This chapter focuses on another class of target processors, namely, those equipped with deep pipelines and parallel functional units such as VLIW architectures for instance. Such architectures are quite popular in embedded system design since they do not require designs to sacrifice software development productivity for the very high-performance processing needed for today’s applications. Naturally, to achieve their peak performance, all parallel functional units must be kept busy during program execution. Thus, a common hardware features to increase the amount of available instruction-level parallelism (ILP) is predicated execution (PE). Basically, this allows to implement if-then-else (ITE) statements without jump instructions that offers a number of optimization opportunities. Furthermore, PE can enable more
KeywordsCost Computation Implementation Scheme Basic Block Setup Cost Embed Processor
Unable to display preview. Download preview PDF.