personal opinion on future role of formal verification in hardware industry
DESCRIPTION
Personal Opinion on Future Role of Formal Verification in Hardware Industry. W. Paul Universität Saarbrücken wiss. Gesamtprojektleiter bmb+f Projekt Verisoft. 3.5 Mio €/year Partners: infineon,… Mission (industrial) Verify high end controllers Mission (academic) Verify stacks - PowerPoint PPT PresentationTRANSCRIPT
Personal Opinion on Future Role of Formal Verification in Hardware
Industry
W. Paul Universität Saarbrückenwiss. Gesamtprojektleiter
bmb+f Projekt Verisoft
Verisoft 1: 2003-2007
• 3.5 Mio €/year• Partners: infineon,…• Mission (industrial)
– Verify high end controllers
• Mission (academic)– Verify stacks– Processor, devices,
drivers, compiler, kernel, OS, client/server, TCP, real time….
• 2003: verified lab prototype (processor, 1.5 Mio gate equivalents)
Industrial Processor VerificationOneSpin Solutions (2006)
• Tools tightly coupled to production flow• Prove hundreds of properties by bounded model
checking (> 90 % effort)• Combine to correctness proof in interactive
prover (< 10 % effort)• Cheaper than conventional testing• New masks…• 14 Mio € Venture Capital
Conclusions 1
• Tools tightly coupled to production flow• It can be done; do it.• Cheaper than conventional testing• Use formal methods as soon as finding bugs by
testing becomes slow• Everybody does it for FPUs.
– among the most complex constructions in computer science
– Learn to do it for more units
Verisoft 2 (2007-2010, planned)
• Formal proof of kernel written in C with in line assembler code expected 2007 (Verisoft 1)
• Mission (Industrial)• Verify and certify OS
related product code,• Partners
– Automotive– Avionics– Microsoft
• High end processors
A Grand Embarassement
• High end processors are flagship products of our field
• We presently don‘t know how to define their ISA formally (i.e. don‘t know what exactly they are doing)
• Necessary for complete formal verification• We better fix that asap
Mathematics Literacy in CS
• Very much of applied computer science as a whole can be presented as a mathematical theory
• Faster/more precise than usual hand waving
• applied CS textbooks will have mix of text/pictures/math familiar from classical engineering disciplines
Conclusion 2The Mathematically Clumsy CS Engineer
• Endangered species
• Near future: jobs moved off shore
• But: off shore countries educate mathematically literate CS engineers…
• Hire mathematically literate engineers as quickly as you can
• Put pressure on universities to educate them