You need to check your system

chung__kristian · ‎04-05-2018

I am running the student version of Intel Parallel Studio XE 2018 (update 2) integrated into Visual Studio 2017 running on Windows 10 64 bit. I am compiling fortran 90 code and trying to use the Extended Eigensolver routines, and I'm currently trying to compile and run the example source file dexample_sparse_f.f in .../mkl/examples/examples_core_f/solvers_eef/source.

I am able to use the MKL lapack routines by simply using the compiler option /Qmkl:sequential.

Using the Intel MKL Link Line Advisor I got the correct compiler and linker options to use the lapack95 routines (https://software.intel.com/en-us/articles/intel-mkl-link-line-advisor/)

However when I try to compile the dexample_sparse_f.f example, I get errors

unresolved external symbol FEASTINIT referenced in function MAIN_

unresolved external symbol DFEAST_SCRGV referenced in function MAIN_

unresolved external symbol DFEAST_SCSREV referenced in function MAIN_

unresolved external symbol DGEMM referenced in function MAIN_

I cannot figure out which library I need to include to use these routines. All I have found is the interface file mkl_solvers_ee.fi in the .../mkl/include directory. Any help appreciated.

Gennady_F_Intel · ‎04-05-2018

You need to check your system paths. Please check MKL and Compiler User's guide how to install these.

below you may see the quick check how this example is compiled and executed from command line prompt.

I used Intel Fortran Compiler and usual option /Qmkl ( linking with LP64 and threading layer of MKL). MKL 2018. Everething works as expected.

...\windows\mkl\examples\solvers_eef\source> ifort /Qmkl dexample_sparse_f.f

Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 18.0.2.185 Build 20180210

Microsoft (R) Incremental Linker Version 14.00.24215.1

-out:dexample_sparse_f.exe

-subsystem:console

-libpath:C:\Apps\Intel2018\compilers_and_libraries\windows\mkl\lib\intel64_win

and Running

C:\Apps\Intel2018\compilers_and_libraries\windows\mkl\examples\solvers_eef\source>dexample_sparse_f.exe

Sparse matrix size 11

Search interval 3.00000000000000 7.00000000000000

Testing dfeast_scsrev

Intel MKL Extended Eigensolvers: double precision driver

Intel MKL Extended Eigensolvers: List of input parameters fpm(1:64)-- if different from default

Intel MKL Extended Eigensolvers: fpm(1)=1

Search interval [3.000000000000000e+00;7.000000000000000e+00]

Intel MKL Extended Eigensolvers: Size subspace 8

#Loop | #Eig | Trace | Error-Trace | Max-Residual

0,6,2.570747125938328e+01,1.000000000000000e+00,6.482954464586578e-06

1,6,2.570747126011605e+01,1.046817236069078e-10,3.387996133676337e-11

2,6,2.570747126011606e+01,5.075305255429287e-16,5.090168922706407e-16

Intel MKL Extended Eigensolvers have successfully converged (to desired tolerance).

FEAST OUTPUT INFO 0

Number of eigenvalues found 6

Computed | Expected

Eigenvalues | Eigenvalues

3.17157287525381 3.17157287525381

4.00000000000000 4.00000000000000

4.12924848418909 4.12924848418909

4.40664990067315 4.40664990067315

6.00000000000000 6.00000000000000

Max value of

| computed eigenvalue -expected eigenvalues | 2.664535259100376E-015

*************************************************

************** REPORT ***************************

*************************************************

# Search interval [Emin,Emax] 3.00000000000000 7.00000000000000

# mode found/subspace 6 8

# iterations 2

Relative error on the trace 5.075305255429287E-016

Eigenvalues/Residuals

1 3.17157287525381 | 1.752559301667446E-016

2 4.00000000000000 | 3.947600454143145E-016

3 4.00000000000000 | 3.904784723663274E-016

4 4.12924848418909 | 1.892424847101820E-016

5 4.40664990067315 | 4.219595942862164E-016

6 6.00000000000000 | 5.090168922706407E-016

Max absolute value of the matrix

(transposed of X)*X-I 4.440892098500626E-016

Testing dfeast_scsrgv

Intel MKL Extended Eigensolvers: double precision driver

Intel MKL Extended Eigensolvers: List of input parameters fpm(1:64)-- if different from default

Intel MKL Extended Eigensolvers: fpm(1)=1

Search interval [3.000000000000000e+00;7.000000000000000e+00]

Intel MKL Extended Eigensolvers: Size subspace 8

#Loop | #Eig | Trace | Error-Trace | Max-Residual

0,7,3.259411364459005e+01,1.000000000000000e+00,4.822065566418076e-01

1,6,2.570747126011606e+01,9.838060549248558e-01,5.007802803472852e-11

2,6,2.570747126011605e+01,1.522591576628786e-15,3.835996311692877e-16

Intel MKL Extended Eigensolvers have successfully converged (to desired tolerance).

FEAST OUTPUT INFO 0

Number of eigenvalues found 6

Computed | Expected

Eigenvalues | Eigenvalues

3.17157287525381 3.17157287525381

4.00000000000000 4.00000000000000

4.12924848418909 4.12924848418909

4.40664990067315 4.40664990067315

6.00000000000000 6.00000000000000

Max | computed eigenvalue -expected eigenvalues | 1.776356839400250E-015

*************************************************

************** REPORT ***************************

*************************************************

# Search interval [Emin,Emax] 3.00000000000000 7.00000000000000

# mode found/subspace 6 8

# iterations 2

Relative error on the trace 1.522591576628786E-015

Eigenvalues/Residuals

1 3.17157287525381 | 1.359734746820817E-016

2 4.00000000000000 | 3.027799317329866E-016

3 4.00000000000000 | 3.820159509369450E-016

4 4.12924848418909 | 2.429464330738821E-016

5 4.40664990067315 | 3.835996311692877E-016

6 6.00000000000000 | 3.042645045144006E-016

Max absolute value of the matrix

(transposed of X)*X-I 7.771561172376096E-016

0

chung__kristian · ‎04-05-2018

Thank you, this works fine from the command line, but it's not working if I try to run it from within Visual Studio, any idea why this might be? I would think this should simply require specifying the /Qmkl compiler option as I can do from the command line, but it can't find the external functions.

Linking to the MKL Extended Eigensolver