aartbik wrote: Just a random note, @fabianmcg , that I really appreciate your refactoring. The GPU "pipeline" for the sparse compiler still had a few rough edges and you really smoothed these out! So, thanks! https://github.com/llvm/llvm-project/pull/66220