C Compilers for ASIPs: Automatic Compiler Generation with LISA

ISBN-13: 9781441911759 / Angielski / Twarda / 2009 / 240 str.

Manuel Hohenauer; Rainer Leupers

Opis

1. 1 Motivation Digital information technology has revolutionized the world during the last few decades. Todayabout98%ofprogrammabledigitaldevicesareactuallyembedded 132]. Theseembeddedsystemshavebecomethemainapplicationareaofinfor- tiontechnologyhardwareandarethebasistodeliverthesophisticatedfunctionality of today's technical devices. As shown in Fig. 1. 1(a), current forecasts predict a worldwideembeddedsystemmarketof$88billionin2009. Millions of Gates 40 25% 300 2004 35 Available Gates 2009 250 20% Used Gates AAGR% 30 200 Design Productivity Gap 25 15% Design Productivity Gap 20 150 10% 15 100 10 5% 50 32 55 25 50 5 20 47 43 10 8 123 0,8 0 0 0% 1993 1995 1997 1999 2001 2003 2005 America Europe Japan Asia-Pacific (a) Global embedded systems revenue and (b) Crisis of complexity 217] average annual growth rate(AAGR) 103] Fig. 1. 1 Embeddedsystemdesign Overthepastfewyears, theever-increasingcomplexityandperformancerequi- mentsofnewwirelesscommunications, automotiveandconsumerelectronicsapp- cations are changing the way embedded systems are designed and implemented today. InconformitywithMoore'slaw 99], onedrivingforceistherapidprogress in deep-submicron process technologies. Chip designers and manufacturers have constantly pushed the envelope of technological and physical constraints. In fact, designers have more gates at their disposal than ever before. However, current M. Hohenauer, R. Leupers, C Compilers for ASIPs, 1 DOI10. 1007/978-1-4419-1176-6 1, C SpringerScience+BusinessMedia, LLC2010 $Billions 2 1 Introduction mainstream-embeddedsystemdesignsarenotusingatleast50%ofthesiliconarea availabletothem(Fig. 1. 1(b)). Thegrowthindesigncomplexitythreatenstooutpace thedesigner'sproductivity, onaccountofunmanageabledesignsizesandtheneed formoredesigniterationsduetodeep-submicroneffects. Thisphenomenonisalso referredtoas crisis of complexity 103]andcomesalongwithexponentiallygr- ing non-recurring engineering (NRE) costs (Fig. 1. 2) to design and manufacture chips. Understandably, thesecostsonlyamortizeforverylargevolumesorhigh-end products. $100. 000. 000. 000,00 $10. 000. 000. 000,00 $1. 000. 000. 000,00 $100. 000. 000,00 RTL Methodology Future Improvements $10. 000. 000,00 1990 1995 2000 2005 2010 2015 Fig. 1.

Spis treści

ASIP Design Methodology.- A Short Introduction to Compilers.- Related Work.- Processor Designer.- Code Selector Description Generation.- Results for Semantics based Compiler Generation.- SIMD Optimization.- Predicated Execution.- Assembler Optimizer.- Summary.

Blurb

C Compilers for ASIPs: Automatic Compiler Generation with LISA

by:

Manuel Hohenauer

Rainer Leupers

The ever increasing complexity and performance requirements of modern electronic devices are changing the way embedded systems are designed and implemented today. The current trend is towards programmable System-on-Chip platforms which employ an increasing number of Application Specific Instruction-set Processors (ASIPs) as building blocks. ASIP design platforms comprise retargetable software development tools that can be adapted quickly to varying target processor configurations. Such tools are usually driven by a processor model given in an Architecture Description Language (ADL), such as LISA. One of the major challenges in this context is retargetable compilation for high-level programming languages like C. First of all, an ADL must capture the architectural information needed for the tool generation in an unambiguous and consistent way. This is particularly difficult for compiler and instruction-set simulator. Moreover, there exists a trade-off between the compiler's flexibility and the quality of compiled code.

This book presents a novel approach for ADL-based instruction-set description in order to enable the automatic retargeting of the complete software toolkit from a single ADL processor model. Additionally, this book includes retargetable optimization techniques for architectures with SIMD and Predicated Execution support. Both allows a high speedup in compiler generation and combines high flexibility with acceptable code quality at the same time. Coverage includes a comprehensive overview of retargetable compilers and ADL based processor design, a methodology and related toolkit to generate a C-compiler fully automatically from an ADL processor model, and retargetable code optimization techniques.

• Presents a strong background and various perspectives of architecture description language (ADL)-based processor design and the retargetable compilation problem;
• Provides the history of ADL based processor design, making the reader knowledgeable about the past research as well as the difficulties faced over time;
• Offers an ADL based modelling formalism and corresponding implementation methods, which can be used for automatic compiler retargeting to quickly obtain compiler support for newly developed ASIPs;
• Presents retargetable optimization techniques for common ASIP features, which can be quickly adapted to varying target processor configurations and help to meet the stringent performance requirements of embedded applications.