Top Banner
Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University
13

Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Dec 22, 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: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Chess ReviewMay 11, 2005Berkeley, CA

Platform Modeling and Analysis

Presented byTivadar SzemethyISIS, Vanderbilt University

Page 2: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 2

Model-Based Design with Components

Design Modelcomponents interacting

according to a “Model of Computation”

Runtime Platformabstraction of

HW/SW/MW services

Platform

HW CPU

Network

HW CPU

Comp.

Comp.

Comp. Comp.

Comp.

Comp. Comp.

Component-Based System

System Synthesis:1. Map the design-time components into platform

objects2. Enforce interaction rules using platform services

Page 3: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 3

Analysis model for verification

• Verification:– does it satisfy requirements specification ?

• Need design requirements, in terms of– observable MoC events (mapped to Platform)– Platform quantities (resources)

Analysis necessitates Platform-level model

Behavior := System Model + Components + MoC semantics + Platform

properties

Page 4: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 4

Platform-level analysis model

• Purpose – formal verification (ideal)– simulation (at least)

• Language for analysis model:– SMV/SPIN model, Timed/Hybrid Automata…– Simulink…

• To be automatically constructed based on– MoC semantics (formal, well-defined)– Platform properties (?)

Page 5: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 5

Example: SMOLES -> UPPAAL 1.

• Application: Camera tracking an object

• Simple MOdeling Lang. for Embedded Sys.– Dataflow-oriented DSML in GME– Components, Ports, Triggers, Timers, Methods

• Platform: DataFlow Kernel– simple OO asynch. dataflow engine in C++/Java– SMOLES model interpreter generates code

• Analysis model: UPPAAL Timed Automata

Page 6: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 6

Example: SMOLES -> UPPAAL 2.

Page 7: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 7

Example: SMOLES -> UPPAAL 3.

KernelrunTimer1

runComponent1

g: (running == kernel &&Timer1Clk > Period1 &&Timer1OutBuff.avaliable())u: Timer1Clk.reset()u: Timer1OutBuf++

.

.

.g: (running == kernel &&

UComp1.Methodm.Triggert)u: running:=Comp1

schedule

g: running == kernel

g: (running == kernel &&

not(UCompn.Methodm.Triggert) &&

not(UTimernClk>Periodn)u: IdleClk.reset()

inv: IdleClk

<IdlePeriod

idle

g: IdleClk >= IdlePeriod

Idle

.

.

.

Method1

g: (running == me &&trigger1 == true)u: input_token_cnt--

inv: (clk < WCET)

g: (clk >= BCET)u: output_token_cnt++, running:=Kernel

Componenti

Methodn

Page 8: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 8

Example: SMOLES -> UPPAAL 4.

Translation rulesetinput: SMOLES modeloutput: UPPAAL TAimplicit: DFK “internals”

Platform-level modelone TA per componentone TA for KernelTranslation using

graph transformations

GReAT graph rewriting tool (GME add-on)

Page 9: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 9

Lessons learned

UPPAAL analysis model too restrictiveno preemptive

schedulingonly for timing analysisno higher-level structures

“Intermediate format”e.g. IF or Metropolishigher-level languageprovides mapping to

multiple analysis tools

DFK model was implicit in transformation:complex, monolithic

transformation spec.

Platform model as1) “Skeleton” for

1) Components2) Kernel

2) Rules to construct synchronizers/guards

3) Composition

results published in:

“Platform Modeling and Model Transformations for Analysis”in Journal of Universal Computer Science vol. 10, pp. 1383-1407, 2004

Page 10: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 10

Generating the analysis model

Well-definedtoolchain

The DSML → IF transformation needs to “know”:• DSML → Platform mapping• Platform → IF mapping• Platform interaction rules

Page 11: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 11

Generating the transformation

Well-definedtoolchain

Encoded in the transformation:• DSML to Platform mapping• Platform to IF (“expansion”)• Platform interaction rules

Transformation generator:

1. rewrite the (DSML→Platform)mapping to (DSML → IF)

2. model composition rules(through Kernel entity, defined in the PM)

Page 12: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 12

The “Big Picture”

Meta-level “operational”

level

Page 13: Chess Review May 11, 2005 Berkeley, CA Platform Modeling and Analysis Presented by Tivadar Szemethy ISIS, Vanderbilt University.

Review, May 11, 2005 13

Preliminary results

• Formalizing the Platform– GME metamodel for DFK– SMOLES → DFK transformation specified as

graph transformation in GReAT

• In search of an Intermediate Format– Evaluating VERIMAG’s IF framework

(metamodel, simple translators and examples)

• Started working on modeling– Giotto with E-machine