Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
mlir/include/mlir/Dialect/SparseTensor/IR/Enums.h | ||
---|---|---|
164 | Why do we want to the double quote in the string? |
mlir/include/mlir/Dialect/SparseTensor/IR/Enums.h | ||
---|---|---|
161 | Rename to toMLIRString or similar? To make it clear that the string is à la MLIR surface syntax | |
161 | Is this really worth inlining? or should the definition be moved to a cpp file instead? | |
mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp | ||
163 | You probably want to assert that the string is nonempty, just in case. Also, you should keep the todo comment (unless it's been resolved by deciding that we shouldn't raise an error for undef). |
mlir/include/mlir/Dialect/SparseTensor/IR/Enums.h | ||
---|---|---|
161 | This way, we don't have dependences on a cpp file, which seemed nice given the nature of the Enums.h being pretty independent | |
161 | that makes sense, used toMLIRString | |
164 | I am neutral, but this way it matched the print/parser values. In the long run, when those are no longer strings we can change this one too | |
mlir/lib/Dialect/SparseTensor/IR/SparseTensorDialect.cpp | ||
163 | It feels like having a undef value is something that should be verified elsewhere, not while printing. The nonempty string will never be reached with your unreachable follow up, so not really worth asserting now, right? |
Rename to toMLIRString or similar? To make it clear that the string is à la MLIR surface syntax