Many parallel computer architectures are especially suited for particular classes of applications. However, there are only a few parallel architectures equally well suited for standard programs. Much effort is invested into research in compiler techniques to make programming parallel machines easier.
This book presents methods for automatic parallelization, so that programs need not to be tailored for specific architectures; here the focus is on fine-grain parallelism, offered by most new microprocessor architectures. The book addresses compiler writers, computer architects, and students by demonstrating the manifold complex relationships between architecture and compiler technology.
Author(s): Günter Böckle (eds.)
Series: Lecture Notes in Computer Science 942
Edition: 1
Publisher: Springer-Verlag Berlin Heidelberg
Year: 1995
Language: English
Pages: 193
Tags: Programming Languages, Compilers, Interpreters; Control Structures and Microprogramming; Arithmetic and Logic Structures; Logic Design; Processor Architectures; Programming Techniques
Introduction....Pages 1-2
Kinds of parallelism....Pages 3-7
Architectures for fine-grain parallelism....Pages 8-17
VLIW machines....Pages 18-27
Constraints on VLIW architectures....Pages 28-31
Architectural support for exploitation of fine-grain parallelism....Pages 32-37
Constraints for instruction scheduling....Pages 38-39
Instruction-scheduling methods....Pages 40-61
Developing instruction-scheduling methods....Pages 62-62
Tools for instruction scheduling....Pages 63-64
The machine model....Pages 65-66
The horizontal instruction-scheduler....Pages 67-141
Resource management....Pages 142-155
Exceptions....Pages 156-157
Vertical instruction-scheduling....Pages 158-163
Conclusion....Pages 164-164
