Maybe i found a bug
Ln 1654 in JSphGpu_ker.cu
here limit the max shift distance for particles. With z direction, shift distance in x direction is used to comparision, which maybe wrong.
And i noticed that the shifting aglorithm are different in singlephase code and twophase code. In singlephase code, the aglorithm is consistent with guide.
But i found it in twophase code is
which is quite different, the calculation of gradient of concentration. in singlephase, it needs only one summation, and needs two in twophase code. It means the efficiency is lower in twophase. so, anyone know the difference of accuarcy between them.