• 1. Don't use floating point data types for precision on any scale.
  2. Don't use floating point data types for currency!
  3. [Insert any other rules that you can probably derive from 1. or 2. here]

Quote:Real-world numerical catastrophes.

• Ariane 5 rocket. Ariane 5 rocket exploded 40 seconds after being launched by European Space Agency. Maiden voyage after a decade and 7 billion dollars of research and development. Sensor reported acceleration that so was large that it caused an overflow in the part of the program responsible for recalibrating inertial guidance. 64-bit floating point number was converted to a 16-bit signed integer, but the number was larger than 32,767 and the conversion failed. Unanticipated overflow was caught by a general systems diagnostic and dumped debugging data into an area of memory used for guiding the rocket's motors. Control was switched to a backup computer, but this had the same data. This resulted in a drastic attempt to correct the nonexistent problem, which separated the motors from their mountings, leading to the end of Ariane 5.
  
• Patriot missile accident. On February 25, 1991 an American Patriot missile failed to track and destroy an Iraqi Scud missile. Instead it hit an Army barracks, killing 26 people. The cause was later determined to be an inaccurate calculate caused by measuring time in tenth of a second. Couldn't represent 1/10 exactly since used 24 bit floating point. Software to fix problem arrived in Dhahran on February 26. Here is more information.
  
• Intel FDIV Bug Error in Pentium hardwire floating point divide circuit. Discovered by Intel in July 1994, rediscovered and publicized by math professor in September 1994. Intel recall in December 1994 cost $300 million. Another floating point bug discovered in 1997.
  
• Sinking of Sleipner oil rig. Sleipner A $700 million platform for producing oil and gas sprang a leak and sank in North Sea in August, 1991. Error in inaccurate finite element approximation underestimate shear stress by 47% Reference.
  
• Vancouver stock exchange. Vancouver stock exchange index was undervalued by over 50% after 22 months of accumulated roundoff error. The obvious algorithm is to add up all the stock prices after Instead a "clever" analyst decided it would be more efficient to recompute the index by adding the net change of a stock after each trade. This computation was done using four decimal places and truncating (not rounding) the result to three.

Quote:I also know for a few days (you know I am learning C# atm) that C# has a decimal data type which is exactly for the cases where you need precision. That's quite nice.

(05-22-2013, 05:58 AM)Psycho_Coder Wrote: Very informative post sir. I have a doubt, see the following :-

Code:

double d = 0.0;
for (int i = 0 ;; i++)
{
    d += 0.1;
    printf("%f\n", d);
    getchar();
}

In printf the format specifier you used is %f , shouldn't this be %lf which is in case of C (Since this is a C code). In the second code too you have used %.16f instead of %.16lf, since you are using double datatype so %lf should be used.

Quote:ISO/IEC 9899, second edition (the C99 Standard), sections 7.19.6.1 and 7.19.6.2.

(05-22-2013, 06:14 AM)ArkPhaze Wrote:

(05-22-2013, 05:58 AM)Psycho_Coder Wrote: Very informative post sir. I have a doubt, see the following :-

Code:

double d = 0.0;
for (int i = 0 ;; i++)
{
    d += 0.1;
    printf("%f\n", d);
    getchar();
}

In printf the format specifier you used is %f , shouldn't this be %lf which is in case of C (Since this is a C code). In the second code too you have used %.16f instead of %.16lf, since you are using double datatype so %lf should be used.

I read that %lf is for long double, you could be right though. There seems to be confusion depending on what you read out there between C89 and C99..

Quote:Specifies that a following d, i, o, u, x,or X conversion specifier applies to a long int or unsigned long int argument; that a following n conversion specifier applies to a pointer to a long int argument; that a following c conversion specifier applies to a wint_t argument; that a following s conversion specifier applies to a pointer to a wchar_t argument; or has no effect on a following a, A, e, E, f, F, g,or G conversion specifier.

Quote:If a length modifier appears with any conversion specifier other than as specified above, the behavior is undefined

Quote:%lf conversion specifier allowed in printf

(05-22-2013, 07:00 AM)Psycho_Coder Wrote: In most of the sites I have read that %lf if specifically to be used for double

Code:

double radius;
    int length;  

    scanf("%lf%d", &radius, &length);
    scanf("%lf%d", &radius, &length);
    printf("%lf  %d", radius, length);

Quote:When we have to use scanf then for double we have to give %f and according to C99 %Lf is used for long double. Here :- http://www.cplusplus.com/reference/cstdio/printf/