Debugging Strange Fortran Code: A Simple Delta Function

[phynewb]

Hi guys,
I just wrote a simple code for delta function and print it out. Here it is.

      program main
        implicit none
        integer i
        real*8 del
        real*8,parameter:: ep=1D-2
        real*8,parameter:: pi=3.1415926
        do i=-10,10,1
        write(*,*) i,del(real(i))  ! use function
        enddo
        do i=-10,10,1
        write(*,*) i,(1.0/pi*ep/((real(i))**2+ep**2)) ! no use function
        enddo
        end

      function del(x)
      implicit none
      real*8 del
      real*8 x
      real*8,parameter::ep=1D-2
      real*8,parameter:: pi=3.1415926

      del=(1.0/pi*ep/(x**2+ep**2))

      return
      end

It turns out to be quite wrong by using function del. I cannot figure out what's wrong with the function because the expression is the same. Can you tell me how to debug the function? Thanks!

 

[Mark44]

Hi guys,
I just wrote a simple code for delta function and print it out. Here it is.

      program main
        implicit none
        integer i
        real*8 del
        real*8,parameter:: ep=1D-2
        real*8,parameter:: pi=3.1415926
        do i=-10,10,1
        write(*,*) i,del(real(i))  ! use function
        enddo
        do i=-10,10,1
        write(*,*) i,(1.0/pi*ep/((real(i))**2+ep**2)) ! no use function
        enddo
        end

      function del(x)
      implicit none
      real*8 del
      real*8 x
      real*8,parameter::ep=1D-2
      real*8,parameter:: pi=3.1415926

      del=(1.0/pi*ep/(x**2+ep**2))

      return
      end

It turns out to be quite wrong by using function del. I cannot figure out what's wrong with the function because the expression is the same. Can you tell me how to debug the function? Thanks!

What output do your two loops produce?

 

[phynewb]

What output do your two loops produce?

It shows
-10 31.830990148286446
-9 4.42379935910539249E-054
-8 3.96072696157977595E-306
-7 31.830990148286446
-6 4.42379936653263965E-054
-5 0.0000000000000000
-4 31.830990148286446
-3 4.42379937841623373E-054
-2 0.0000000000000000
-1 31.830990148286446
0 4.42380394766179441E-054
1 31.830990148286446
2 4.42380242655975552E-054
3 31.830990148286446
4 4.42380241467614869E-054
5 31.830990148286446
6 4.42380240873434528E-054
7 31.830990148286446
8 4.42380240279254245E-054
9 31.830990148286446
10 4.42380239982164045E-054

-10 3.18309583173281331E-005
-9 3.92974701861929112E-005
-8 4.97358443944407051E-005
-7 6.49610718106421125E-005
-6 8.84191714697638248E-005
-5 1.27323451299340604E-004
-4 1.98942445036508832E-004
-3 3.53673738606087140E-004
-2 7.95754859835665312E-004
-1 3.18278073675496917E-003
0 31.830990148286446
1 3.18278073675496917E-003
2 7.95754859835665312E-004
3 3.53673738606087140E-004
4 1.98942445036508832E-004
5 1.27323451299340604E-004
6 8.84191714697638248E-005
7 6.49610718106421125E-005
8 4.97358443944407051E-005
9 3.92974701861929112E-005
10 3.18309583173281331E-005The upper part is from the function and the lower part is calculated without function. It shall look like sharp gaussian distribution around 0 instead of random distribution.

 

[Mark44]

I'm pretty sure this line is causing the problem:

write(*,*) i,del(real(i))  ! use function

In the call to your del function, you are converting an integer value, i, to a four-byte real (a single-precision floating point value), but the function is expecting an eight-byte real (a double-precision floating point value).

I think this will fix the problem:

write(*,*) i,del(dble(i))  ! use function

Here's a link to a page with more information - http://www.liv.ac.uk/HPC/HTMLF90Course/HTMLF90CourseNotesnode114.html .

 

[phynewb]

I'm pretty sure this line is causing the problem:

write(*,*) i,del(real(i))  ! use function

In the call to your del function, you are converting an integer value, i, to a four-byte real (a single-precision floating point value), but the function is expecting an eight-byte real (a double-precision floating point value).

I think this will fix the problem:

write(*,*) i,del(dble(i))  ! use function

Here's a link to a page with more information - http://www.liv.ac.uk/HPC/HTMLF90Course/HTMLF90CourseNotesnode114.html .

Thanks Mark44. It really works! I am always troubled by this accuracy stuff...
Thank you a lot :p

 

[Mark44]

It's more about the size, in bytes, of the variables you're using. For floating point variables, the choices are real*4 (single precision - 4 bytes) and real*8 (double precision - 8 bytes). real() converts whatever to a 4-byte floating point value and dble() converts whatever to an 8-byte floating point value.

Make sure that what you pass to a function is what it is expecting - otherwise you'll have problems.