national taiwan university peakaso: peak-temperature aware scan- vector optimization minsik cho and...
TRANSCRIPT
National Taiwan UniversityNational Taiwan University
PEAKASO: Peak-Temperature Aware Scan-
Vector Optimization
Minsik Cho and David Z. Pan
Dept. of ECE
The University of Texas at Austin, Austin
VTS’06
National Taiwan University
Outline
• INTRODUCTION• PRELIMINARIES• HEAT DISSIPATION & PEAK TEMPERATURE• PEAKASO ALGORITHM
– Hotspot Prediction:– Global Scan Vector Ordering:– Local Scan Vector Reordering:
• EXPERIMENTAL RESULTS• My Comment
National Taiwan University
Outline
• INTRODUCTION• PRELIMINARIES• HEAT DISSIPATION & PEAK TEMPERATURE• PEAKASO ALGORITHM
– Hotspot Prediction:– Global Scan Vector Ordering:– Local Scan Vector Reordering:
• EXPERIMENTAL RESULTS• My Comment
INTRODUCTION
• Minimizing power consumption can reduce the temperature of CUT,
• but it does not necessarily provide the best solution for peak temperature minimization.– 1. peak temperature is a localized event by nature.– 2. peak temperature depends on both heat generation
from power consumption and heat dissipation to ambient air.
National Taiwan University
National Taiwan University
Outline
• INTRODUCTION• PRELIMINARIES• HEAT DISSIPATION & PEAK TEMPERATURE• PEAKASO ALGORITHM
– Hotspot Prediction:– Global Scan Vector Ordering:– Local Scan Vector Reordering:
• EXPERIMENTAL RESULTS• My Comment
PRELIMINARIES
National Taiwan University
heat conduction equation
Power Model for Scan-based Testing
PRELIMINARIES
National Taiwan University
National Taiwan University
Outline
• INTRODUCTION• PRELIMINARIES• HEAT DISSIPATION & PEAK TEMPERATURE• PEAKASO ALGORITHM
– Hotspot Prediction:– Global Scan Vector Ordering:– Local Scan Vector Reordering:
• EXPERIMENTAL RESULTS• My Comment
National Taiwan University
National Taiwan University
Outline
• INTRODUCTION• PRELIMINARIES• HEAT DISSIPATION & PEAK TEMPERATURE• PEAKASO ALGORITHM
– Hotspot Prediction:– Global Scan Vector Ordering:– Local Scan Vector Reordering:
• EXPERIMENTAL RESULTS• My Comment
PEAKASO ALGORITHM-Hotspot Prediction:
National Taiwan University
• Computational complexity can be reduced if hotspot can be identified.
PEAKASO ALGORITHM-Hotspot Prediction:
National Taiwan University
PEAKASO ALGORITHM-Hotspot Prediction:
• Example:
National Taiwan University
PEAKASO ALGORITHM-Hotspot Prediction:
National Taiwan University
PEAKASO ALGORITHM-Global Scan Vector Ordering:
• Ordering Problem can be converted into a complete graph,
• where each scan vector becomes a vertex, and an edge cost from Vi to Vj on hotspot
• Cost function:
National Taiwan University
PEAKASO ALGORITHM-Global Scan Vector Ordering:
• Example:
National Taiwan University
PEAKASO ALGORITHM-Global Scan Vector Ordering:
National Taiwan University
PEAKASO ALGORITHM-Local Scan Vector Reordering:
National Taiwan University
PEAKASO ALGORITHM-Local Scan Vector Reordering:
• Example:
National Taiwan University
PEAKASO ALGORITHM-Local Scan Vector Reordering:
National Taiwan University
National Taiwan University
Outline
• INTRODUCTION• PRELIMINARIES• HEAT DISSIPATION & PEAK TEMPERATURE• PEAKASO ALGORITHM
– Hotspot Prediction:– Global Scan Vector Ordering:– Local Scan Vector Reordering:
• EXPERIMENTAL RESULTS• My Comment
EXPERIMENTAL RESULTS
National Taiwan University
National Taiwan University
Outline
• INTRODUCTION• PRELIMINARIES• HEAT DISSIPATION & PEAK TEMPERATURE• PEAKASO ALGORITHM
– Hotspot Prediction:– Global Scan Vector Ordering:– Local Scan Vector Reordering:
• EXPERIMENTAL RESULTS• My Comment
My Comment
• Peak-Temperature is different from Peak-Power.
• Reordering does not affect peak-power very much.
• We can add Peak-Temperature minimization to our Low-Power ATPG : LPtest.
National Taiwan University