How to avoid problem with CALL AA(N*3)

[jc_dauphin]

Hi everybody,

I noticed that the Intel Compiler has a problem with computed arguments like:

Call AA(N*5)

....

SUBROUTINE AA(NN)

INTEGER*4 NN

I = NN; ! NN is wrong here

It would work if I use a declared variable

LEN = N*5

CALL AA(LEN)

I suppose there is a compiler or linker parameter to avoid this.

Thank you in advance for any suggestion

Jean-Claude

[Steven_L_Intel1]

The Intel compiler can handle expressions in arguments fine - at least in cases I know about. If you believe you have found a situation where it doesn't, please submit a complete example to Intel Premier Support.

[jc_dauphin]

That's strange Steve because I found the problem in 2 different places. I think I understand the problem. It seems that the type of value generated depends on the type of the variables involved. In my case they are defined as INTEGER*2 and I suppose it generates also a temporary variable of the same type, i.e INTEGER*2 and the subroutine called is expecting an integer*4 thus the value is wrong.

By the way it was working properly on Digital Visual Fortran 6.0a.

I think it's a compiler bug but I'm not sure.

I have summarized the source code of the 2 issues (file attached) I have until now find out. I am afraid that there are other statements like this and we will have to check each call statement.

Best wishes,

Jean-Claude

[jc_dauphin]

Sorry, I forgot to attach the file. Here is the file.

[Steven_L_Intel1]

That isn't a complete program, but I suspect the problem you're seeing is due to the coding error where L and RBAR are declared INTEGER*2 in the caller but you're passing to an INTEGER*4 argument. Passing the character scalar ILINE to array C is also illegal Fortran, though you may be able to get away with it in some cases.