Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han ([email protected]) Brian L....

16
Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han ([email protected]) Brian L. Evans ([email protected]) Earl E. Swartzlander, Jr. ([email protected]) Dept. of Electrical and Computer Engineering The University of Texas at Austin Austin, TX 78712 USA Asilomar Conference on Signals, Systems & Computers November 2nd, 2005

Transcript of Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han ([email protected]) Brian L....

Page 1: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

Low-Power Multipliers withData Wordlength Reduction

Kyungtae Han ([email protected]) Brian L. Evans ([email protected])

Earl E. Swartzlander, Jr. ([email protected])

Dept. of Electrical and Computer EngineeringThe University of Texas at Austin

Austin, TX 78712 USA

Asilomar Conference on Signals, Systems & Computers November 2nd, 2005

Page 2: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

2

Outline

• Introduction

• Wordlength reduction

• Power consumption

• Analysis in switching expectation

• FPGA dynamic power estimation

• Conclusion

Page 3: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

3

Introduction

• Minimize power dissipation due to limited battery power and cooling system

• Multipliers often a major source of power consumption in typical DSP applications

• Multi-precision multipliers can select smaller multipliers (8, 16 or 24 bits) to reduce power consumption

• Wordlength reduction to select any word size [Han, Evans, and Swartzlander 2004]

Page 4: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

4

Wordlength Reduction in Multiplication

• Input data wordlength reduction– Smaller bits enough to

represent, e.g. π x π ≈ 9

• Truncation

• Signed right shift– Move toward the least

significant bit (LSB)

– Signed bit extended for arithmetic right shift

0001 0010 0011 01001101 1100 1010 1001

(a) Original Multiplication

0001 0010 0000 00001101 1100 0000 0000

(b) Reduction by Truncation

0000 0000 0001 00101111 1111 1101 1100

(c) Reduction by Signed Right Shift

Sign bit

Page 5: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

5

Power Reduction via Wordlength Reduction

• Power dissipation – Switching power consumption– Static power consumption

• Switching power consumption– Switching activity parameter, α– Reduce α by wordlength

reduction

clkddLswitching fVCP 2

What is relationship between wordlength and switching parameter, α, in power consumption?

CL Load capacitance

Vdd Operating voltage

fclk Operating frequency

Page 6: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

6

Switching Activity in Multipliers

• Logic delay and propagation cause glitches• Proposed analytical method

– Hard to estimate glitches in closed form– Analyze switching activity w/r to input data wordlength– Does not consider multiplier architecture

• Simulation method– Count all switching activities

(transition counts in logic)– Power estimation (Xilinx XPower)– Considers multiplier architecture

Page 7: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

7

Analytical Method• Consider stream of data for one of the multiplicands• Compare two adjacent numbers in stream after

reduction• Expectation of bit

switching, x, withprobability Px

– L-bit input data– Truncate input data

to M bits (N bits areremoved)

– N-bit signed rightshift in L-bit input(Y is sign bit)

2)(

LXEL

22)(

MNLXEtr

2

)1|(2

1)0|(

2

1)(

L

YXEYXEXErs

L

xX xPxXE

0

)()(

S … …

L bits

M bits N bits

S … …

S S … SS …

Page 8: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

8

Analytical Method

)1|(2

1)0|(

2

1)( YXEYXEXErsX has binomial

distribution

Always L/2 (independent on M and N)

Page 9: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

9

Analytical Method

Input Switching expectation

Full length used

L/2

Truncate N bits

M/2

N-bit signed

right shift

L/2

Wordlength (L) = 16

Page 10: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

10

Wallace vs. Booth Multipliers

Tree dot diagram in 4-bit Wallace multiplier

Radix-4 multiplier based on Booth’s recoding (Χ ● a = P)

Asymmetric (one operand

recoded)

Symmetric

Page 11: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

11

Dynamic Power Consumption for Wallace Multiplier (1MHz)

Reduction(56%)

16-bit x 16-bit multiplier (Simulated on XC3S200-5FT256 FPGA)

Swapping(recode,nonrecode)

Page 12: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

12

Dynamic Power Consumption for Radix-4 Modified Booth Multiplier (1MHz)

Reduction(31%)

Sensitive(13%)

Swapping(recode,nonrecode)

16-bit x 16-bit multiplier (Simulated on XC3S200-5FT256 FPGA)

Page 13: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

13

Conclusion• Truncation to 8 bits reduces est. power consumption

by 56% in Wallace and 31% in Booth 16-bit multipliers

• Signed right shift exhibits no est. power reduction in Wallace multiplier (for any shift) and 25% reduction in Booth multipliers (for 8-bit shift)

• Power consumption in tree-based multiplier– Highly depends on input data– Simulation of all switching activity matches analysis of

switching activity in reduced multiplicands in Wallace mult.

• Operand swapping can reduce power consumption– In Booth multiplier, non-recoded operand 13% more sensitive

in power consumption

Page 14: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

14

Thank You!

Page 15: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

15

Backup Slides

Page 16: Low-Power Multipliers with Data Wordlength Reduction Kyungtae Han (khan@mail.utexas.edu) Brian L. Evans (bevans@ece.utexas.edu) Earl E. Swartzlander, Jr.

16

Dynamic Power Consumption

• 16-bit x 16-bit multiplier (Simulated on XC3S200-5FT256 FPGA)

Wallace multiplier (1 MHz) Radix-4 modified Booth multiplier (1 MHz)

56%31%

Swapping