Nuclear shell-model code for massive parallel computation, "KSHELL"

A new code for nuclear shell-model calculations, "KSHELL", is developed. It aims at carrying out both massively parallel computation and single-node computation in the same manner. We solve the Schr\"{o}dinger's equation in the $M$-scheme shell-model model space, utilizing Thick-Restart Lanczos method. During the Lanczos iteration, the whole Hamiltonian matrix elements are generated "on-the-fly" in every matrix-vector multiplication. The vectors of the Lanczos method are distributed and stored on memory of each parallel node. We report that the newly developed code has high parallel efficiency on FX10 supercomputer and a PC with multi-cores.