The extent of structural variants (SVs), in particular copy number variations (CNVs) in plant and animal genomes remains unknown, mainly due to the lack of population-scale truly high-quality genomes. Here, we de novo assembled and annotated 31 gold-standard reference genomes from varieties representing all major cultivated rice subtypes, and accurately detected 107,251 non-redundant SVs affecting 644.91Mbp. A position-resolved pan-genome comprising 66,636 genes enabled our discovery that more than 38% of protein-coding genes in cultivated rice harbor CNVs, a far greater proportion than previous estimate. Illustrating functional consequences of these variations, CNVs of Awn3-1, OsVIL1 and OsMADS18, as well as a 345kb inversion likely contributed to major events in rice evolution, domestication and environmental adaptation. Beyond fully resolving the SVs of cultivated rice and comprehensively cataloguing CNVs among protein-coding genes, our study suggests that SVs-affected genes likely contribute to many mechanisms underlying domestication and phenotypic variation in rice.