srdatalog.ir.dialects¶
Registered dialects, organized by category.
Folder layout mirrors the dotted dialect-name convention used in docs/ir_lowering_semantics.md. Each category subdirectory groups sibling dialects of the same kind:
relation/ — data-structure dialects (sorted_array, lsm, uf, …) iir/ — index-aware IR cross-cutting (cf, future: mem) target/ — codegen backends (cuda, future: hip, cpp_tbb, …) par/ — parallelism strategies (future: data.warp_strided, data.tbb_for, task, simd, scalar)
Adding a new dialect: pick the right category, drop a subpackage in. No edits to other categories or to ir_core needed (Property P1 in design_principles.md).