Soil　&　Rock　Aggregate　(SRA)　is　ubiquitous　in　nature　and　an　excellent　fill　material　in　engineering.　The　presence　of　SRA　rock　blocks　in　number　and　size-difference　makes　it　impossible　to　simulate　each　rock　block　separately　in　engineering　computations.　A　practical　way　is　to　regard　SRA　as　an　equivalent　continuum　and　to　evaluate　prop-erties　of　the　equivalent　continuum　using　numerical　methods.　In　this　way,　it　is　hopeful　to　overcome　big　errors　of　laboratory　tests　due　to　removing　big　blocks.　Randomness　of　SRA　rock　blocks　in　shape　and　orientation　renders　clumsy　mesh　dependent　methods　such　as　finite　element　methods.　By　using　finite　elements　to　form　mathematical　patches,　the　numerical　manifold　method　(NMM),　called　FE-NMM,　reduces　to　some　degree　mesh　dependency,　but　still　involves　a　huge　number　of　operations　in　cover-cutting　and　information-rebuilding　while　generating　physical　covers.　Specifying　node　influence　domains　in　the　moving　least　squares　(MLS)　as　mathematical　patches,　and　MLS　shape　functions　as　the　weight　functions　of　the　physical　patches,　this　study　develops　the　procedure　MLS-NMM　to　evaluate　SRA　permeability,　completely　avoiding　extra　labor　in　FE-NMM.　Many　complicated　deployments　of　rock　blocks　in　SRA　are　designed　and　analyzed,　suggesting　significant　advantages　of　MLS-NMM　over　FE-NMM.　Meanwhile,　some　interesting　results　are　observed.