forums.silverfrost.com

[christyleomin]

I would like to know the impact of the following calls with the arguments as stated:

[davidb]

The code isn't valid Fortran so its difficult to answer this.

You shouldn't pass a two-dimensional array actual argument when a one-dimensional dummy argument is expected.
_________________
Programmer in: Fortran 77/95/2003/2008, C, C++ (& OpenMP), java, Python, Perl

[JohnCampbell]

In each call to A_temp3, you are effectively passing a starting address in the array crtem.
Subroutine A_temp3 takes this starting address as the start of a vector temp2(iramp).

Presumably you have the declaration dimension crtem(iramp,5)

call 1: crtem provides the first iramp locations, probably equivalent to crtem(1:iramp,1)
call 2: provides crtem(1:iramp,2)
call 3: provides crtem(1:iramp,3)
call 4: provides crtem(1:iramp,4)
call 5: provides crtem(1:iramp,4)

Note:
what is iret ; it is not provided in the calls to A_temp3
temp1 could be declared local in A_temp3
you should be careful with the kind of your real arguments

I think your routine is equivalent to:

[christyleomin]

If crtem is actually an allocatable array, then how do wee deal with it. I do not want temp1 to be local in A_temp3 . Is it possible?

[JohnCampbell]

There should not be any problem as long as you have allocated crterm first, with:
ALLOCATE ( crterm(n,m) )
where n >= iramp and m >= 4
The memory address of crtemp(1,2) is determined and transferred when the subroutine is called.
There is no problem transferring temp1 through the subroutine call, as long as it is dimensioned to be >= idist

Also, you have declared iret in the subroutine definition, but not provided it as an argument in the calls. This would probably produce an access violation when you run the program.

John

[christyleomin]

Thanks John..

Yes, I'm taking care of iret.

Yes, crtem is allocated as you said :

[JohnCampbell]

Once you allocate the arrays, they can be supplied in a subroutine call as a "normal" array. You do not declare them as allocatable in A_temp3. (That is possible in F2003, although it means something different.)

eg

[christyleomin]

I trried this bit of code and it works fine (even in Fortran90)

That is, all changes done in subroutine checktopo_1 below are refelected correctly in TOPOLOGY array (in the write statement after the call to checktopo_1) below.

[JohnCampbell]

I know you are allowed to do it, but why would you use an array section in call checktopo_1(TOPOLOGY(2,1:n2),n2) ?

You are just asking for trouble.
I'd never write this, although there are possibly many who do.

You have TOPOLOGY in the module, so wouldn't it be cleaner to USE it.

My recommendation,

[christyleomin]

Thanks...I know it is not a good coding ethic.

I have a huge pice of code that needs to eb understood and upgraded.

Why do you say "asking for trouble"? Is it incorrect to pass an allocatable array as an argument?

[christyleomin]

Hi John,

Please see my post immediately above this as well..I'm facing a peculiar problem. There are times my code compiles/builds without any issues/problems (having used above approach).

But there are times when it gives me a bunch of errors stating

[PaulLaidler]

You will need to show us the code that is generating the error report.

[mecej4]

[JohnCampbell]

Calls of the type: call checktopo_1(TOPOLOGY(2,1:n2),n2)
This can cause problems as it can create a temporary storage for the array section. Repeated use can lead to a "Stack Overflow". I would recommend to avoid array sections, as they can readily lead to Stack Overflow problems.

To answer your othe problem: