Modern variable precision and integer multipliers

[avikarto]

I learned Fortran from ancient people who disregard modern standards. As such, my code consists mainly of that older syntax. I am wondering how well this meshes with current Fortran environments, specifically regarding double precision calculations. This comes up mainly because I have been told that real*8 and similar declarations should be abandoned in favor of KIND.

For example, in the following code section

real*8 x,y,z
x=12.3456
y=2*x
z=2.d0*x

will y maintain full double precision, or will a statement like z be required?

Similarly, if the fractional nature of the initial variable is defined only implicitly,

real*8 pi,y,z
pi=4*atan(1)
y=2*pi
z=2.d0*pi

does this change the nature of y or z?

Or, for that matter, does pi maintain an appropriate value here instead of being declared as pi=4.d0*atan(1.d0)?

Thanks.

 

[jim mcnamara]

If I understand what you are asking: can I use what seems to be an integer value and apply some arithmetic operator to it on a double precision variable without losing precision? In general, yes. Double precision on most systems is 15 digits of precision. That means the two operands and the result all have to fall within those bounds, or things may go wonky, precision-wise.

Promotion: from http://www.cs.uwm.edu/~cs151/Bacon/Lecture/HTML/ch06s09.html FORTRAN class notes.

Implicit Data Conversions
Promotions occur when a two different data types are operands to a mathematical operator. In all cases, the value of the lower ranking type is converted to the higher ranking type.

1. double complex [ complex(8) ]
2. complex [ complex(4) ]
3. double precision [ real(8) ]
4. real [ real(4) ]
5. integer(8)
6. integer [ integer(4) ]
7. integer(2)

Promotion is defined as taking the variable with least precision and boosting it up to the precision of the other variable.

 

[avikarto]

Specifically, is real*8 still fully compatible this way with newer version compilers? Also, do I lose anything with my simpler definition of pi?

 

[SteamKing]

Specifically, is real*8 still fully compatible this way with newer version compilers? Also, do I lose anything with my simpler definition of pi?

You can always run a test. Calculate π using REAL*8 and calculate it using the KIND parameter. Compare the two results.