The MKL reference manual does not say much about eps to be specified for Jacobian calculation.
The examples such as ex_nlsqp_c.c etc actually specify the same eps (an array of 6 elements) for dtrnlsp_init() to djacobi().
Which element of eps is used for the Jacobian calculation?
Is it eps (The trial step precision.)?
"If eps(6) = 0, then the trial step meets the required precision (≤ 1.0D-10)." is too hard for me to understand.
I'm sure djacobi/djacobix internally allocate the arrays(2 arrays?) holding function values each time they are called(and free them on exit). I do not think this is a good idea...