Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.
6680 Discussions

Problem using MKL Trust Region EXAMPLE_EX_NLSQP_F90_X

YeltsinVega
Novice
664 Views

Hello,

I am using the example provided EXAMPLE_EX_NLSQP_F90_X.f90 in the folder of Intel oneAPI Math Kernel Library directory. This compile successfully, but when I run it in debug mode, the result of dtrnlsp_init in line 108 is TR_INVALID_OPTION.

Can you please offer me some suggestions on how to run it correctly? Thank you in advance.

The Fortran code of the example is:

 

!===============================================================================
! Copyright 2004-2021 Intel Corporation.
!
! This software and the related documents are Intel copyrighted  materials,  and
! your use of  them is  governed by the  express license  under which  they were
! provided to you (License).  Unless the License provides otherwise, you may not
! use, modify, copy, publish, distribute,  disclose or transmit this software or
! the related documents without Intel's prior written permission.
!
! This software and the related documents  are provided as  is,  with no express
! or implied  warranties,  other  than those  that are  expressly stated  in the
! License.
!===============================================================================

!   Content : TR Solver F90 example
!
!********************************************************************************


module u_data
type, public :: my_data
      integer a
      integer sum
end type my_data
end module

!    nonlinear least square problem without boundary constraints
    include 'mkl_rci.f90'
program EXAMPLE_EX_NLSQP_F90_X
    use MKL_RCI
    use MKL_RCI_type
    use u_data
    implicit none
!   supplementary routine to track/free memory
    external mkl_free_buffers
!   user's objective function
    external extended_powell 
!   n - number of function variables
!   m - dimension of function value
    integer n, m
    parameter (n = 4)
    parameter (m = 4)
!   precisions for stop-criteria (see manual for more details)
    real*8 eps(6)
!   solution vector. contains values x for f(x)
    real*8 x(n)
!   iter1 - maximum number of iterations
!   iter2 - maximum number of iterations of calculation of trial-step
    integer iter1, iter2
    parameter (iter1 = 1000)
    parameter (iter2 = 100)
!   initial step bound
    real*8 rs
!   reverse communication interface parameter
    integer RCI_Request
!   controls of rci cycle
    integer successful
!   function (f(x)) value vector
    real*8 fvec(m)
!   jacobi matrix
    real*8 fjac(m*n)
!   number of iterations
    integer iter
!   number of stop-criterion
    integer st_cr
!   initial and final residuals
    real*8 r1, r2
!   TR solver handle
    type(HANDLE_TR) :: handle
!   cycle's counter
    integer i
!   results of input parameter checking
    integer info(6)

!   Additional users data
    type(my_data) :: m_data
    m_data%a = 1
    m_data%sum = 0
    rs = 0.0

!   set precisions for stop-criteria
    do i = 1, 6
        eps(i) = 0.00001
    end do
!   set the initial guess
    do i = 1, n/4
        x(4 * (i-1) + 1) = 3.0
        x(4 * (i-1) + 2) = -1.0
        x(4 * (i-1) + 3) = 0.0
        x(4 * (i-1) + 4) = 1.0
    end do
!   set initial values
    do i = 1, m
        fvec(i) = 0.0
    end do
    do i = 1, m * n
        fjac(i) = 0.0
    end do
!   initialize solver (allocate memory, set initial values)
!       handle       in/out: TR solver handle
!       n       in:     number of function variables
!       m       in:     dimension of function value
!       x       in:     solution vector. contains values x for f(x)
!       eps     in:     precisions for stop-criteria
!       iter1   in:     maximum number of iterations
!       iter2   in:     maximum number of iterations of calculation of trial-step
!       rs      in:     initial step bound
    if (dtrnlsp_init (handle, n, m, x, eps, iter1, iter2, rs) /= TR_SUCCESS) then
!       if function does not complete successfully then print error message
        print*,'| error in dtrnlsp_init'
!       Release internal Intel(R) oneAPI Math Kernel Library (oneMKL) memory that might be used for computations.
!       NOTE: It is important to call the routine below to avoid memory leaks
!       unless you disable  Intel oneMKL Memory Manager
        call MKL_FREE_BUFFERS
!       and exit
        stop 1
    end if
!   Checks the correctness of handle and arrays containing Jacobian matrix, 
!   objective function, lower and upper bounds, and stopping criteria.
    if (dtrnlsp_check (handle, n, m, fjac, fvec, eps, info) /= TR_SUCCESS) then
!       if function does not complete successfully then print error message
        print*,'| error in dtrnlsp_init'
!       Release internal  Intel oneMKL memory that might be used for computations.
!       NOTE: It is important to call the routine below to avoid memory leaks
!       unless you disable  Intel oneMKL Memory Manager
        call MKL_FREE_BUFFERS
!       and exit
        stop 1
    else
        if ( &
!           The handle is not valid. 
            info(1) /= 0 .or. &
!           The fjac array is not valid.
            info(2) /= 0 .or. &
!           The fvec array is not valid.
            info(3) /= 0 .or. &
!           The eps array is not valid.
            info(4) /= 0 ) then
            print*, '| input parameters for dtrnlsp_solve are not valid'
!           Release internal  Intel oneMKL memory that might be used for computations.
!           NOTE: It is important to call the routine below to avoid memory leaks
!           unless you disable  Intel oneMKL Memory Manager
            call MKL_FREE_BUFFERS
!           and exit
            stop 1
        end if
    end if
!   set initial rci cycle variables
    RCI_Request = 0
    successful = 0
!   rci cycle
    do while (successful == 0)
!       call tr solver
!       handle               in/out: tr solver handle
!       fvec         in:     vector
!       fjac         in:     jacobi matrix
!       RCI_request in/out:  return number which denote next step for performing
        if (dtrnlsp_solve (handle, fvec, fjac, RCI_Request) /= TR_SUCCESS) then
!           if function does not complete successfully then print error message
            print*, '| error in dtrnlsp_solve'
!           Release internal  Intel oneMKL memory that might be used for computations.
!           NOTE: It is important to call the routine below to avoid memory leaks
!           unless you disable  Intel oneMKL Memory Manager
            call MKL_FREE_BUFFERS
!           and exit
            stop 1;
        end if
!       according with rci_request value we do next step
        if (RCI_Request == -1 .or. &
            RCI_Request == -2 .or. &
            RCI_Request == -3 .or. &
            RCI_Request == -4 .or. &
            RCI_Request == -5 .or. &
            RCI_Request == -6) then
!           exit rci cycle
            successful = 1
        end if
        if (RCI_Request == 1) then
!           recalculate function value
!               m            in:     dimension of function value
!               n            in:     number of function variables
!               x            in:     solution vector
!               fvec    out:    function value f(x)
            call extended_powell (m, n, x, fvec, m_data)
        end if
        if (RCI_Request == 2) then
!           compute jacobi matrix
!               extended_powell      in:     external objective function
!               n               in:     number of function variables
!               m               in:     dimension of function value
!               fjac            out:    jacobi matrix
!               x               in:     solution vector
!               jac_eps         in:     jacobi calculation precision !/
            if (djacobix (extended_powell,n,m,fjac,x,eps(1),%VAL(LOC(m_data))) /= &
                TR_SUCCESS) then
!               if function does not complete successfully then print error message
                print*, '| error in djacobix'
!               Release internal  Intel oneMKL memory that might be used for computations.
!               NOTE: It is important to call the routine below to avoid memory leaks
!               unless you disable  Intel oneMKL Memory Manager
                call MKL_FREE_BUFFERS
!               and exit
                stop 1;
            end if
        end if
    end do
!   get solution statuses
!       handle            in:        TR solver handle
!       iter              out:       number of iterations
!       st_cr             out:       number of stop criterion
!       r1                out:       initial residuals
!       r2                out:       final residuals
    if (dtrnlsp_get (handle, iter, st_cr, r1, r2) /= TR_SUCCESS) then
!       if function does not complete successfully then print error message
        print*, '| error in dtrnlsp_get'
!       Release internal  Intel oneMKL memory that might be used for computations.
!       NOTE: It is important to call the routine below to avoid memory leaks
!       unless you disable  Intel oneMKL Memory Manager
        call MKL_FREE_BUFFERS
!       and exit
        stop 1
    end if
    print*, 'Iterations : ',iter
    print*, 'Final residual : ',r2
    print*, 'Stop-criteria : ',st_cr
!   free handle memory
    if (dtrnlsp_delete (handle) /= TR_SUCCESS) then
!       if function does not complete successfully then print error message
        print*, '| error in dtrnlsp_delete'
!       Release internal  Intel oneMKL memory that might be used for computations.
!       NOTE: It is important to call the routine below to avoid memory leaks
!       unless you disable  Intel oneMKL Memory Manager
        call MKL_FREE_BUFFERS
!       and exit
        stop 1
    end if

!   Release internal  Intel oneMKL memory that might be used for computations.
!   NOTE: It is important to call the routine below to avoid memory leaks
!   unless you disable  Intel oneMKL Memory Manager
    call MKL_FREE_BUFFERS
!   if final residual less then required precision then print pass

    print*, 'User data ', m_data%sum

    if (r2 < 0.00001) then
        print*, '|         dtrnlsp Powell............PASS'
        stop 0
!   else print failed
    else
        print*, '|         dtrnlsp Powell............FAILED'
        stop 1;
    end if
end program EXAMPLE_EX_NLSQP_F90_X

!     nonlinear system equations without constraints
!     routine for extended Powell function calculation
!     m     in:     dimension of function value
!     n     in:     number of function variables
!     x     in:     vector for function calculating
!     f     out:    function value f(x)
!     user_data in: additional users data

subroutine extended_powell (m, n, x, f, user_data)
    use u_data
    implicit none
    integer m, n
    real*8  x(n), f(m)
    type(my_data) :: user_data
    integer i

    if (n /= m) then
        print *, 'm must equal n for this example function'
        stop 1
    end if

    user_data%sum = user_data%sum + user_data%a

    do i = 1, m/4
        f(4*(i-1)+1) = x(4*(i-1)+1) + 10.0 * x(4*(i-1)+2)
        f(4*(i-1)+2) = 2.2360679774998 * (x(4*(i-1)+3) - x(4*(i-1)+4))
        f(4*(i-1)+3) = ( x(4*(i-1)+2) - 2.0 * x(4*(i-1)+3) )**2
        f(4*(i-1)+4) = 3.1622776601684 * (x(4*(i-1)+1) - x(4*(i-1)+4))**2
    end do
end subroutine extended_powell

 

 

0 Kudos
1 Solution
mecej4
Black Belt
467 Views

There is one inconsistency that stood out when I looked at the source file and the build log. You have default (i.e., 4-byte) integers in the program variables. In particular, the integer type arguments to MKL routines are 4-byte integers. Therefore, your linking with the mkl_intel_ilp64.lib is inappropriate. You should be using the mkl_intel_lp64.lib.

Consult the MKL Link-Line Advisor for the correct options to use.

View solution in original post

5 Replies
VidyalathaB_Intel
Moderator
530 Views

Hi,

 

Thanks for reaching out to us.

 

>>the result of dtrnlsp_init in line 108 is TR_INVALID_OPTION

We have tested the example Fortran code from our end which comes with the latest version of oneMKL 2022 in Debug Mode in VS2022 but we are not getting any such results as TR_INVALID_OPTION.

 

Output from VS2022:

VidyalathaB_Intel_0-1646899195762.png

 

>>Can you please offer me some suggestions on how to run it correctly?

Could you please let us know MKL version being used and the compiler along with the OS details? (If you are using VS2022 please let us know the exact version of VS 2022 you are working with.)

 

Alternately, you can try running the same example from Intel oneAPI command prompt and check whether you are getting the same results as in VS2022.

Command: ifort /Qmkl <filename>

 

If the issue still persists please get back to us with the steps you followed and also attach the project file.

 

Regards,

Vidya.

 

YeltsinVega
Novice
474 Views

Hello,

Thanks for your reply. When I run it in debug mode from de IDE I obtain de print of line 110:

YeltsinVega_0-1646921730218.png

When I see the result from dtrnlsp_init function, I obtain 1502 that correspond to TR_INVALID_OPTION.

When I run this from Intel oneAPI command prompt I obtain a good result:

YeltsinVega_1-1646921963362.png

The versions of the IDE and the MKL library are:

Microsoft Visual Studio Enterprise 2022 (64-bit) Version 17.1.0
Intel® oneAPI Math Kernel Library (oneMKL) 2022.0.3
Intel Libraries for oneAPI – toolkit version: 2022.1.3, extension version 22.0.0.17, Package ID: w_oneAPI_2022.0.3.226, Copyright © 2019-2021 Intel Corporation. All rights reserved.
Intel® C++ Compiler – toolkit version: 2022.1.3, extension version 22.0.17, Package ID: w_oneAPI_2022.0.3.226, Copyright © 2002-2021 Intel Corporation. All rights reserved.
Intel® Fortran Compiler – toolkit version: 2022.1.3, extension version 22.0.0065.17, Package ID: w_oneAPI_2022.0.3.226, Copyright © 2002-2021 Intel Corporation. All rights reserved.

I am attaching you the project file and BuildLog with the Verbose. I change the extensions because I couldn't attach files with extension htm and vfproj.

Thanks in advance,

 

Yeltsin

mecej4
Black Belt
468 Views

There is one inconsistency that stood out when I looked at the source file and the build log. You have default (i.e., 4-byte) integers in the program variables. In particular, the integer type arguments to MKL routines are 4-byte integers. Therefore, your linking with the mkl_intel_ilp64.lib is inappropriate. You should be using the mkl_intel_lp64.lib.

Consult the MKL Link-Line Advisor for the correct options to use.

YeltsinVega
Novice
453 Views

Thank you very much. It works.

VidyalathaB_Intel
Moderator
448 Views

Hi,

 

>>There is one inconsistency that stood out

Yes, the issue comes from linking the application against mkl_intel_ilp64.lib as pointed out by mecje4.

This is the reason behind the error that you are getting while running the code.

I've checked it from my end and could see that linking with mkl_intel_ilp64.lib causes the error while linking against mkl_intel_lp64.lib gives correct results.

 

So please try changing the option which you have included in the additional options in your VS and then build the project.

 

In most of the cases, the automatic settings which we do just by enabling the MKL library under Project → Property Pages →Fortran → Libraries →Use Intel oneAPI Math Kernel Library is enough for building the application, and no need of adding any additional options until and unless if we want to do the settings manually.

 

If you still have any concerns regarding the MKL settings in VS please refer to the below article for full details

https://www.intel.com/content/www/us/en/developer/articles/training/how-to-build-onemkl-application-...

 

Thanks for accepting the solution.

Glad to know that your issue is resolved.

As your issue is resolved we are closing this thread. Please post a new question if you need any additional information from Intel as this thread will no longer be monitored.

Have a Great Day!

 

Regards,

Vidya.

 

Reply