ACCESS IC LAB
Graduate Institute of Electronics Engineering, NTU
101-1 Under-Graduate ProjectTechniques in VLSI design
Speaker: Yu-Min LinAdvisor: Prof. An-Yeu Wu
Date: 2012/10/23
Some slides come from Prof. Chien’s course — DSP in VLSI Design
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page2
OutlineTechniques in VLSI design
Iteration BoundPipelining & ParallelRetimingUnfolding / Folding
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page3
Critical PathThe path with the longest computation time
among all paths that contain zero delays
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page4
Loop BoundLoop — A directed path that begins and ends
at the same node
Loop Bound — The lower bound on the loop computation timeLoop Bound T = t / wt — loop computation timew — # of delays in the loop
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page5
Iteration BoundCritical Loop —
The loop with the maximum loop boundIteration Bound —
Loop bound of the critical loop
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page6
Iteration Bound
Iteration bound is the lower bound on the sample/clock period of the DSP program regardless of the amount of computing resources available
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page7
Pipelining & ParallelPipelining and parallel are the most important
design techniques in VLSI DSP systems
Pipelining — Different function units working in parallel
Parallel — Duplicated function units working in parallel
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page8
Pipelining & Parallel
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page9
PipeliningHow to do pipelining?
Put pipelining registers across an feed-forward cutset
Drawbacks of Pipelining — Increasing latencyIncreasing the number of registers
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page10
Parallel
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page11
Parallel
Drawbacks of Pipelining — Large hardware cost
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page12
RetimingRetiming —
A transformation technique used to change the locations of delay elements in circuit without affecting the input/output characteristics
Applications of retiming — Reducing the clock periodReducing the number of registers
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page13
Retiming
Pipelining is a special case of retiming
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page14
RetimingReducing the clock period —
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page15
RetimingReducing the number of registers —
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page16
UnfoldingUnfolding is similar to parallel processing
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page17
UnfoldingSample/Clock period reduction —
ACCESS IC LAB Graduate Institute of Electronics Engineering, NTU
page18
Folding