parallel version of Pardiso solver won't converge

[cjl88]

https://coin-or.github.io/Ipopt/INSTALL.html says:
"
If you are using the parallel version of Pardiso, you need to specify the number of processors it should run on with the environment variable OMP_NUM_THREADS, as described in the Pardiso manual.
If you want to compile Ipopt with the Pardiso library that is included in Intel MKL, it is sufficient to ensure that MKL is used for the linear algebra routines (Blas/Lapack), see BLAS and LAPACK.
"

I am using Pardiso because I specified:
"./configure --with-lapack="-L${MKLROOT}/lib/intel64 -Wl,--no-as-needed -lmkl_intel_lp64 -lmkl_gnu_thread -lmkl_core -lm"
When compiling IpOpt.

If I don't set OMP_NUM_THREADS env var, the solver works fine. But if I set OMP_NUM_THREADS=4 to use the parallel Pardiso, frequently the solver won't converge.

I tried both -lmkl_gnu_thread -lgomp for GNU openMP and -lmkl_intel_thread -liomp5 for Intel OpenMP. Both have same issues.

Any reasons for this?

Thanks.