why?? it should be another thing!! 1

• Sep 20, 2006

1
• #2

IRstuff

Aerospace

Jun 3, 2002

44,888

US

Why would you think that a numerical processor has infinite precision?

TTFN

 

Upvote 0 Downvote

• Sep 22, 2006

• #3

corus

Mechanical

Nov 6, 2002

3,165

GB

The exponential is an infinite series. If you have an infinite amount of time you can get a better answer.

corus

 

Upvote 0 Downvote

• Sep 23, 2006

• #4

FrancisL

Computer

Aug 29, 2001

337

GB

"If you have an infinite amount of time you can get a better answer"
Not quite, you would also need a computer with an infinite number of bits. That is to say this is a quantisation problem not a time one.

 

Upvote 0 Downvote

• Sep 23, 2006

• #5

johnwm

Computer

Sep 26, 2002

806

GB

Actually FrancisL since its a converging infinite series any computer will do - you just need a bigger stack of paper in your printer

Good Luck
johnwm
________________________________________________________
To get the best from these forums read faq731-376 before posting

Steam Engine enthusiasts:

http://www.essexsteam.co.uk

 

Upvote 0 Downvote

• Sep 25, 2006

• #6

zeitghost

Computer

Oct 8, 2003

400

GB

So long as you don't use floating point... the source of the rounding errors.

bcd maths would be better...

 

Upvote 0 Downvote

• Sep 25, 2006

• #7

IRstuff

Aerospace

Jun 3, 2002

44,888

US

bcd math would have to be floating point as well, to represent cos(). These are irrational numbers and cannot be represented in any format exactly.

TTFN

 

Upvote 0 Downvote

• Sep 25, 2006

• #8

FrancisL

Computer

Aug 29, 2001

337

GB

That is hat I was trying to say earlier!
Francis

 

Upvote 0 Downvote

• Sep 25, 2006

• #9

zeitghost

Computer

Oct 8, 2003

400

GB

Indeed.

But bcd math doesn't have the binary/decimal conversion error.

)

 

Upvote 0 Downvote

• Sep 25, 2006

• #10

IRstuff

Aerospace

Jun 3, 2002

44,888

US

The conversion error is at the limit of the machine's precision, so it really doesn't matter whether you were able to express 0.1 exactly in BCD, when there is an infinite number of cases that BCD doesn't help anyway.

TTFN

 

Upvote 0 Downvote

• Sep 25, 2006

• #11

johnwm

Computer

Sep 26, 2002

806

GB

You can calculate using scaled integers to any arbitrary precision, as long as the function can be reduced to a converging series. About 20 years ago I was teaching a BASIC programming class in which we calculated pi and e to 100000 places using integer arithmetic.

Took some time to calculate though

Good Luck
johnwm
________________________________________________________
To get the best from these forums read faq731-376 before posting

Steam Engine enthusiasts:

http://www.essexsteam.co.uk

 

Upvote 0 Downvote