Accelerator Data-Path Synthesis for High-Throughput Signal Processing Applications

1 Introduction.- 1.1 Design of ASICs for high-throughput RSP.- 1.2 Overview and contributions of the book.- 2 Applications and Architecture.- 2.1 Characteristics of the application domain.- 2.2 The architectural style.- 2.3 Summary.- 3 The Underlying Synthesis Data Models.- 3.1 Operations and building blocks.- 3.2 Signal flow graph related issues.- 3.3 Summary.- 4 Literature Study.- 4.1 Traditional synthesis.- 4.2 Synthesis with accelerator data paths.- 4.3 Conclusions.- 5 The HLDM Script.- 5.1 The Cathedral-3 script.- 5.2 A script based approach.- 5.3 Initial flow graph transformations.- 5.4 ASU synthesis tasks.- 5.5 Motivation for the script.- 5.6 Synthesis of homogeneous architectures.- 5.7 Summary.- 6 Operation Clustering.- 6.1 Motivation.- 6.2 Related work.- 6.3 Operation clusters.- 6.4 Clustering techniques.- 6.5 Summary.- 7 Cluster to ASU Assignment.- 7.1 The cluster compatibility model.- 7.2 Solution techniques.- 7.3 Assignment in the presence of loops.- 7.4 Extensions.- 7.5 Summary.- 8 ASU Synthesis.- 8.1 The merging problem.- 8.2 Merging of a pair of structures.- 8.3 Extensions of pairwise merging.- 8.4 Ordering schemes for pairwise merging.- 8.5 N-way merging by iterative improvement.- 8.6 Summary.- 9 Demonstrator Designs.- 9.1 A 2 MHz ISDN modem.- 9.2 DCT/IDCT design for a videophone application.- 9.3 Address generation for a 2D motion estimator.- 9.4 Conclusions.- 10 Conclusions.- 10.1 Motivation of our approach.- 10.2 Contributions.- 10.3 Further work.- A Timing Analysis on Flow Graphs.- A.1 The model.- A.2 Timing vector algebra.- A.3 The critical path algorithm.- B The Constructive Clustering Algorithms.- B.1 Definitions.- B.2 Algorithms.- C Algorithms of the Assignment Chapter.- D Assignment Benchmarks.- D.1 The Silage code.- D.2 The results.- E Proof ofTheorem 8.1..- F ASU Synthesis Benchmarks.- F.1 Results for the pair-merging techniques.- F.2 Results for the ordering schemes.- F.3 Area correlation.- References.