This change is mostly for brevity's sake; but it also paves the way for the Policy enum to be reuseable for other situations that require the same three-way semantics.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
mlir/lib/Dialect/SparseTensor/IR/Detail/Var.h | ||
---|---|---|
393 | Since we (at least at the current moment) requires forward declaration (when needed) for the dim/lvl variables. Do we still need the Policy? Or do you have the plan to completely get rid of forward declaration in the future? |
mlir/lib/Dialect/SparseTensor/IR/Detail/Var.h | ||
---|---|---|
393 | The long-term plan is to get rid of the forward-decls (of level-vars; the symbol-vars will still be there, of course). In D155533 (DimLvlMapParser.cpp:181–207) I add a longer explanation of why the forward-decls cause problems and how we can avoid the need for the forward-decls. The original design of VarEnv, Policy, VarInfo/VarInfo::ID vs Var/Var::Num, etc was set up precisely to solve this issue without the need for forward-decls. |
Since we (at least at the current moment) requires forward declaration (when needed) for the dim/lvl variables. Do we still need the Policy?
Or do you have the plan to completely get rid of forward declaration in the future?