I have been using MKL non-symmetric solver for my finite element code (up to MKL Version10), but have a probelm with updated version of MKL when call pardiso_unsym ( reordering):
phase = 11 ! only reordering and symbolic factorization
CALL pardiso (pt, maxfct, mnum, mtype, phase, n, a, ia, ja,
1 idum, nrhs, iparm, msglvl, ddum, ddum, error)
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
libpthread.so.0 00002AAAAC8B754B Unknown Unknown Unknown
Stack trace terminated abnormally.
The compiler link I used are:
Currently Loaded Modulefiles:
1) Barrine/2.0 2) mpt/2.00 3) intel-cmkl-11/126.96.36.199.8.273 4) intel-fc-11/188.8.131.52.273
COM = ifort -m64 -w -fpp -I/sw/sdev/intel/ComposerXE/composer_xe_2011_sp1.8.273/mkl/include
LIBS = -L"/sw/sdev/intel/ComposerXE/composer_xe_2011_sp1.8.273/mkl/lib/intel64" -lmkl_intel_lp64 \
-liomp5 -lpthread -lm
In addtion, my code works well with the SuperLU, but not with MKL Pardiso at the moment. Any advices to fix it? Thanks in advance
Could you provide testcase or matrix elements with iparm array to reproduce issue on our side? It could help us to investigate it
With best regards,
Did you follow a checklist such as
to track this down? No doubt, you're aware that the most common situation is a requirement to increase stack limit or try heap-arrays option.
Thanks Alexander and Tim.
I had a probelm to post it here and thus I emailed you directly this Monday. I used to check the code up to item 5) as listed by Tim.
My probelm is not solved yet and thus I attached the matrix element and solver code (iparm) below. Look forward to your further advice.
I've turn on matrix checker on your matrix (iparm(27)=1) and got next problem:
*** Error in PARDISO (incorrect input matrix ) error_num= 24
*** Input check: j=4908159, ja(j)=35835, ja(j+1)=35830 are incompatible
The following ERROR was detected: -1
Looks like your ja array is incorrect.
With best regards,