adj_reset affects subsequent solutions

[Evan Cummings]

Hello,

I've recently discovered that if I call the function adj_reset, later solutions derived by the solve method are affected. I have compared all variables affecting the solution before and after the call to adj_reset, and they appear to be identical.

Is there some other reason that a call to adj_reset would affect the solution process later on?

[Simon Funke]

I'm am not sure what you mean with later solutions derived by the solve method are affected.

adj_reset deletes the tape and intermediate variables that are stored by dolfin-adjoint. It should not have an impact on the solutions of normal dolfin function. Is this what you mean? In that case, it is probably a bug and worth making a small example so that I can try and fix it.

[Evan Cummings]

Yes, you understood my meaning. It was my understanding that adj_reset should not affect dolfin solve calls, but this is what I have experienced. I'll see if I can reproduce it in a small example.