1 / 21

Probabilistic CMOS & Probabilistic Boolean Logic

Probabilistic CMOS & Probabilistic Boolean Logic. Zheng -Shan Yu & Ching - Yi Huang & Fox 2013/01/21. Outline. PCMOS CMOS with noise CMOS with varying voltages Correctness v.s . energy PBL Previous work Research direction Future work. CMOS with noise. NOT:.

Download Presentation

Probabilistic CMOS & Probabilistic Boolean Logic

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. Probabilistic CMOS & Probabilistic Boolean Logic Zheng-Shan Yu & Ching-Yi Huang & Fox 2013/01/21

  2. Outline • PCMOS • CMOS with noise • CMOS with varying voltages • Correctness v.s. energy • PBL • Previous work • Research direction • Future work

  3. CMOS with noise • NOT:

  4. CMOS with noise (cont.) • Interconnection noise • Simultaneous switching noise • Crosstalk • Thermal noise

  5. CMOS with noise (cont.) • Noise models of PCMOS:

  6. CMOS with varying voltages With noise Add noise Without noise Raise voltage to avoid noise

  7. Correctness v.s. energy • Energy per switching: Energy ratio p

  8. Probabilistic Boolean logic • Probabilistic OR: ∨p • Probabilistic AND: ∧p • Probabilistic NOT: ¬p • Example: ∨p , assume p=0.9

  9. Probabilistic Boolean logic • A probabilistic Boolean formula & circuit • F = (a ∧0.8 b)∨0.7(¬0.8c) a n 0.8 b 0.7 F 0.8 c m

  10. Probabilistic Boolean logic (cont.) • OR: Pout = (1–Pa)*(1–Pb)*(1–p) + [1–(1–Pa)*(1–Pb)]*p • AND: Pout = Pa*Pb*p + (1 – Pa*Pb)*(1–p) • NOT: Pout= (1–Pin)*p + Pin*(1–p) Pa Pout p Pb Pa Pout p Pb p Pout Pin

  11. Probabilistic Boolean logic (cont.) • F = (a ∧0.8 b)∨0.7(¬0.8c) • P(n) = 1*1*0.8 + 0*0*0.2 = 0.8 • P(m) = 0*0.8 + 1*0.2 = 0.2 • P(F) = 0.2*0.8*0.3 + (1-0.8*0.2)*0.7 = 0.636 1 n 0.8 1 0.7 F 0.8 1 m

  12. Probabilistic Boolean logic (cont.) • F = (a ∧0.8 b)∨0.7(¬0.8c) • When (AND,NOT,OR) = (C,C,C) (C,I,C) (I,I,C) (I,C,I), F=1 • The probability of each situation 0.448, 0.112, 0.028, 0.048 • P(F=1) = 0.636 1 n 0.8 1 0.7 F 0.8 1 m

  13. Previous work • A Statistics Approach to Correctness Analysis for Probabilistic Boolean Circuits • Problem formulation: • Given a probabilistic Boolean circuit and an input pattern • Report the output probabilities of correctness • Monte Carlo Method • Random Pattern Generation • Sampling Rule • Scoring • Error Estimation

  14. Previous work (cont.) • Monte Carlo simulation • simulate until reaching the stop point RPG 1 n 0.8 1 0.7 F 0.8 1 ‧ ‧ ‧ m ‧ ‧

  15. Research direction • Problem formulation 1: • Given a deterministic Boolean circuit and a correctness constraint • Report a low-energy-consumption probabilistic Boolean circuit satisfying the correctness constraint Low power & Correctness >N Required Correctness >N Our program ///////////////////////////////////////////////////////////////////////////

  16. Research direction (cont.) • Problem formulation 2: • Given a deterministic Boolean circuit and an energy constraint • Report a high-correctness probabilistic Boolean circuit satisfying the energy constraint High correctness & energy consumption <N Required energy consumption <N Our program ///////////////////////////////////////////////////////////////////////////

  17. Research direction (cont.) • Correctness • If the correct value at the output under an input pattern is 0 • c i= 1 – p1i • If the correct value at the output under an input pattern is 1 • c i= p1i • For all n input patterns • Average correctness =

  18. Research direction (cont.) • Function control • Assign probability parameters • Add side inputs • Issue • Which gates are good choices to assign? • Which probabilities should be assigned? Add p 0.9 Add a side input 0.8 0 0.9 0.9

  19. Research direction (cont.) A x C y • Assign probability parameters • Low testability B 0.9 F z A x 0.9 C y B F z

  20. Research direction (cont.) • Energy control • Basic energy consumption • Issue • Load capacitance changes in different manufacture processes. Static power? Per switching? 0.9 0.8

  21. Future work • Function control • Testability • Benefits of adding a side input • Output probabilities under all input patterns • More… • Energy control • Energy consumption of a basic probabilistic Boolean gate • Hspice • Cell libraries • More…

More Related