1 / 21

JouleTrack - A Web Based Tool for Software Energy Profiling

JouleTrack - A Web Based Tool for Software Energy Profiling. Amit Sinha and Anantha Chandrakasan Massachusetts Institute of Technology June 19, 2001. Outline. Motivation for fast software energy profiling Experimental setup – StrongARM and SH-4

juro
Download Presentation

JouleTrack - A Web Based Tool for Software Energy Profiling

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. JouleTrack -A Web Based Tool for Software Energy Profiling Amit Sinha and Anantha Chandrakasan Massachusetts Institute of Technology June 19, 2001

  2. Outline • Motivation for fast software energy profiling • Experimental setup – StrongARM and SH-4 • Current profiling measurements and implications • Instruction level • Program level • Models • Leakage energy estimation methodology • Processor leakage model • Leakage vs. switching • JouleTrack design • Conclusions

  3. Motivation • Proliferation of portable devices • 50X increase in μP power vs. 4X in battery life • Battery technology lags behind • Programmable solutions preferred • Fast software profiling techniques required

  4. Scenarios • Portable devices downloading apps • Predictive energy estimation • Wireless sensor networks • Lifetime estimation • Energy-Quality tradeoffs • Integrated Development Environments • Energy aware software

  5. Processor Power Supply Memory A A Software Energy Estimation • [Tiwari96] proposed instruction level analysis • Disadvantages • Tedious instruction level analysis • Large simulation time and storage requirement • Very fine grained • Macro-estimation?

  6. Experimental Setup • Intel StrongARM SA-1100 Brutus platform • Hitachi SH-4 based SH7750 platform Freq 59 – 206 MHz VDD 0.8 – 1.5 V SH7750 StrongARM DVS

  7. 45 [Montanaro, JSSC ‘96] 40 35 30 Power (%) 25 20 15 10 5 0 Cache Control GCLK EBOX I/O,PLL SA-1100 Instruction Current • Little variation in instruction currents! Supply VDD 1.5 V Frequency 206 MHz Load/Store 0.196 A ALU 0.178 A

  8. SH-4 Instruction Currents • Similar variation with MACs and Branches consuming more current

  9. Program Current for SA-1100 • Current almost independent of program! • First order model, I = f(VDD,frequency)

  10. Second Order Estimation • Determine energy differentiated instruction classes and weights for those classes SA-1100 Example • Measure statistics for those classes at runtime • Program current = Weighted average • Error < 3%

  11. Benchmark Programs Measure Statistics Measure Current Execute Determine Weights Model Calibration

  12. First and Second Order Models • Second order model reduces average error to 2.8% (based on 66 test programs)

  13. Eleak = VDDI0 exp(Vdd/nVT ) t Eswitch = CtotVDD2 Optimal Voltage Frequency Variation

  14. Total Charge flow Leakage Current (mA) FFT Execution Time Supply Voltage (V) Leakage Model • Leakage currents account for 10% of energy dissipation • Leakage behavior is exponential with supply voltage • I0 = 1.2mA, n = 21.3 for the StrongARM Ileak = I0 exp(Vdd/nVT )

  15. ( ) æ ö - - g × - h × V V V V V GS th 0 S DS DS - ç ÷ nV V = - I Ae 1 e T T ç ÷ ç ÷ è ø W eff 2 1 . 8 = m A C V e 0 ox T L eff Explanation of Exponential Model • Processor leakage model can be explained from transistor subthreshold leakage model • Model error < 6% VDD VDD VDD I3 I2 I1 [Gu96]

  16. Isolating Leakage Energy • Leakage energy component can dominate switching energy for high supply voltage operation!

  17. Duty Cycle : D = T1/T Switching Energy Idle Mode Energy Leakage Energy Time T1 T Task completes Low Duty Cycle Effects • Leakage occurs at all times! D = 100%  Leakage is 10% D = 10%  Leakage > 50% • Just in time processing will reduce leakage energy

  18. Logs Energy Estimation Compiler Linker JouleTrack Design Libraries source.c source.exe Simulator Errors www Common Gateway Interface Cycle countsStatistics Web Server Remote User Energy Profile Average Currents, Weights

  19. Executable Image Clock Speed Memory Map Memory Timing 00000000 8000 SRAM 4 rw 1/1 1/1 00008000 8000 ROM 2 r 100/100 100/100 00010000 8000 DRAM 2 rw 150/100 150/100 Memory Models Co-processor OS Models Profiler Tracer RDI Cycle Accurate Simulator Core Execution Statistics Timing Estimation Engine

  20. JouleTrack Snapshot http://dry-martini.mit.edu/JouleTrack 4000 hits since Sept-2000

  21. Conclusions • Fast macro-level program current estimation demonstrated • No need for elaborate trace profiling • Can be incorporated in an energy aware application development environment • Very little instruction current variation because of common overheads • Leakage currents becoming significant • 10% at 100% duty cycle on StrongARM • Leakage measurement technique presented

More Related