is IMSL feature set limited?

[Brian_Murphy]

I recently bought IVF with IMSL. The compiler will let me call some routines like D_GVCCG but it won't let me call LIN_GEIG_GEN CALL D_GVCCG(B_SYS_COMPLEX, A_SYS_COMPLEX, ALPHA, BETAV, CEIG%VEC) CALL lin_geig_gen(B_SYS, A_SYS, ALPHA, BETAV, v=CEIG%VEC) Here is the message I get from the compiler Error 1 error #6285: There is no matching specific subroutine for this generic subroutine call. [LIN_GEIG_GEN] Am I missing something??

[mecej4]

The compiler will refuse to accept a CALL statement when there is sufficient information from the context of the call for it to determine that an explicit interface is required and either such an interface is not provided or the types of the actual arguments do not match those specified in one of the corresponding subroutine.

In other cases, the compiler may decide (either correctly, or because there is no information to tell it otherwise) that an implicit interface can be used. In such cases, no error occurs at compilation time but a linker error or a runtime error may occur instead.

You should show the declarations of the variables in the argument list and any USE, INCLUDE and INTERFACE statements/blocks in the calling routine.

[Steven_L_Intel1]

The problem you're having is unrelated to "feature limits" You're calling lin_geig_gen with arguments that don't match one of the declared signatures for that routine. I would also discourage you from calling D_GVCCG - call the generic form. If you call the specific you may end up using the wrong signature, especially if you don't USE the appropriate _INT module.

As mecej4 says, we need to see the declarations of all the arguments to lin_geig_gen.

To answer your question, no, there is not a limited feature set. There is a handful of routines marked "deprecated" in a table that are omitted from the library, but everything documented normally is there.

[Brian_Murphy]

I'm working on getting the arguments figured out. Meanwhile, the Roguewave web site says the IMSL package includes ARPACK eigensolvers for sparse matrices, but ARPACK or Arnoldi are nowhere to be found in the MATH.PDF documentation file that I've got. That's what got me thinking I'm missing something.

[Steven_L_Intel1]

ARPACK is new in IMSL 7.0, which is not yet available from us. We expect it to arrive sometime in 2014.

[Brian_Murphy]

Bummer.  I've been using an Arnoldi eigensolver in another program since about 1999, and it runs rings around anything else I've every tried.

[mecej4]

Brian Murphy wrote:

 I've been using an Arnoldi eigensolver in another program since about 1999, and it runs rings around anything else I've every tried.

IMSL 7 is available directly from Roguewave now. It is straightforward to build ARPACK from the sources at http://www.caam.rice.edu/software/ARPACK/ , as was recently mentioned in a sister forum thread: http://software.intel.com/en-us/forums/topic/302621 .

[Brian_Murphy]

The problem was indeed in the declrations.  It works fine now that I've got those straightended out.

Cheers,

Brian