search

mammasmias / IterativeRotationsAssignments

https://mammasmias.github.io/IterativeRotationsAssignments/
Other
8 stars 2 forks source link

ira_eq.x doesn't work. #1

Closed FanMover closed 2 years ago

FanMover commented 2 years ago

Hi , @mgoonde .

I am a student working on a project of Off-lattice Kinetic Monte Carlo. I found your paper and source code with excitement. As you mentioned in the paper(IRA: A shape matching approach for recognition and comparison of generic atomic patterns), in self-Learning KMC, atomic assignment is unknown amd number of atoms unset. Traditional optimal method like SVD and assignment method like Hungarian algorithm cannot work on my system. So I downloaded your code and tried.

However, there was something wrong after I combined two .xyz system and executed ira_eq.x.

First, variable _hdfin not defined. To be honest, I don't know what it means. So i changed hd_out = hd_fin to hd_out = hd image

Second, the way you permute the coordinates is not recognized by my machine, it goes image , whenerver there is a permutation operation. So I changed it from coords1(:,:) = coords1(:,nint(d_o(2,:))) to do i = 1, nat1 coords1(:,i) = coords1_tmp(:,nint(d_o(2,i))) end do So does the array typ1/typ2/coords2

After doing these rescue measures, I ran your example ./ira_eq.x < temp_ira. temp_ira is combined from _benchmark_test/data/ljclusters/xyz/47.xyz, same structure file in run_single.sh, and the randomized one. The two xyz structures didn't found scattered using ovito. The result goes infinity.

I don't know what is wrong and I am not that farmiliar with fortran. So I appreciate your help about it. Best wishes.

mgoonde commented 2 years ago

Hey @FanMover, I would be interested in your project, since we are actually doing the exact same thing (off-lattice kmc). We will have a release of the kmc quite soon, and it will be posted also in the description here. I would appreciate if you could write me a short email of what you do.

Regarding the code problem: