- #01 Inconsistent between results on Mac and Linux
- Actually it is a bug in the code, where I didn't keep the variable for
gsl_vector_view. Under clang, the complier keep it for me, GNU-gcc didn't
- Actually it is a bug in the code, where I didn't keep the variable for
- #02 Nan in intermeida nextPoints
- The nan is due to the minimization problem. I didn't check the convergence status.
- #03 Need to Accept extra parameters for potential and its derivatives
- #04 GSL insufficient number of points for interpolation error
- For number of points larger or equal to 3, use the gsl_interp_steffen (GSL >= 2.1)
- For number of points equal to 2, use the gsl_interp_linear
- For number of points equal to 1, no interpolation
- #05 Two phases have just one common point, how to treat such case in removeRedundantPhases
- For some situation, this case happens when two phases have tiny overlap in temperature due to the poor precision in determining the local minima around that temperature.
- At first place, I though this was the precision problem (the program didn't get the local minimum with sufficient precision). Thus I reconstruct the precision control parameter, put them into a structure which can be modified anywhere. However, I found that just checking the step-size after each step for for minimization convergence is really bad. Thus I change back to checking gradient.
- For the ISSUE05 itself, I need to get the test-case.
- For some situation, this case happens when two phases have tiny overlap in temperature due to the poor precision in determining the local minima around that temperature.
- #6 The precision in Tunneling1D when finding the profile. dphi/dr can have tremendously different scale from phi itself.