Another kind of rounding can occur when a floating number is
converted between different formats, for example with
processors that internally use a higher precision.
Consider the following two almost identical programs.
-
small powerOf2 exp
|
Frank Christoph writes:
It seems to me there is a tension between using show as a way of doing quick
and dirty pretty-printing, and as a way of getting a portable representation of
data.
This is a "bug" in hugs.
To illustrate the problem, the next floating point number after 5.0 is
Hi John,
| This is a "bug" in hugs.
|
| To illustrate the problem, the next floating point number after
| 5.0 is 5.0047,
| which hugs also prints as 5.0. One might argue that to display it
| as 5.005 would
| be misleading, since this number is the closest representable to
|
On 09-Feb-2000, Mark P Jones [EMAIL PROTECTED] wrote:
The Numeric.showFloat function is there for the more expert programmers
who care about the last few bits after the floating point. That's the
function that a Haskell programmer should use if they need this kind of
functionality. I can
(LONG and about floating point, so I suspect many Haskellers are not
going to be interested in this message . . .)
Julian Assange wrote:
The precission and or rounding used by hugs/ghc seems strange, to wit:
Prelude sin(pi)
-8.74228e-08
Prelude pi
3.14159
sin(3.14159265358979323846)
Jerzy Karczmarczuk wrote:
but one should always try to perform all
the needed reductions before the floating precision gets out of
control. There are nice recurrences which simplify the computations
with trigonometric functions. But between 0 and Pi/4 it should be
well done.
No one should
Prelude 1.0 - 0.8 - 0.2
-1.49012e-08
Since the Paleozoic Era Hugs is distributed with HAS_DOUBLE_PRECISION
desactivated (can some gurus explain why?...), and the first thing
I do with, is its recompilation. Such nasties as above become then less
dangerous.
"less dangerous" ... perhaps I
Prelude 1.0 - 0.8 - 0.2
-1.49012e-08
Taking Ian Stark's example a little bit further,
Main let x=6.0e-8 in (1.0,1.0+x,1.0==1.0+x)
(1.0,1.0,False)
is a useful reminder that show isn't one-to-one.
John Hughes
George Russell wrote:
Julian Assange wrote:
Once you are within a few UDP, the underlaying grainyness of the
representation is going to get you, so that smoothe, monotonic line
segment you have below, will look like an appalling zigzag at
best. This is my point. Near the limits of
-Original Message-
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]
...
George rightly points out how tricky trig functions are. My own
favourite curious operation is subtraction:
Prelude 1.0 - 0.8 - 0.2
-1.49012e-08
Yes, floating point arithmetic is fascinating, isn't it!
(LONG and about floating point, so I suspect many Haskellers are not
going to be interested in this message . . .)
This one too.
-Original Message-
From: George Russell [mailto:[EMAIL PROTECTED]]
...
Let's summarise what
IEEE 754 gives you.
IEEE 754 only
Hello ASCII artists,
Maybe the rounding issue, which doesn't seem to have much to
do with Haskell anymore at this point, can be taken offline?
Regards,
Koen.
--
Koen Claessen http://www.cs.chalmers.se/~koen
phone:+46-31-772 5424 e-mail:[EMAIL PROTECTED]
I had at first thought that this might be a discussion of the US style
definition of round in the standard prelude that always bites me...
... Then that it might be the bug in floor within the standard issue
of GMP.
Instead its a discussion of FP problems.
(1) My friend Jean-Michel Muller at
Since the Paleozoic Era Hugs is distributed with HAS_DOUBLE_PRECISION
desactivated (can some gurus explain why?...), and the first thing
I do with, is its recompilation. Such nasties as above become
then less dangerous.
The newer STG Hugs which we are developing has "real" Doubles as
a straight line, EG for example
XX
X
XX
X
X
XX
X
XX
(blah blah)
the point being that this is a straight line with slope a
bit less than
1/3.
Code to test this attached. Try (in Hugs) plotSin pi 100.
You should a
George Russell [EMAIL PROTECTED] writes:
(LONG and about floating point, so I suspect many Haskellers are not
going to be interested in this message . . .)
Excellent, thanks george.
Now I think the original suggestor wanted the library to spot that the answer
is "very close to" 0 and
John Hughes wrote:
Taking Ian Stark's example a little bit further,
Main let x=6.0e-8 in (1.0,1.0+x,1.0==1.0+x)
(1.0,1.0,False)
is a useful reminder that show isn't one-to-one.
Dunno much about FP arithmetic, but maybe there should be a primitive
showFloatExact which yields the
Julian Assange wrote:
The precission and or rounding used by hugs/ghc seems strange, to wit:
Prelude sin(pi)
-8.74228e-08
Prelude pi
3.14159
sin(3.14159265358979323846)
-8.74228e-08
ghc:
module Main where
main = do
print pi
print (sin pi)
./a.out
Lennart Augustsson [EMAIL PROTECTED] writes:
Haskell performs no worse or better than C. Your comment about how math
libraries round might be accurate for some languages, but for C it's pure nonsense.
It seems that you are right in this instance. However I recall seeing
comments about error
19 matches
Mail list logo
