Problem linking to Math kernel in Fortran Composer XE for OSX

Problem linking to Math kernel in Fortran Composer XE for OSX

Imagen de John Kornak
John Kornak
Sáb, 30/04/2011 - 16:39

Hi,

I would really appreciate any help to solve the following problem (or to point me in the direction as to where I can find a solution):

I recently installed Intel Fortran Composer XE for OS X (Snow Leopard on a new MacBook Pro) and I am trying to link a program to the fast Fourier Transform functions, but am having problems:

> ifort CGoct3D.f90 -L$MKLPATH -I$MKLINCLUDE -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -openmp -lpthread
CGoct3D.f90(171): error #6284: There is no matching specific function
for this generic function reference. [DFTICREATEDESCRIPTOR]
StatusBig(1) = DftiCreateDescriptor(FFT_BIG, DFTI_DOUBLE, DFTI_COMPLEX, 2, DimBig)
---------------^
CGoct3D.f90(176): error #6284: There is no matching specific function
for this generic function reference. [DFTICOMPUTEFORWARD]
StatusBig(i) = DftiComputeForward(FFT_BIG,newA(:,:,i),DFTBig(:,:,i))
-------------------^
CGoct3D.f90(227): error #6284: There is no matching specific function
for this generic function reference. [DFTICOMPUTEBACKWARD]
StatusBig(i) = DftiComputeBackward(FFT_Big,rmatBig(:,:,i))
---------------^
CGoct3D.f90(240): error #6284: There is no matching specific function
for this generic function reference. [DFTICOMPUTEFORWARD]
StatusBig(i) = DftiComputeForward(FFT_BIG,bmat(:,:,i),DFTBig(:,:,i))
-------------------^
compilation aborted for CGoct3D.f90 (code 1)
>

I had compiled mkl_dfti.f90 as root:

> cd /opt/intel/composerxe-2011.3.167/mkl/include
include > ifort -I$MKLINCLUDE -c mkl_dfti.f90
include >

> ls -l mkl_dft*
-rwxrwxrwx 1 kornak staff 9577 Apr 30 16:19 mkl_dft_type.mod
-r--r--r-- 1 27945 wheel 28055 Mar 18 03:42 mkl_dfti.f90
-r--r--r-- 1 27945 wheel 10561 Mar 18 03:42 mkl_dfti.h
-rwxrwxrwx 1 kornak staff 52484 Apr 30 16:19 mkl_dfti.mod
-rw-r--r-- 1 root wheel 783 Apr 30 16:19 mkl_dfti.o

In addition, I have the following environment variables set:

> echo $MKLROOT
/opt/intel/composerxe-2011.3.167/mkl
> echo $MKLPATH
/opt/intel/composerxe-2011.3.167/mkl/lib
> echo $MKLINCLUDE
/opt/intel/composerxe-2011.3.167/mkl/include
>

Finally, here are the relevant parts of the frotran 90 code:

PROGRAM bayes3D
 ! .. Use Statements ..
  Use MKL_DFTI
 ! none in USE
 ! .. Implicit None Statement ..
 IMPLICIT NONE
 ! .. Default Accessibility ..

! --- snip !

ALLOCATE (DimBig(2))
DimBig(1) = nl1
DimBig(2) = nl2
StatusBig(1) = DftiCreateDescriptor(FFT_BIG, DFTI_DOUBLE, DFTI_COMPLEX, 2, DimBig)
StatusBig(1) = DftiCommitDescriptor(FFT_BIG)

ALLOCATE (DFTBig(nl1,nl2,zl))
DO i=1,zl
    StatusBig(i) = DftiComputeForward(FFT_BIG,newA(:,:,i),DFTBig(:,:,i))
END DO

Thanks for any insight.

Cheers

John

RSS Superior
publicaciones de 11 / 0 nuevos
Último envío
Para obtener más información sobre las optimizaciones del compilador, consulte el aviso sobre la optimización.
Imagen de mecej4
mecej4
Sáb, 30/04/2011 - 21:05

All the errors are related to the types of the arguments to the FFT subroutines: "There is no matching specific function
for this generic function reference". Yet, you seem to think that the declarations of those variables were not relevant and did not show them. Without that information, it is impossible to diagnose the problem.

Superior
Imagen de John Kornak
John Kornak
Mié, 04/05/2011 - 11:30

Thank you for the very quick response.

I apologize for the incomplete information. I assumed that I was having a problem linking to the library without considering alternatives!

Redirecting me to realize it was a problem with my coding meant that I was able to debug it and get my code running correctly.

Thanks again

Cheers

John

Superior
Imagen de John Kornak
John Kornak
Mié, 04/05/2011 - 11:32

PS. Is there a way to delete this posting from the forum?

Given my poor initial description of the problem I doubt this will be of much use to anyone else.

Cheers

John

Superior
Imagen de TimP (Intel)
TimP (Intel)
Mié, 04/05/2011 - 12:29

You should be able to delete your own posts, at least those which no one has yet replied, using the delete link which should appear next to the reply link. It's generally preferred not to delete posts which leave a dangling reply. I'll delete this reply if you delete your preceding ones.

Superior
Imagen de John Kornak
John Kornak
Mié, 04/05/2011 - 12:53

Sorry if this is a dumb question but what do you mean by a "dangling reply"?

Cheers

John

Superior
Imagen de TimP (Intel)
TimP (Intel)
Mié, 04/05/2011 - 15:07

If someone has replied to your post, and you delete your post later on, the value of that reply may be destroyed.

Superior
Imagen de John Kornak
John Kornak
Mié, 04/05/2011 - 15:42

Got it. Thanks.

I will just leave the post as is then.

Cheers

John

Superior
Imagen de yuan z.
yuan z.
Mar, 05/03/2013 - 22:50

I am facing the same question as you once did. how did you solve your problem? Can you give me some help . I am looking forward to your reply.

My codes is as follows:

error #6284: There is no matching specific function for this generic function reference.   [FFT]

USE FOURIER, ONLY : &
  FFT, &                ! The Fast Fourier Transform routine.
  GetFFTDims, &
  CleanFFT, &
  GetFFTComplexDims

VWPot = 0.5 * FFT( FFT(VWPot) * qTable**2 ) / VWPot

Superior
Imagen de John Kornak
John Kornak
Mié, 06/03/2013 - 11:54

Sorry yuan z. I do not remember exactly what the problem. I think it was an error in my code where I declared an array to have the wrong type (e.g., integer array instead of real).

Superior
Imagen de mecej4
mecej4
Mié, 06/03/2013 - 16:33

Yuan Z.:

Show the interface for function FFT, and the declarations of the variables VWPot and qTable. Apparently, your module Fourier contains one or more specific functions with the generic name FFT, and the type of the argument in the invocation of the function FFT does not match the types of the arguments of any of the specific functions.

Superior

Inicie sesión para dejar un comentario.