Abstract: In this paper, a new sorting method which accords bucket number of bytecode to distribution and linking is put forward, and its sorting algorithm,its flowchartsand its experimental results in C-language are also given. The algorithmic analysis andexperimental results show that the time eomplexity of this method is O（N） and it hasnothing to do with data distribution. Addi-nonal memory cost is （N+512）ε （εis thesize of needed space of linking field）1. The method is not only obviously quicker thanthat of quick sort, but also much quicker than the bucket sorting method,even in thecase of non-uniformly distributed data.