double - Dealing with small numbers and accuracy -
i have program deal lot of small numbers (towards lower end of double limits).
during execution of application, of these numbers progressively smaller meaning "estimation" less accurate.
my solution @ moment scaling them before calculations , scaling them down again?
...but it's got me thinking, gaining more "accuracy" doing this?
thoughts?
are numbers really in region between 10^-308 (smallest normalized double) , 10^-324 (smallest representable double, denormalized i.e. losing precision)? if so, scaling them indeed gain accuracy working around limits of exponent range of double type.
i have wonder though: kind of application deals numbers extremely small? know of no physical discipline needs that.
Comments
Post a Comment