Intel® Software Network

Hello

I´m a doubt: I need use double float numbers but I wish that these numbers have a fixed accuracy of only 3 or 4 digits. For example:

REAL*8      TOUT, BUF(5000,17)

TOUT = (BUF(i,17) - BUF(i-1,17))*60.0D0

resulting in (for example):

TOUT = 10.2839999999995

But I need that this number is rounded to 10.2840000000000.

Does anybody know how I can resolve this problem?

Thanks

IlexSP

From what you posted, I don't know whether you have unrealistic expectations, or have overlooked the Fortran FORMAT specifications, or ..... ?
When you take the difference of two nearly equal binary values, with bits just sufficient to represent 15 or more decimal digits, the results you got are to be expected.