prev

next

out of 11

View

3.848Download

92

Embed Size (px)

DESCRIPTION

Fortran quick reference

Appendix C

A Fortran Primer: (and cheat sheet)This section will provide a basic intro to most of the commonly occuring features of Fortran that you will need for the course. This list is by no means exhaustive, but it should be enough to get you where you need to go. For more information, We have extensive fortran manuals scattered about the observatory. By the end of this section you should understand the basic data types, the structure of arrays, standard arithmetic operations, loops and conditional statements, subroutines and functions, and basic IO. This section will also briey discuss how to build programs using make and how to debug them using dbxtool/debugger.

C.1 Basic Fortran Concepts and CommandsData types for the most part there will be only three basic types of data you will have to deal with, integers, oating point numbers and characters. In fortran these data types are declared asinteger exact whole numbers (-3, 0, 5 234), usually stored in 4 bytes real inexact representation of decimal numbers (3.1415, 27.2, 1.e23). Usu-

ally stored as 4 bytes, good for about 6-9 signicant digits, ranges from about 1038 1038double precision same as real but much larger range and precision.

Usually stored as 8 bytes, good for about 15-17 signicant digits, ranges from about 10308 10308 . You can get more precision than this but you should have a good reason.character, character*n either a single character or a string of characters of length n.

Constant and Variable names A constant or variable can have any name with up to 32 characters (To play it safe though, Standard Fortran allows only 6 characters). The name must start with a letter, but can have most of the 19

20 alphanumeric characters in it. Fortran also has the annoying feature of implicit typing so that if undeclared, any variable name that starts with the letters i through n are assumed to be integers, everything else is assumed real. Rigorous programming practice suggests that you start every program with implicit none and declare every variable. Its annoying but it keeps you honest. I will try to do this in my programs. A sample program declaration might look like implicit none integer npnts, n real c, dcdt real car(1000),tar(1000) character*40 fileout arrays Perhaps the most important feature for modeling is arrays. Arrays are simply ordered sets of numbers that can be addressed by index. Every array has a name (e.g. car or tar above) and a length (here both car and tar are arrays of real numbers of length 1000). Arrays can be of integers, reals, double precision or even characters (fileout is actually an array of 40 characters). Each member of an array can be addressed as by specifying its index in the array (car(10), tar(n) etc.). Arrays can also have up to 7 dimensions. A two dimensional array a(10,10) can be thought of as 10 1-dimensional arrays of 10 numbers each (a total of 100 elements). Most importantly in fortran, the leading index increases fastest i.e. for a(i,j), the i=1 and i=2 are next to each other in memory. Simple operations integer and oating point numbers and arrays of them can all be operated on by standard mathematical options. The most commonly used arithmetic are= assignment x=y ** exponential a=x**y /, * divide, multiply a=x/y, or a=x*y +, - add subtract a=x+y, or a=x-y

The order of evaluation is standard algebraic and parentheses can be used to group operands. In addition to the simple operations, Fortran also has some built in intrinsic functions. The most commonly occuring are trigonometric functions sin(x), cos(x), tan(x), asin(x), acos(x), atan(x), atan2(x) (inverse trig functions) sinh(x), cosh(x), tanh(x) etc. exponential functions exp(x), log(x) (natural log), log10(x) (log base ten), sqrt(x) square-root. conversion functions these functions will convert one data type to another, e.g. int(x) returns the integer value of x, real(?) converts anything

Fortran

21 to a real, dble(?) converts anything to a double (also operations for complex numbers)

misc. functions see table 6.1 Program ow control Any program will just execute sequentially one line at a time unless you tell it to do something else. Usually there are only one of two things you want it to do, loop and branch. The control commands for these aredo loops Do loops simply loop over a piece of internal code for a xed

number of loops and increment a loop counter as they go. do loops come in the standard line number avour do 10 i=1,n,2 j=i+1 other stuff to do continue

10

(note the spacing of the line number and starting commands at least 6 characters over is important (stupid holdover from punch cards). Or in the non-standard but more pleasant looking form do i=1,n,2 j=i+1 other stuff to do enddo Do loops can be nested a fair number of times (but dont overdo it) conditional statements Being able to make simple decisions is what separates the computers from the calculators. In fortran separate parts of the code can be executed depending on some condition being met. The statements that control this conditional execution areif a single line of the form

if (iam.eq.crazy) x=5. will assign a 5. to the variable x if the expression in parentheses is true, otherwise it will simply skip the statement. block if statements More useful blocks of execution can be delimited by block ifs of the form if ( moon.eq.full ) then call howl(ginsberg) x=exp(ginsberg) endif the block if statements can also be extended to have a number of condition statements. If there are only two, it looks like if ( moon.eq.full ) then call howl(ginsberg) x=exp(ginsberg)

22

6Table 6-1 Intrinsic Function Absolute Value Denition Arithmetic Functions (continued) No. of Args 1 Generic Name ABS Specic Name IABS ABS DABS CABS CQABSo QABS o ZABS o CDABSo MOD AMOD DMOD QMOD o ISIGN SIGN DSIGN QSIGN o IDIM DIM DDIM QDIM o DPROD QPROD o MAX MAX0 AMAX1 DMAX1 QMAX1 o AMAX0 MAX1 Choosing Smallest Value min(a1, a2, ) 2 MIN MIN0 AMIN1 DMIN1 QMIN1 o AMIN0 MIN1 Type of Argument Integer Real Double Complex Complex*32 Real*16 Complex*16 Complex*16 Integer Real Double Real*16 Integer Real Double Real*16 Integer Real Double Real*16 Real Double Integer Real Double Real*16 Integer Real Integer Real Double Real*16 Integer Real Function Integer Real Double Real Real*16 Real*16 Double Double Integer Real Double Real*16 Integer Real Double Real*16 Integer Real Double Real*16 Double Real*16 Integer Real Double Real*16 Real Integer Integer Real Double Real*16 Real Integer

|a| Read Note 6 (ar2 + ai2)**(1/2)

Remainder

a1-int(a1/a2)*a2 Read Note 1

2

MOD

Transfer of Sign

|a1| if a2 0 -|a1| if a2 < 0

2

SIGN

Positive Difference

a1-a2 if a1 > a2 0 if a1 a2

2

DIM

Double and Quad Products Choosing Largest Value

a1 * a2

2

max(a1, a2, )

2

280

FORTRAN 2.0.1 Reference ManualOctober 1992

Fortran

23 else call wait(1,month) endif or if there are several conditions if ( expression ) then statements galore.... elseif (expression2) more statements elseif (expression3) are you tired of this yet else default statements endif Note: a block if of this type will execute the rst true expression and the jump to the end if (even if several conditions are true). relational operators The statement moon.eq.full is a conditional statement that evaluates to true if the variable moon is logically equivalent to the variable full (dont use = for .eq. very different animals). The operator .eq. is a relational operator that tests for equivalence. Other relational operators are .lt. (lessthan),.le. (less-than-or-equal-to) ,.ge. (greater-than-or-equalto),.gt] (greater-than). Individual conditional statements can also be linked together using the operators .and., .or., .not. (and a few less useful things like exclusive or .xor.. Examples include& if ((moon.eq.full).and.(i.eq.werewolf)) call runlikehell() if ( (x.eq.0.5).or.(i.le.2) ) then x=x*i endif

Subroutines and Functions Subroutines and functions are separate pieces of code that are passed arguments, do something and return. Subroutines and functions are similar except that functions return a value and are used like the intrinsic functions i.e. integer myfunction, k,j external myfunction real x ... k=myfunction(x,j) and are declared like integer function myfunction(r,i) real r integer i myfunction=mod(int(r),i))

24 return end Note the variables r and i are dummy variables that hold the place of x and j in the main routine. Subroutines also have dummy variables but are called and dont return a value (although it will often change the values of what it is passed). Example subroutine call would be integer k,j real x(100),a .... j=10 a=2.5 call arrmult(x,j,a) and the subroutine itself would look like subroutine arrmult(ar,n,con) integer n real ar(n),con integer i do i=1,n ar(i)=ar(i)*con enddo return end Note that arrays are passed to subroutines and functions by name. Most importantly, the dimension of the array within the subroutine can be passed. In the above routine, only the rst 10 elements of x are multiplied by 2.5. If we wanted to multiply the elements from 6 to 16 we could also do call arrmult(x(6),11,a) Note that there are actually 11 elements between indexes 6 and 16. In addition, arrays that are declared as 1-D arrays in the main program can be operated on as variable length n-d arrays in a subroutine (and vice-versa). E.g. integer i,j real x(1000),a ... i=10 j=10 a=2.5 call arrmultij(x,i,j,a)

Fortran... subroutine arrmultij(ar,m,n,con) integer m,n real ar(m,n),con integer i,j do j=1,n do i=1,m ar(i,j)=ar(i,j)*j*con enddo enddo return end

25

We will make extensive use of this feature for efcient programming in ndimensions. This is the one thing you cannot do w