NavList:

   

Reply

Alexandre E Eremenko wrote:
>
> The reason of these errors is probably the rounding
> subroutine. To get the last digit really true,
> yoiu have to do all computations with one more digit,
> and then apply an exact rounding algorithm, while
> in most cases they just cut off the last digit
> instead. Because nobody really cares about the exactness
> of the
> decimal minute digit.

Generally, programming languages don't provide any way to carry one
more digit in computations. You can control precision in a coarse
manner by the choice of data type, e.g., single precision or double
precision floating point. But the machine stores and computes that type
to full accuracy, whether you need it or not.

On output you can specify precision. In all the C family languages
I've used, floating point output is rounded. For example, if you print
pi to three decimal places, the result is 3.142, not 3.141. This control
over precision only affects the output formatting, though.

To reduce the precision of the variable itself, the Microsoft .NET
Framework (accessible with Visual Basic, Visual C#, et al) has routines
to round a value to a specified number of decimal places:
http://msdn2.microsoft.com/en-us/library/system.math.round.aspx

But since that's extra work for the machine, normally you'd round only
the human-readable output, and maintain variables to full precision.

Astronomical computing has changed a lot since the days when "computers"
were human beings running desk calculators. Nowadays even a cheap
machine has a floating point processor in the CPU. Many of the old labor
saving methods disappeared because on today's hardware there's no
penalty for carrying extra significant digits.

--
I block messages that contain attachments or HTML.

--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to NavList@fer3.com
To , send email to NavList-@fer3.com
-~----------~----~----~----~------~----~------~--~---

   

Reply