Automatic Behavioral Transformations

The input HardwareC description is parsed and translated first into an abstract syntax tree representation, which provides the underlying model for semantic analysis and behavioral transformations. The transformations are categorized into user-driven and automatic transformations. User-driven transformations are optional, and allow the designer the capability of modifying the model calls and hierarchy of the input description. They include the following ... 

This chapter is organized as follows. Section 3.1 describes the BIF model. Sections 3.2 and 3.3 present the user-driven and automatic behavioral transformations, respectively. Finally, Section 3.4 summarizes the behavioral transformations applied in Hercules. 

The Hercules system provides an integrated environment for the behavioral synthesis and optimization of hardware behavior. It performs both user-driven and automatic behavioral transformations, and produces as output one or more sequencing graph abstractions of the optimized behavior, described in SIF. The flow of operation in Hercules is described below. 

Perform automatic behavioral optimizations. In-line expansion and operator to library mapping are transformations that are guided by the designer. Upon completion of these optional user-driven transformations, a suite of automatic behavioral transformations is performed to optimize the behavior. Note that these optimizations caimot be applied to a block model because of its declarative semantics (the list of optimizations is described in Section 3.3). Optimizations include compilo -like optimizations such as dead-code elimination and variable unfolding. 

Assuming that a polynomial has been found which adequately represents the response behavior, it is now possible to reduce the polynomial to its canonical form. This simply involves a transformation of coordinates so as to express the response in a form more readily interpreted. If a unique optimum (analogous to a mountain peak in three dimensions) is present, it will automatically be located. If (as is usual in multidimensional problems) a more complex form results, the canonical equation will permit proper interpretation of it. 

In this paper we have introduced the LARES formalism for the specification of dependability models of reconfigurable systems. The formalism follows a structured hierarchical approach and includes features that allow the user to describe arbitrarily complex dynamic behavior. The paper described the textual user interface, concretized by two examples from the literature. Next steps will include the definition of a formal semantics for LARES. Based on this semantics, we will develop model transformations in order to automatically convert different application-specific modeling formalisms to the LARES formalism, and implement interfaces to existing analysis engines, which will be used to perform the quantitative analysis of the specified models. 

Automatic transformations optimize the behavior by performing transformations similar to those found in optimizing compilers [17,16, IS]. The automatic... 

---