On Jan 28, 6:17 am, Paul Rubin <http://[EMAIL PROTECTED]> wrote: > Paddy <[EMAIL PROTECTED]> writes: > > Given the complexity of current microprocessors i'm guessing that > > their previous testing methods would be too good to just junk in > > totality because the FDIV bug was not found. Similarly if they were > > not using formal methods then it makes sense to add it too your > > arsenal; and unfortunately it takes a mistake like that to allow > > different methods to be explored and incorporated. > > Fair enough. My main issue was against the notion that random testing > is the only thing necessary.
Sorry Paul if I may have given that impression, its just that when you bring in random testing to a design that until then had only directed tests you can see the bug rate jump up! Think of a hysteresis curve that has gone flat with current testing methods as not many new bugs are being found; add a new test methodology - random testing and you get a new hysteresis curve added as bugs found jump up again. We eventually ship the chip and get awarded by one of our large customers for quality - which happened - so thats why I put it forward. - Paddy. -- http://mail.python.org/mailman/listinfo/python-list
