Edward Lee and Stephen Neuendorffer EECS Department University of California at Berkeley Berkeley, CA 94720, U.S.A.

Abstract. New embedded signal processing architectures are emerging that are composed of loosely coupled heterogeneous components like CPUs or DSPs, specialized IP cores, reconfigurable units, or memories. We believe that these architectures should be programmed using the Process Network model of computation. To ease the mapping of applications, we are developing the Compaan compiler that automatically derives a Process Network (PN) description from an application written in input Matlab. In this paper, we investigate a particular problem in nested loop programs, which is about classifying the interprocess communication in the PN representation of the nested loop program. The global memory arrays present in the Matlab code have to be replaced by a distributed communication structure used for sending data to the network processes. We will show that four types of communication exists, each exhibiting different requirements when realizing them in hardware of software. We present two compile time tests that decide the type of the communication corresponding to a particular static array. These tests are based on Integer Linear Programming and have become an important part of our Compaan compiler. 1

DCharts, a formalism for modeling and simulation of complex reactive software systems, is proposed and studied. The DCharts formalism is based on UML statecharts and DEVS, but provides better modularity and expressiveness. DCharts semantics is rigorously defined in both an operational way and in a denotational way. Abstract, textual, and visual syntax for DCharts are presented. SVM, a DCharts simulator implemented in Python, is presented. It accepts textual model descriptions and simulates them. Multiple types of simulations, as well as real-time execution, are discussed in detail with examples. Model verification is supported by means of repeated simulations in SVM and rule-checking of the simulation traces with extended regular expressions. SCC is a tool to synthesize executable code from DCharts models. It statically optimizes the models to achieve high run-time performance. Multiple target languages are supported. Applications of the DCharts formalism are studied, by means of the the above-mentioned tools. They demonstrate how DCharts are ready for practical use.

Abstract. Although current programming models provide adequate performance, many prove inadequate to support the effective development of efficient Grid applications. Many of the hard issues, such as the dynamic nature of the Grid environment, are left to the programmer. We are developing a programming model that incorporates a familiar, formal computational model and a reflective interface. The programming model, called PAGIS, provides a desirable abstract computer with an interface to introduce and customize Grid functionality. Using PAGIS, an application programmer constructs applications that are implicitly parallel and distributed transparently. This paper describes the basic components of the PAGIS framework for constructing and executing applications, and the reflective techniques to customize applications for computation on the Grid. 1

The COMPAAN/LAURA [18] tool chain maps nested loop applications written in Matlab onto reconfigurable platforms, such as FPGAs. COMPAAN rewrites the original Matlab application as a Process Network in which the control is parameterized and distributed. This control is given as parameterized polytopes that are expressed in terms of pseudo-linear expressions. These expressions cannot always be mapped efficiently onto hardware as they contain multiplication and integer division operations. This obstructs the data flow through the processes. Therefore, we present in this paper the Expression Compiler that efficiently maps pseudolinear expressions onto a dedicated hardware datapath in such a way that the distributed and parameterized control never obstructs the data flow through processors. This compiler employs techniques like number theory axioms, method of difference, and predicated static single assignment code. 1