Table of Contents
DATAFLOW ARCHITECTURES
Literature
Dataflow Processors - Motivation
Dataflow vs. Control-Flow
Dataflow model of computation
Dataflow languages
Example in VAL
Example in Id
Two important characteristics of dataflow graphs
Dataflow architectures
Pure Dataflow
Static Dataflow
Dataflow graph and Activity template
Acknowledgement signals
MIT Static Dataflow Machine
Deficiencies of static dataflow
Dynamic Dataflow
The U-interpreter (U = unraveling)
The U-interpreter
MERGE and SWITCH nodes
Branch Implementations
L, L-1, D, and D-1 Operators for Loop Implementation
Basic Loop Implementation
A, A-1, BEGIN, and END Operators for function calls
Function application
I-structures (I = incremental)
I-structures
I-structures
I-structure
I-structure select and assign
MIT Tagged-Token Dataflow Architecture
Manchester Dataflow Machine
Advantages and Deficiencies of Dynamic Dataflow
Explicit Token Store Approach
Explicit Token Store
Explicit Token Store Matching Scheme
k-bounded Loop Scheme
k-bounded Loop Scheme - new Operators
Monsoon, an Explicit Token Store Machine
Monsoon, an Explicit Token Store Machine
Monsoon, an Explicit Token Store Machine
Monsoon Prototype
Advantages and Deficiencies of Dynamic Dataflow
Augmenting Dataflow with Control-Flow
Augmenting Dataflow with Control-Flow
Threaded Dataflow
Threaded Dataflow (continued)
Direct token recycling of Monsoon
Epsilon and EM-4
Strongly connected blocks in EM-4
Direct matching: Instruction Memory and Operand Memory
EM-4
Large-Grain (coarse-grain) Dataflow
Large-Grain Dataflow: StarT
Dataflow with Complex Machine Operations
Dataflow with Complex Machine Operations and combined with LGDF
Augsburg Structure-Oriented Architecture (ASTOR)
RISC and Dataflow
RISCifying dataflow
P-RISC Architecture
P-RISC Characteristics
Other Hybrids
Lessons Learned from Dataflow
Comparing dataflow computers with superscalar microprocessors
Lessons Learned from Dataflow (Pipeline Issues)
Lessons Learned from Dataflow (Continued)
Lessons Learned from Dataflow (Continued)
Lessons Learned from Dataflow (Memory Latency)
Lessons Learned from Dataflow (Continued)
Lessons Learned from Dataflow (Continued)
Lessons Learned from Dataflow (alternative instruction window organizations)
|