Re: [ECOS] One question.

[Michele Portolan <michele.portolan@imag.fr>]

Zhao.Green@inventec-inc.com wrote:
> Dear Sir:
> 	Sure, if I try to less precision in output. The result can be round back to 0.1.
> But if  the dnum1 = 0.00000000001 , dnum2 = dnum1.
> Result = dnum1+dnum2 
> The result is 0.00000000002
> This amaze me.
> 
This is quite normal: floating point use relative precision (depends on the size of the mantissa. When you write 0.00000000001 all zero are normalized, but when you use 1234567 0.1 must be too little if compared to 1234567 in your representation.

> When dnum1  = 0.00000000001, the register show this value is 9.9999999999999994e-12.

You should try to see the binary representation, once more there must is too much precision in the visualisation primitive.


hoping to have helped you,

Michele
> 
> 
> 
> BR
> Green
> 
> 
> 
> -----Original Message-----
> From: ecos-discuss-owner@ecos.sourceware.org [mailto:ecos-discuss-owner@ecos.sourceware.org] On Behalf Of Paul D. DeRocco
> Sent: Wednesday, May 31, 2006 2:29 PM
> To: eCos Discuss
> Subject: RE: [ECOS] One question.
> 
>> From: Zhao.Green@inventec-inc.com
>>
>> 	I use the ecos package , and I found the double type question Double 
>> dnum1 = 1234567.1 Double dnum2 = 1234567 Double result = 0 ;
>>
>> Result = dnum1-dnum2 ;
>>
>> The result is 0.10000000009
>>
>> Anybody know how to solve this question?
> 
> 0.1 can't be represented exactly in binary, so there's an inevitable roundoff error. Sometimes these errors are hidden by the implementation of the binary to decimal conversion used in printing out the answer, but sometimes not. I don't know if you're using printf in C, or << in C++, but you might try requesting less precision in the output, to see if that rounds the result back to 0.1.
> 
> --
> 
> Ciao,               Paul D. DeRocco
> Paul                mailto:pderocco@ix.netcom.com
> 
> 
> --
> Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
> and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss
> 
> 


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss