As　the　best　connection　model　for　the　multi-pin　net　of　the　non-Manhattan　architecture　global　routing　problem,　the　X-architecture　Steiner　Minimum　Tree　(XSMT)　construction　is　a　Non-deterministic　Polynomial　hard　(NP-hard)　problem.　The　Differential　Evolution　(DE)　algorithm　has　shown　good　application　effect　in　solving　various　NP-hard　problems.　For　this　reason,　based　on　the　idea　of　DE　algorithm,　this　paper　proposes　an　XSMT　construction　algorithm　for　solving　this　problem.　First　of　all,　because　the　traditional　DE　algorithm　is　designed　for　continuous　problems,　the　optimization　ability　is　limited　in　solving　discrete　problems.　This　paper　proposes　a　novel　crossover　operator　and　mutation　operator.　At　the　same　time,　in　order　to　maintain　the　effectiveness　of　the　evolutionary　algorithm,　an　Edge-to-Point　coding　strategy　suitable　for　evolutionary　algorithms　is　proposed　to　better　preserve　the　optimal　substructure　of　the　population.　Finally,　in　order　to　speed　up　the　convergence　speed　and　quality　of　the　algorithm,　this　paper　proposes　an　initial　solution　based　on　the　minimum　tree　generation　algorithm.　Experiments　show　that　the　effectiveness　of　the　proposed　algorithm　and　related　strategies　can　construct　a　high-quality　XSMT　solution.　©　Springer　Nature　Switzerland　AG,　2020.