Skip to content

Latest commit

 

History

History
331 lines (242 loc) · 10.5 KB

chapter2.md

File metadata and controls

331 lines (242 loc) · 10.5 KB

Getting started: simple code

Simple code in FORTRAN

FORTRAN is a bit different from, say, Matalab. It is a low level language. Concretely this means that the steps needed to run some code are distinct:
1. writing the code
2. compiling the code
3. executing the code

How to start

1. Writing the code

To write the code use any text editor. It might be more convenient to use an editor specialized in FORTRAN such as Atom (from Github) or SublimeText. In these editors the syntax is colored according to FORTRAN conventions.

To start writing code, create a new a file and call itmyprogram.f90. That's your main file. It has the following structure.

program main
  the code here
end programtr

You do not need to call the programmain. You can also call it myprogram or any other name, as long as the file starts withprogram and ends withend program.

The the code here part is some FORTRAN code that we will cover later. For now, note that the code must be closed, meaning that the whole structure must be consistent. In this case it means that we start with a keyword program and ends with the keywords end program. This is different from Matlab where your code might only contains something like

a = 3
b = 2
c = a + b

2. Compiling the code

To compile the code one needs a compiler (see Chapter 1). On Unix machine, open the terminal, navigate to the directory where your fortran file is located (using cdcommand, see Appendix - Bash) and type

gfortran myprogram.f90

This will produce a new file (an executable). By default, the new file is called a. You can change this to give the file a more convenient name, for example myprog_exe. To do so, use the option -o((small letter o, for output):

gfortran -o myprogr_exe myprogram.f90

This tells the compiler to name the output file myprog_exe. It uses the flag (or option) -o followed by the argument to this flag.

To compile a program with modules or a library, see next chapters.

3. Executing the code

It is only when executing the file that the code will be executed and the computations run. To execute the file, type in the terminal

./a

Or, if you used the -o flag with a different output name:

./myprog_exe

Declaring variables

You need to declare all the variables you will use. Here is how

program myprogram
  implicit none
  real :: a, b, c
  the code that uses a b and c here
end program

The key words implicit none means that you want that all the variables used in the code have to be first declared. Hence, if in the the code here part you use other letters, such as d, the code will return an error when you try to compile it. It is possible no to use implicit none which allows you to use variables not declared. However, you should never do this (mostly because each letter of the alphabet has a default type if not declared). The double colon after real is optional. Note that here we declared the variables asreal. There are other types, and the most usual types are

real
integer
character
logical

The type character is for strings. To declare a string you need to use single or double quotes and write

string :: stringvar="text in the variable"

You can specify the number of bytes to be used for your variables (hence the precision) by using the kindspecifier.

integer(kind=2) :: shortinteger
integer(kind=4) :: longinteger
integer(kind=16) :: verylonginteger

Types and arrays, parameters

To declare an array (vector) a and a matrix b use

real :: a(5)
real :: b(5,4)

or, equivalently

real, dimension(5) :: a
real, dimension(5,4) :: b

The first way is convenient to declare arrays of different dimensions on the same line

real :: a(5), b(5,4)

You can also use the keyword parameter. This means that the variable cannot change value. You must therefore initialize it at the same time you declare it.

real, parameter :: d = 5

Arithmetic operations

The usual operations are: addition +, subtracction -, division / , multiplication * , exponention ** .

A very common error is to use integer instead of real when doing arithmetic operations. Operations on integers always return an integer. For example 9/4 will return 2.

Functions and Subroutines

In FORTRAN you can use two types of blocs of code which you provide with input: the function and the subroutine. They both work the same way but with slight differences.

Functions

A function is defined as follows:

function myfunction(a,b,c)
  implicit none
  real :: a,b,c, myfunction
  the code here
end function

As you can see, one needs to declare all the variables that are passed as arguments (here a, b and c) as well as the output variable which must have the same name as the function. For example, the function could be

function myfunction(a,b,c)
  implicit none
  real :: a,b,c, myfunction
  myfunction = a + b + c
end function

One can actually choose a different name for the output (rather than the same name as the function) but one must then use the following syntax

function myfunction(a,b,c) res(var)
   implicit none
   real:: a,b,c
   real:: var
   var = a + b + c
end function

To call this function into the main program you would write altogether

program myprogram
  implicit none
  real :: x1,x2,x3, easy_sum
  x1 = 1
  x2 = 2
  x3 = 3
  easy_sum = myfunction(x1,x2,x3)

  contains
  function myfunction(a,b,c)
    implicit none
    real :: a,b,c, myfunction
    myfunction = a + b + c
  end function
end program

keywordThe key word contains tells the compiler that the code following this key word contains the definitions of functions and subroutines.

In the definition of the function, the variables passed as argument are local: they can have the same name as variables in the main part (above contains), this will not affect them. One can however use variables defined above containsinside the function, such as:

program myprogram
  implicit none
  real :: x1,x2,x3, easy_sum, big_number
  big_number = 1000
  x1 = 1
  x2 = 2
  x3 = 3
  easy_sum = myfunction(x1,x2,x3)

  contains
  function myfunction(a,b,c)
    implicit none
    real :: a,b,c, myfunction
    myfunction = a + b + c + bignumber
  end function
end program

In this example, easy_sum will be equal to 1000 + x1 + x2 + x3,, i.e. 1+2+3+1000 = 1006.

Subroutines

A subroutine is similar to a function but it does not necessarily return a variable. Instead, it executes the code inside it. Declaring a subroutine is done as follows

subroutine mysubroutine(a,b,c)
  implicit none
  real :: a, b, cr
  a = 0
  b = -1000
  c = a + b
end subroutine

This subroutine will change the value of the three variables. You use a subroutine into your program with the following code:

program myprogram
  implicit none
  real :: x1,x2,x3
  call mysubroutine(x1,x2,x3)

  contains
  subroutine mysubroutine(a,b,c)
    implicit none
    real :: a, b, c
    a = 0
    b = -1000
    c = a + b
  end subroutine
end program

More on functions and subroutines

In both functions and subroutines, one can specify if the arguments of the function/subroutine are to be altered by using the keywords intent(in), intent(out) and intent(inout). intent(in) means that the variable should not be altered inside the function (input only). intent(out) means that whatever the value given to the variable before it enters the function as an argument, this value is ignored. With intent(inout) the variable can be altered and the value it has before entering the function is still assigned to it at the beginning of the function. For example

function myfunction(a,b,c)
  implicit none
  real, intent(in) :: b, c
  real :: myfunction
  real, intent(inout) :: a
  a = a + b + c
  myfunc = a
end function

Note that you should not have an intent(out) before the return of the function (here myfunction).

A function can also be a pure function. This implies that the function will only change the behavior of the output (i.e. the variable myfunction above). It is a keyword used to make sure that nothing else is altered.

pure function myfunction(a,b,c)
  implicit none
  real, intent(in) :: b, c
  real :: myfunction
  real, intent(inout) :: a
  a = a + b + c
  myfunc = a + b + c
end function

This will not work (to check) because a is an argument of a pure function, but the function tries to modify it (the inoutclassification is enough to provoke an error as a pure function cannot modify an input argument). To have something working one needs to change the intent(inout) :: a into intent(in) :: a as well as delete a = a + b + c (or create an intermediate variable intvar = a + b + c , and not forgetting to declare it with: real :: intvar).

Save keyword, initialization-declaration

The keyword save is used to make a variable global (see next chapter). More specifically, save is a specification statement which can be used to ensure that variables used within a procedure (local variable) preserve their values between successive calls to the procedure (often the procedure is in a module).
Another way to make a variable global is to assign it a value at the same time as the declaration.

real, save :: a
real :: b = 5

That is why it is dangerous to assign a value at the same time one declares a variable: at each iterative call, the starting value of the variable is the last one assigned to it, not the one assigned in the declaration (here b = 5).

Overall, it is preferable to use modules instead of save (see next chapter).

Comments

To write a comment in FORTRAN, use ! at the beginning of the comment.

! this is a line with only comments
real, save :: a ! this is a comment
real :: b

Break lines

To break a line in FORTRAN, use & at the end of the line you want to break, and start the next line right below.

program main
implicit none
real :: a
real :: b

a = 5.0
b = 2.3 + 3.0 &
    + a
end program

It is useful to break lines because each line has a maximum length (which depends on the compiler). It is wise to limit the length to no more than 300 characters.

STOP

The STOP statement terminates the program. One can pass a message as an argument to STOP (either an integer or a string). The message will be displayed when STOP is reached and the program terminates.

stop "Error"
stop 1