This paper introduces the concept of programming with sketches, an approach for the rapid development of high-performance applications. This approach allows a programmer to write clean and portable reference code, and then obtain a high-quality implementation by simply sketching the outlines of the desired implementation. Subsequently, a compiler automatically fills in the missing details while also ensuring that a completed sketch is faithful to the input reference code. In this paper, we develop StreamBit as a sketching methodology for the important class of bit-streaming programs (e.g., coding and cryptography).A sketch is a partial specification of the implementation, and as such, it affords several benefits to programmer in terms of productivity and code robustness. First, a sketch is easier to write compared to a complete implementation. Second, sketching allows the programmer to focus on exploiting algorithmic properties rather than on orchestrating low-level details. Third, a sketch-aware compiler rejects "buggy" sketches, thus improving reliability while allowing the programmer to quickly evaluate sophisticated implementation ideas.We evaluated the productivity and performance benefits of our programming methodology in a user-study, where a group of novice StreamBit programmers competed with a group of experienced C programmers on implementing a cipher. We learned that, given the same time budget, the ciphers developed in StreamBit ran 2.5× faster than ciphers coded in C. We also produced implementations of DES and Serpent that were competitive with hand optimized implementations available in the public domain.

Abstract

No abstract

ÐWe describe a VLIW architecture designed specifically as a target for dynamic compilation of an existing instruction set architecture. This design approach offers the simplicity and high performance of statically scheduled architectures, achieves compatibility with an established architecture, and makes use of dynamic adaptation. Thus, the original architecture is implemented using dynamic compilation, a process we refer to as DAISY (Dynamically Architected Instruction Set from Yorktown). The dynamic compiler exploits runtime profile information to optimize translations so as to extract instruction level parallelism. This work reports different design trade-offs in the DAISY system and their impact on final system performance. The results show high degrees of instruction parallelism with reasonable translation overhead and memory usage.

No abstract