Re: Fuzzy Logic Operating Systems

Feb 13, 2006 3 Replies

In comp.os.linux.misc Jean-David Beyer wrote: :> : Let me remark that the the things I _insist_ upon in a piece of computer : hardware or operating system is predictability, repeatability, and : deterministic behavior. If I do the same thing (e.g., run a program with the



So you don't use any programs on binary computers that do floating point math? If so, ever heard of round-off error or wonder why



5/5 is not necessarily equal to 2-1?

Don't foget, ALL binary computers work in finite precision, so insist away!



Stan


Stan Bischof ("stan" at the below domain) www.worldbadminton.com

You are missing the point. What you get when doing floating point arithmetic on any given machine, with a given compilation system _is_ deterministically defined. So if floating point 5 divided by floating point 5 does not give exactly floating point zero (although in every case I know of, it does), and floating point 2 minus floating point 1 does not give exactly floating point

1 (and it does in every case I know of), the floating point difference between the ratio and the difference, above, is _always the same_ or the machine is broken.

You do not get it.

.~. Jean-David Beyer Registered Linux User 85642. /V\ PGP-Key: 9A2FC99A Registered Machine 241939. /( )\ Shrewsbury, New Jersey http://counter.li.org ^^-^^ 18:20:00 up 24 days, 9:48, 4 users, load average: 4.15, 4.16, 4.16

In comp.os.linux.misc Jean-David Beyer wrote: :> :> Don't foget, ALL binary computers work in finite precision, so insist :> away! :> : You do not get it.

Sure- if you use exactly the same software running on exactly the same hardware starting from exactly the same initial conditions, and with all exactly the same inputs, in a deterministic system you end up in exactly the same ending point.

However, that would apply to all binary computers running any OS/software, barring true random events like alpha-particle induced flipped bits and the like, no?

can you name ANY OS/software that isn't determinitstic given this restrictivce definition?

I was pointing out that if you are a little less restrictive then you end up with all sorts of affects - like FP bit errors- that creep in and appear to lend a randomness.

regards

Stan

Stan Bischof ("stan" at the below domain) www.worldbadminton.com

Jean-David> You are missing the point. What you get when doing Jean-David> floating point arithmetic on any given machine, with a Jean-David> given compilation system _is_ deterministically Jean-David> defined. So if floating point 5 divided by floating Jean-David> point 5 does not give exactly floating point zero Jean-David> (although in every case I know of, it does),=20

In very case I know of, it doesn't. 5/5 !=3D 0 That's simple arithmetic that one learns in primary school.

Jean-David> and floating point 2 minus floating point 1 does not Jean-David> give exactly floating point 1 (and it does in every Jean-David> case I know of), the floating point difference between Jean-David> the ratio and the difference, above, is _always the Jean-David> same_ or the machine is broken.

--=20 Lee Sau Dan =A7=F5=A6u=B4=B0 ~= {@nJX6X~}

E-mail: snipped-for-privacy@informatik.uni-freiburg.de Home page:

formatting link

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required