Formal Methods Michael Collins. 18-849, Section B Spring 1999 - - PowerPoint PPT Presentation

formal methods
SMART_READER_LITE
LIVE PREVIEW

Formal Methods Michael Collins. 18-849, Section B Spring 1999 - - PowerPoint PPT Presentation

Formal Methods Michael Collins. 18-849, Section B Spring 1999 Formal Methods Why Formalisms? Relationships Flaws Some systems Conclusions Why Formal Methods? We already can build systems. Roman Engineers built


slide-1
SLIDE 1

Formal Methods

Michael Collins. 18-849, Section B Spring 1999

slide-2
SLIDE 2

Formal Methods

■ Why Formalisms? ■ Relationships ■ Flaws ■ Some systems ■ Conclusions

slide-3
SLIDE 3

Why Formal Methods?

■ We already can build

systems.

■ Roman Engineers

built aqueducts

■ Neither group has

math for the job

■ Both groups waste(d)

time & effort

slide-4
SLIDE 4

Why Formal Methods? (2)

■ Correctness is proven, not observed

◆ Automatic Proof

■ Provides a neutral description

◆ Good for documentation ◆ Good for standardization

■ Legal guarantees

slide-5
SLIDE 5

Relationships

■ SW Reliability

◆ Fault Avoidance

■ Fault Tolerant Computing

◆ Vide Supra

■ Verify/Validate/Certify

◆ Serves as a validation system

■ (Ultra Dependability)

slide-6
SLIDE 6

How do we use them?

■ Build a model using a Modeling Language

◆ Algebra

■ Verify the correctness of the model

◆ Theorem Provers exist (Boyer-Moore)

■ Translate the model to implementation

◆ Again, tools exist

slide-7
SLIDE 7

Flaws

■ Idealized models ■ Design vs. Implementation ■ Learning curve ■ How do you prove a prover? ■ Can’t apply models to existing systems

slide-8
SLIDE 8

LARCH

■ Two-level language

◆ Versions for C++, VHDL...

■ One language for modeling, one for

implementation

■ Similar systems include VDM & Z

slide-9
SLIDE 9

Petri Nets

■ Purely graphical

modeling language

■ Model Concurrency ■ Can be used to

define protocols

slide-10
SLIDE 10

SML

■ Theorem proving language

◆ Literally designed for provability ◆ Functional and strongly typed

■ Proofs are limited in scope

◆ No side effects

■ Similar projects include Haskell

slide-11
SLIDE 11

HOL

■ Higher order logic ■ Mechanized prover ■ Most (in)famous project: Viper Chip

◆ Couldn’t handle interrupts

■ Other provers include Boyer-Moore

slide-12
SLIDE 12

Conclusions

■ Formal methods are attractive in theory ■ Very few benefits right now ■ Current methods provide unsatisfactory

models

■ Engineers may have to start thinking like

mathematicians