Re: This calculation is just wrong / computer can't count!
"David Wilkinson" <no-reply@effisols.com> wrote in message
news:uObZQmmCIHA.2324@TK2MSFTNGP03.phx.gbl...
GT wrote:
... which the rest of the world calls mathematics. You might have been
brainwashed by some ancient Fortran 'lessons', but everyone else in the
world knows that if you express 1/3 1 as a decimal you get a recursive
number and the final digit is always a 3 - NEVER a 7. If your brainwashed
view of the world allows you to accept a 7 as the last digit, then your
fundamental grasp of mathematics has been tainted and you can never rely
on any of your calculations ever again! A 7 as the last digit is WRONG.
Plain and simple - it is WRONG.
GT:
IEEE floating point arithmetic is not wrong; it is just different from the
way you would like it to be.
Computers work in binary arithmetic.
The analogy of what is going on in this thread is: floating point captures
the essence of ideal math about the same as a color picture captures the
moment in real life. Obviously it is not the real thing, but for most uses
it's close enough. GT's complaint is that in this otherwise perfectly fine
color picture, one pixel that should be entirely black is instead entirely
white. He says that is wrong, and I tend to agree with him. The rest say
it is an artifact of the color picture system and should be tolerated.
Either way, it makes no difference. Like it or not, the color picture is
the best we've got and if a little tweaking can fix it's flaw (such as just
printing out less digits that we know are right and ignoring the rest), then
that is what we should do and say no more about it.
-- David