The　application　of　GPU　to　accelerate　large-scale　smoke　simulation　is　a　hot　research　topic　in　computational　fluid　dynamics.　However,　the　current　smoke　parallel　computing　methods　for　different　scale　smoke　flow　field,　there　are　some　problems,　such　as　the　need　to　manually　adjust　some　algorithms,　task　allocation　does　not　consider　the　difference　between　the　nodes　computing　power.　In　order　to　meet　the　adaptive　requirements　of　smoke　parallel　computing,　a　distributed　smoke　simulation　system　with　adaptive　task　partitioning　is　designed.　The　computing　power　of　nodes　and　the　transmission　cost　of　data　exchange　are　fully　considered.　The　problem　of　boundary　data　synchronization　is　solved　and　the　correctness　of　parallel　computing　results　is　ensured.　　©　2023　IEEE.