float f1 = 0.0f, f2;
do
{
f2 = f1;
f1 = f1 + 1;
}
while (f2 != f1);
what will the o/p of this question & why its more important why..??
O/P 1.677722e+7
About Floats
Computers, CPUs, and electronic devices store numbers in binary format. Floating
point numbers, or floats, are the most commonly used system for representing real
numbers in computers. In industrial automation applications, analog values read
from an I/O unit are an example of floats. Floats represent real numbers in scientific
notation: as a base number and an exponent.
The IEEE Standard for Binary Floating-Point Arithmetic (IEEE 754) is the most widely
used standard for floating-point computation. It defines how to store real numbers
in binary format and how to convert between binary and float notations.
Opto 22’s SNAP PAC System uses IEEE single-precision floats, which have 32 binary
digits (bits). The IEEE 754 32-bit float format is as follows:
Float calculation: (-1)Sign x [1 + Significand/223] x 2 (Exponent-127)
While this is an excellent standard for the purpose, it has limitations that could cause
issues if you’re not aware of them. Squeezing infinitely many real numbers into a
finite number of bits requires an approximate representation. Most floats cannot be
exactly represented using this fixed number of bits in a 32-bit IEEE float. Because of
this, rounding error is inherent in floating-point computation.
In PAC Control (and in PAC Manager and the OptoMMP protocol), a float is a 32-bit
IEEE single-precision number ranging from ±3.402824 x 10-38 to ±3.402824 x
10+38. These single-precision floats give rounding errors of less than one part per
million (1 PPM). You can determine the limit of the rounding error for a particular
float value by dividing the value by 1,000,000.
This format guarantees about six and a half significant digits. Therefore,
mathematical actions involving floats with seven or more significant digits may incur
errors after the sixth significant digit. For example, if the integer 555444333 is
converted to a float, the conversion yields 5.554444e+8 (note the error in the 7th
digit). Also, converting 5.554444e+8 back to an integer yields 555444352 (note the
error starting in the 7th digit).
Float Issues and Examples
Accumulation of Relatively Small Floating-point Values
When adding float values, the relative size of the two values is important. For
example, if you add 1.0 to a float variable repeatedly, the value of the float variable
will correctly increase in increments of 1.0 until it reaches 1.677722e+7 (16,777,220).
1 bit 8 bits 23 bits
x xxxxxxxx xxxxxxxxxxxxxxxxxxxxxxx
Sign Exponent Significand
Using Floats Technical Note
PAGE
2 TECHNICAL NOTE • Form 1755-080130
Then the value will no longer change, because 1.0 is too small relative to 1.677722e+7 to
make a difference in the significant digits. The same thing will occur if you add 0.0001 to
2,048.0, or add 1,000.0 to 1.717987e+10. The key is the relative size of the numbers.
Here’s another way to think of it. Suppose your bank could only keep track of seven digits. If
you were fortunate enough to have one million dollars ($1,000,000) in your account and
tried to add 10 cents ($0.10) to it, you would not be able to, because the 10 cents is not big
enough relative to the total to be significant. Since the bank has only seven digits to keep
track of your money (in this example), one digit has to fall off the end: either the 10 cents
falls off the right side or the million digit falls off the left side. Which would you rather see in
your bank account?
Note that moving the point indicator doesn’t help, because the exponent is separate. If the
seven digits for the account represent millions of dollars (1.000000) rather than dollars
(1,000,000), the 10 cents would be 0.0000001—still too small to be represented by the
seven digits:
The key is that it is not the size of the numbers that matter, but rather their relative size.
So if you are accumulating relatively small values in a float variable over a long period of
time, at some point, the float value will stop increasing even though you continue to try to
add to it.
Comparing Floating-point Values for Equality
Due to rounding errors and the way floating-point calculations are performed, comparing
two floats for equality can yield inaccurate results. The precision of comparisons depends
on the relative size of the float values as compared to the difference between them.
For example, if 2,097,151.0 is compared for equality with 2,097,152.0, the result will
indicate that the two floats are equal, even though it’s obvious they are not. The reason is
that the difference between the two values is 1.0, and 1.0 compared to one of the
compared values (2,097,151.0) is too small; it is less than one part per million.
In this case, 2,097,152.0 divided by 1,000,000 is 2.1. If the difference between the two
values is at least 2.1, then the equality comparison is guaranteed to be correct. So if
2,097,152.0 and 2,097,149.0 were compared for equality, the result will indicate they are
not equal, because the difference (3.0) is greater than one part per million (2.1). Any time
Seven digits available: 1 2 3 4 5 6 7
Amount in account: 1 0 0 0 0 0 0
Add 10 cents (0.10)? (digit falls off on right) 1 0 0 0 0 0 0
Add 10 cents (0.10)? (digit falls off on left) 0 0 0 0 0 0. 1 Oops!
Seven digits available: 1 2 3 4 5 6 7
Amount in account 1. 0 0 0 0 0 0
Add 10 cents (0.0000001)? (digit falls off on right) 1. 0 0 0 0 0 0
Add 10 cents (0.0000001)? (digit falls off on left) .0 0 0 0 0 0 1 Oops again!
Using Floats Technical Note
PAGE
TECHNICAL NOTE • Form 1755-080130 3
the difference is at least one part per million, the result is guaranteed to be accurate. If the
difference is less than 1 PPM, it may or may not be accurate.
One method that programmers use to work around this issue is to subtract one float from
the other and then compare the absolute value of the result to a limit.
For example:
Float_Diff = Float1 - Float2;
If (AbsoluteValue(Float_Diff) < 1.0 ) then
SetVariableTrue(EqualityFlag);
Else
SetVariableFalse(EqualityFlag);
Endif
Source http://www.opto22.com/documents/1755_Using_Floats_Technical_Note.pdf
No comments :
Post a Comment