrophysics showing that the motion of the planets in our solar system is chaotically unstable. 1 Introduction In 1989, researchers at M.I.T. and Hewlett-Packard began a joint eeort to create the Supercomputer Toolkit , a set of hardware and software building blocks to be used for the construction of special-purpose computational instruments for scientiic applications. Earlier work ((6],,7]) had shown that partial evaluation of numerical programs that are mostly data-independent converts a high-level, abstractly speciied program into a low-level, special-purpose program, providing order-of-magnitude performance improvement and exposing vast amounts of low-level parallelism. A central focus of the Supercomputer Toolkit project was to nd a way to exploit this extremely ne-grained parallelism. By combining the performance improvements available from partial evaluation with novel parallel compilation techniques and a parallel processor architecture speciically designed to execute partially evaluated programs, the Supercomputer Toolkit system enabled scientists to run an important class of abstractly-speciied programs approximately three orders of magnitude faster than a conventionally compiled program executing on the fastest available workstation. This paper presents an overview of the role played by partial evaluation in the Supercomputer Toolkit system, describes the novel parallelism grain-size adjustment technique that was developed to make eeective use of the ne-grained parallelism exposed by partial evaluation, and summarizes the various real-world scientiic projects that have made use of the Supercomputer Toolkit system.
منابع مشابه
Partial Evaluation for Scienti c Computing : The Supercomputer Toolkit Experience
We describe the key role played by partial evaluation in the Supercomputer Toolkit, a parallel computing system for scienti c applications that e ectively exploits the vast amount of parallelism exposed by partial evaluation. The Supercomputer Toolkit parallel processor and its associated partial evaluation-based compiler have been used extensively by scientists at M.I.T., and have made possibl...
متن کاملPartial Evaluation for Scienti
We describe the key role played by partial evaluation in the Supercomputer Toolkit a parallel comput ing system for scienti c applications that e ectively exploits the vast amount of parallelism exposed by partial evaluation The Supercomputer Toolkit parallel processor and its associ ated partial evaluation based compiler have been used exten sively by scientists at M I T and have made possible...
متن کاملPartial Evaluation for Scientific Computing: The Supercomputer Toolkit Experience
We describe the key role played by partial evaluation in the Supercomputer Toolkit, a parallel computing system for scientiic applications that eeectively exploits the vast amount of parallelism exposed by partial evaluation. The Supercomputer Toolkit parallel processor and its associated partial evaluation-based compiler have been used extensively by scientists at M.I.T., and have made possibl...
متن کاملExploiting the Parallelism Exposed by Partial Evaluation
We describe an approach to parallel compilation that seeks to harness the vast amount of ne-grain parallelism that is exposed through partial evaluation of numericallyintensive scienti c programs. We have constructed a compiler for the Supercomputer Toolkit parallel processor that uses partial evaluation to break down data abstractions and program structure, producing huge basic blocks that con...
متن کاملParallel Object Instruction Scheduler
We describe an approach to parallel compilation that seeks to harness the vast amount of ne-grain parallelism that is exposed through partial evaluation of numerically-intensive scientiic programs. We have constructed a compiler for the Supercomputer Toolkit parallel processor that uses partial evaluation to break down data abstractions and program structure, producing huge basic blocks that co...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 1994