Académique Documents
Professionnel Documents
Culture Documents
Guided By:
Presentation By:
Mr.MANU.T.S DHANESH.M.S
Lecturer Roll.No:2007
Dept. of ECE M.Tech (Electronics) with specialization in
VLSI and Embedded systems
TKI, Karuvelil
CONTENTS:
INTRODUCTION
DVFS
TGPM-ALL FLOWCHART
TGPM-ALL ALGORITHM
TGPM-N FLOWCHART
TGPM-N ALGORITHM
PERFORMANCE COMPARISON
CONCLUSION
REFERENCES
2
INTRODUCTION:
Increasing the performance is one of the
fundamental aim of embedded system design.
Our aim is to increase performance while keeping
energy consumption low.
To achieve this we define TGPM algorithm.
3
DYNAMIC VOLTAGE AND FREQUENCY
SCALING(DVFS):
P α f V2
Vαf
Pαf 3
3 approaches: interval, inter-task , intra-task
4
TGPM-ALL ALGORITHM.
TGPM-ALL Start
FLOW CHART:
Pre processing steps
Queue
Compute f*
9
TGPM-ALL ALGORITHM:
10
COMPLEXITY:
Checks test cases until appropriate test cases are
selected.
If no: of test cases are very large and no test case
gives correct frequency, then after examining all the
test case task has to be completed at full speed.
This will increase the complexity with out any
benefit.
11
TGPM-N ALGORITHM.
TGPM-N Start
FLOW CHART:
Pre processing steps & k=0
Queue
Compute f*
Yes 13
Yes Execute task at full
speed. stop
This algorithm keeps track of the number of failed
test case attempts for task T.
If it meets the predefined value N, it will discard
remaining test cases for the task.
Execute the task at full speed until completion.
14
TGPM-N ALGORITHM
15
PERFORMANCE COMPARISON;
EFFECTS OF Poff,C AND ti*off
Effects of these parameters are similar on all
DVFS-based power management schemes.
As (poff/pon(f)) increases, off-chip components
consume relatively more energy.
Increased switching capacitance (C) and off-chip
workload reduce energy-efficient frequencies.
16
EFFECTS OF TEST CASE SIZE ON ENERGY
CONSUMPTION.
20
REFERENCES:
i. Edward Y. Y. Kan , W. K. Chan and T. H. Tse.
Leveraging Performance and Power Savings for
Embedded Systems using Multiple Target
Deadlines, Quality Software (QSIC), 2010 10th
International Conference, IEEE.
ii. T. K. Tan , A. Raghunathan , and N. K. Jha.
Embedded Operating System Energy Analysis
and Macro-modeling, IEEE.
iii. S. Boyd and L. Vandenberghe. Convex
Optimization. Cambridge University Press,
Cambridge, 2004.
iv. http://www.stanford.edu/class/ee364b/lectures/tru
nc_newton_slides.pdf 21