The coming back to the original problem, I’m still unsure of how to diagnose
the problem.
I am not sure if the listserv supports image attachments, but I dumped the
snes_check_jacobian_view to disk and inspected the matrices with the spy
command. I find that no entry in the saved matrix has absolute value larger
than 10^{-2}. There are some that have absolute value larger than 10^{-3}, but
the actual values of the entries at those indices are on the order of 100 or
1000, meaning that they are small in a relative sense.
-gideon
> On Sep 9, 2015, at 12:19 PM, Jed Brown <[email protected]> wrote:
>
> Gideon Simpson <[email protected]> writes:
>
>> Checking the Jacobian, this is a pretty standard output:
>>
>> Testing hand-coded Jacobian, if the ratio is O(1.e-8), the hand-coded
>> Jacobian is probably correct.
>> Finite difference Jacobian
>> Hand-coded Jacobian
>> Hand-coded minus finite difference Jacobian
>> 6.08281e-10 = ||J - Jfd||//J|| 0.000151055 = ||J - Jfd||
>>
>>
>> Things look favorable in terms of relative difference, but the absolute
>> difference is a bit more suspect.
>
> It likely just means you have big matrix entries (perhaps due to poor
> nondimensionalization/scaling).
