Top Banner
Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010
25

Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

Dec 21, 2015

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

Micro-Architecture Techniques for Sensor Network Processors

Amir Javidi

EECS 598

Feb 25, 2010

Page 2: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

2

Motivation

• Low performance tasks

• Long duration

• Small energy supplies

Page 3: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

3

Papers

• [1] L. Nazhandali, B. Zhai, J. Olson, A. Reeves, M. Minuth, R. Helfand, S. Pant, T. Austin, and D. Blaauw, “Energy optimization of subthreshold voltage sensor network processors,” in Proc. Int. Symp. Computer Architecture, 2005, pp. 197–207.

• [2] S. Hanson, M. Seok, Y-S. Lin, Z. Foo, D. Kim, Y. Lee, N. Liu, D. Sylvester, and D. Blaauw, “A low voltage processor for sensing applications with picowatt standby mode,” IEEE Journal of Solid-State Circuits, pp. 1145-1155, April 2009.

Page 4: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

4

Energy Budget

• 2g Vanadium oxide battery: 720 mAh• Powers ARM 720T processor at 100MHz for

45hrs

• Thin film zinc/silver oxide battery: 100 μAh/cm2, 1.55 V• For area of 1mm2 average current must be 114pA

(power consumption of 177 pW) for 1 year lifetime

Page 5: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

5

Performance Requirement

• Blood pressure monitoring (low rate):• Sensing @ 800 bps 10,000 inst/sec

• EEG brain signal monitoring (high rate):• 3200 bps 56,000 inst/sec for filtering,

analysis, compression, and storage

Page 6: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

6

Architecture/Circuit Techniques• Sub-threshold implementation (Vdd < Vth)

• ISA optimization

• Voltage scaling

• Power gating

• Stack forcing

• Data/instruction compression

Page 7: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

7

Subthreshold design

• Why subthreshold?• Processor operating in lowest super-threshold

voltages deliver too much performance

Performance of sensor network processor applications on embedded targets. Number of times faster than real-time the processor can handle the worst case data stream rate [1].

Page 8: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

8

Subthreshold Circuit Design

Page 9: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

9

Subthreshold Energy Optimization

21 ( )

2dd

inst cycle cycle s dd dd leak clk

kVclk sub

E E CPI E C V V I t

t e

SubthresholdEnergy as a function of Voltage[1]

Vmin

energy optimal supply voltage

Page 10: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

10

ISA Optimization

• Why ISA optimization?• Memory dissipates static/dynamic energy

• Memory size leakage

• Tradeoff between

memory size and

control logic size

Logic Vs memory energy tradeoff [1]

Page 11: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

11

ISA Optimization

• Impact of ISA optimization on code size and control logic complexity

Page 12: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

12

Micro-Architecture

• Sensor network processor micro-architecture

Page 13: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

13

Performance vs. Energy

Page 14: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

14

Results

• Sensor network processor• ROM/RAM memory

• 8 bit data path

• 235 mV supply

• 182 KHz

• 1.38 pJ/inst

• 4.1x faster than necessary for mid-bandwidth

• 25 years lifetime with 2g vanadium oxide battery (720 mAh)

Page 15: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

15

Phoenix Processor

• Focus on lowering standby power

• Older 0.18μm technology

• Custom leakage-optimized instruction set

• Simple data memory compression

• Ultra-low-leakage memory cell

• Huge tradeoff between standby power and area and active energy

Page 16: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

16

Phoenix Processor

Page 17: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

17

CMOS Technology

• Newer technology (65 nm)• High subthreshold leakage

• Small capacitance

• Older technology (180 nm)• 7.7x larger

• 647x less total energy consumption

Page 18: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

18

Voltage Scaling

• Supply voltage of 0.5 V• Mix of subthreshold and near-subthreshold

devices

• Retentive gates high-Vth ~ 0.7 V

• Non-retentive gates medium-Vth ~ 0.5 V

• High-Vth consumes ~ 1000x less leakage power

Page 19: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

19

Power Gating

• Medium-Vth power switch

• Smaller switch ~ 1000x • Less area overhead

• Less charging/discharging power overhead

Page 20: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

20

CPU Architecture

• 2 stage, 8bit data width, 10bit inst. Width

• ALU (add, subtract, shift)

• No multiplier

• Simple decoder (min set of operations)

Page 21: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

21

ISA Optimization

• Minimized instruction width (10 bit)• Reduces IMEM standby power dissipation

• Efficient operand encoding• Explicit operand: more flexibility, more frequently used

• Implicit operand: less flexibility, less frequently used

Page 22: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

22

Memory Design

• 64x10b SRAM (IMEM)• Application specific instructions

• No power gating

• 64x10b ROM (IROM)• Commonly used instructions

• Power gated

• 52x40b SRAM (DMEM) • Data compression

• Fine grain power gating

Page 23: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

23

Memory Design

• Leakage reduction• High-Vth bitcell transistors

• Cross coupled inverters:• Stacked transistors

• Increased length • (0.35μm to 0.50μm)

~2x leakage reduction

• Robustness• Full swing read-buffer

• Power gated

Page 24: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

24

Results

• Phoenix processor• 0.5 V power supply

• 106 KHz

• 2.8 pJ/cycle 297 nW

• 226 nW active mode

• 35.4 pW standby mode

• 915 x 915 μm2

Page 25: Micro-Architecture Techniques for Sensor Network Processors Amir Javidi EECS 598 Feb 25, 2010.

Questions?