Systems for Hyperspectral Imaging Cancer Detection Daniel Madroal - - PowerPoint PPT Presentation

systems for hyperspectral imaging
SMART_READER_LITE
LIVE PREVIEW

Systems for Hyperspectral Imaging Cancer Detection Daniel Madroal - - PowerPoint PPT Presentation

Energy Consumption Reduction on High Performance Embedded Systems for Hyperspectral Imaging Cancer Detection Daniel Madroal Universidad Politcnica de Madrid (UPM) School of Telecommunications Engineering and Systems (ETSIST) Research


slide-1
SLIDE 1

Universidad Politécnica de Madrid (UPM) School of Telecommunications Engineering and Systems (ETSIST) Research Center for Software Technologies and Multimedia Systems (CITSEM)

CPS Summer School, Alghero Monday 25th September, 2017

Energy Consumption Reduction on High Performance Embedded Systems for Hyperspectral Imaging Cancer Detection

Daniel Madroñal

slide-2
SLIDE 2
  • D. Madroñal

Energy Consumption Reduction on High Performance Embedded Systems for Hyperspectral Imaging Cancer Detection

2

Introduction

PAPI

Energy estimation Reconfiguration

Real time Energy awareness

slide-3
SLIDE 3
  • D. Madroñal

Energy Consumption Reduction on High Performance Embedded Systems for Hyperspectral Imaging Cancer Detection

3

Procedure

PAPI registers

Total instructions Type of operations Memory ussage

slide-4
SLIDE 4
  • D. Madroñal

Energy Consumption Reduction on High Performance Embedded Systems for Hyperspectral Imaging Cancer Detection

4

Evaluation

8 10 12 14 16 18 1 2 3 4

Power (W) Workpoint

PAPI estimation

8 10 12 14 16 18 1 2 3 4

Power (W) Workpoint

Real measurements

Reconfiguration

slide-5
SLIDE 5
  • D. Madroñal

Energy Consumption Reduction on High Performance Embedded Systems for Hyperspectral Imaging Cancer Detection

5

PhD Research – MPPA Platform

2 I/O (8 DMAs)  16 Clusters  256 PEs @ 400-500MHz

PE PE PE PE PE PE PE PE PE PE PE PE PE PE PE PE Shared Memory (2 MB)

DMA DMA DMA DMA DMA DMA DMA DMA

2GB DDR I/O South I/O North 2GB DDR

slide-6
SLIDE 6
  • D. Madroñal

Energy Consumption Reduction on High Performance Embedded Systems for Hyperspectral Imaging Cancer Detection

6

PhD Research – Preliminary Results

Power (W)

STRESS 8,55 8,93 13,31 15,18 14,55 3,26 0,06 0,05 0,04 29,14 0,82 0,80 0,56

0,01 0,1 1 10 100 8 9 10 11 12 13 14 15 16 0-400-0 1-400-1 256-400-1 256-500-1 256-500-8

Time (s) / Energy (J) Power (W) PEs / Freq / Total DMAs

Power (W) Time (s) Energy (J)

SVM

slide-7
SLIDE 7
  • D. Madroñal

Energy Consumption Reduction on High Performance Embedded Systems for Hyperspectral Imaging Cancer Detection

7

Thank you for your attention

daniel.madronal@upm.es