djacobix aborts after some iterations

Thomas_H_2 — Thu, 18 Jan 2018 15:53:41 GMT

Hello,

i want to calculate the Jacobian of a complicated function. I sucessfully implemented djacobi for a simple test case using the reference page https://software.intel.com/en-us/mkl-developer-reference-fortran-2018-beta-jacobi

The actual function is complicated so i'll start in a compressed form, if you need to see more snippets just tell me.

module ContactForceTest_mod
    use MKL_RCI
    use MKL_RCI_type
    implicit none
contains
    subroutine TestContactForce
        type(ContactForce) :: Contact
        external :: wrapper
        ...
        if (djacobix(wrapper,12,12,Jacobian,Displacement,1e-6,%VAL(LOC(Contact))) /= TR_SUCCESS) then
            write(*,*), '| error in djacobix'
            call MKL_FREE_BUFFERS
            stop 1;
        end if
    end subroutine
end module

subroutine wrapper(m,n,Displacement, Force, Contact)
    use ContactForceTest_mod
    implicit none
    integer, intent(in) :: m,n
    real, dimension(12) :: RelativeDisplacement
    real, dimension(12) :: Force
    type(ContactForce_class) :: Contact

    call Contact%NonlinearContactForce(12,12,RelativeDisplacement, ContactForce)
end subroutine wrapper

I'm using the wrapper subroutine as i need to call a type bound procedure as the function of which i want to have the jacobian (this is apparently working). All variables are properly declared and if i commend the jacobian part, the program is working flawlessly.

After some iterations (i can see Relativedisplacement change its values by 1e-06 one after another), bash just tells me "aborted" without any further message. My problem at the moment is to locate the error. It happens apparently somewhere in djacobix. Is there a way to find out where exactly (as the routine is at least doing some iterations)?

Or any suggestions?

Edit: Here's something from Valgrind:

==14058== Memcheck, a memory error detector
==14058== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
==14058== Using Valgrind-3.10.1 and LibVEX; rerun with -h for copyright info
==14058== Command: ./datar_test
==14058== 
==14058== Conditional jump or move depends on uninitialised value(s)
==14058==    at 0x603C8D: __intel_sse2_strcpy (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058==    by 0x57F277: for__add_to_lf_table (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058==    by 0x5BC548: for__open_proc (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058==    by 0x58620F: for__open_default (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058==    by 0x5A9C23: for_write_seq_lis (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058==    by 0x40E2CA: fruit_mp_init_fruit_ (fruit.f90:635)
==14058==    by 0x44B553: MAIN__ (datar_test.f90:18)
==14058==    by 0x4032DD: main (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058== 
 
 Test module initialized
 
    . : successful assert,   F : failed assert 
 
 --------------
 Starting contact force spectrum test
==14058== Syscall param sched_setaffinity(mask) points to unaddressable byte(s)
==14058==    at 0x6B6EFB9: syscall (in /lib64/libc-2.19.so)
==14058==    by 0x40A0787: __kmp_affinity_determine_capable (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libiomp5.so)
==14058==    by 0x4089F97: __kmp_env_initialize(char const*) (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libiomp5.so)
==14058==    by 0x4081175: ??? (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libiomp5.so)
==14058==    by 0x4074428: __kmp_get_global_thread_id_reg (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libiomp5.so)
==14058==    by 0x407FF98: __kmp_parallel_initialize (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libiomp5.so)
==14058==    by 0x406EC7D: omp_get_num_procs (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libiomp5.so)
==14058==    by 0x54211FA: MKL_get_N_Cores (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libmkl_intel_thread.so)
==14058==    by 0x5420CA8: mkl_read_threads_env (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libmkl_intel_thread.so)
==14058==    by 0x5420694: mkl_serv_mkl_domain_get_max_threads (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libmkl_intel_thread.so)
==14058==    by 0x57BB632: mkl_dft_commit_descriptor_d_c2c_1d_omp (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libmkl_intel_thread.so)
==14058==    by 0x4F4279C: mkl_dft_dfti_commit_descriptor_external (in /global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/libmkl_intel_lp64.so)
==14058==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==14058== 
OMP: Warning #2: Cannot open message catalog "libiomp5.cat":
OMP: System error #2: No such file or directory
OMP: Hint: Check NLSPATH environment variable, its value is "/global/linux/64_bit/opt/intel/mkl/10.2.2.025/lib/em64t/locale/%l_%t/%N".
OMP: Info #3: Default messages will be used.
==14058== Conditional jump or move depends on uninitialised value(s)
==14058==    at 0x5888D5: for_realloc_lhs (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058==    by 0x46584B: contactforceparameters_mod_mp_getnormalizedequilibrium_ (ContactForceParameters_mod.f90:196)
==14058==    by 0x50CE29: contactforce_mod_mp_init_ (ContactForce_mod.f90:53)
==14058==    by 0x406269: contactforcetest_mod_mp_testcontactforce_ (ContactForceTest.f90:71)
==14058==    by 0x44B586: MAIN__ (datar_test.f90:26)
==14058==    by 0x4032DD: main (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058== 
==14058== Conditional jump or move depends on uninitialised value(s)
==14058==    at 0x5888D5: for_realloc_lhs (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058==    by 0x4666FA: contactforceparameters_mod_mp_getcontactforcecorrection_ (ContactForceParameters_mod.f90:241)
==14058==    by 0x50E076: contactforce_mod_mp_init_ (ContactForce_mod.f90:55)
==14058==    by 0x406269: contactforcetest_mod_mp_testcontactforce_ (ContactForceTest.f90:71)
==14058==    by 0x44B586: MAIN__ (datar_test.f90:26)
==14058==    by 0x4032DD: main (in /home/SERVER-hoffmann/sync/04_DATAR/datar_shared/Unittest/datar_test)
==14058== 
.........
 
     Start of FRUIT summary: 
 
 SUCCESSFUL!
 
   No messages 
 Total asserts :              9
 Successful    :              9
 Failed        :              0
Successful rate:   100.00%
 
 Successful asserts / total asserts : [            9 /           9  ]
 Successful cases   / total cases   : [            0 /           0  ]
   -- end of FRUIT summary
==14058== 
==14058== HEAP SUMMARY:
==14058==     in use at exit: 2,594 bytes in 8 blocks
==14058==   total heap usage: 2,462 allocs, 2,454 frees, 7,276,564 bytes allocated
==14058== 
==14058== LEAK SUMMARY:
==14058==    definitely lost: 0 bytes in 0 blocks
==14058==    indirectly lost: 0 bytes in 0 blocks
==14058==      possibly lost: 0 bytes in 0 blocks
==14058==    still reachable: 2,594 bytes in 8 blocks
==14058==         suppressed: 0 bytes in 0 blocks
==14058== Reachable blocks (those to which a pointer was found) are not shown.
==14058== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==14058== 
==14058== For counts of detected and suppressed errors, rerun with: -v
==14058== Use --track-origins=yes to see where uninitialised values come from
==14058== ERROR SUMMARY: 4 errors from 4 contexts (suppressed: 0 from 0)

Hi Thomas,

Khang_N_Intel — Fri, 19 Jan 2018 21:00:31 GMT

Hi Thomas,

It would be helpful if you can provide us with the following information:

1) Testing system configuration: hardware and software (OS and software use including version numbers)

2) Your whole test harness, not just some code snippets, so that we can recreate the errors.

Thanks,

Khang

topic Hi Thomas, in Intel® oneAPI Math Kernel Library

djacobix aborts after some iterations

Hi Thomas,