in the first phase AST (Type I) nmigen abstraction was achieved (see https://libre-soc.org/3d_gpu/architecture/dynamic_simd/) which left DSL (Type II, high-level language constructs) to complete. The work required is very clear but complex. The first phase of Dynamic Partitioned SIMD was completed in previous grants: the second phase is much more complex and much more powerful, extending nmigen so that adding other language-filters is possible (strong typechecking, width enforcement just like VHDL, complex number types etc).
https://groups.google.com/g/comp.arch/c/mpn4OPNkyoA/m/0d4SEqNqAgAJ