This provides a significantly more efficient encoding for locations.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
| mlir/lib/IR/BuiltinDialectBytecode.cpp | ||
|---|---|---|
| 405 | I wonder why you split the writer/reader? | |
| mlir/lib/IR/BuiltinDialectBytecode.cpp | ||
|---|---|---|
| 405 | It was just easier for me to reason about in the beginning. I don't have a huge preference and could move them next to each other, let me know if you want that I'll move them in a follow up. | |
| mlir/lib/IR/BuiltinDialectBytecode.cpp | ||
|---|---|---|
| 405 | Take it as a suggestion for a future revision :) | |
| mlir/lib/IR/BuiltinDialectBytecode.cpp | ||
|---|---|---|
| 379 | Should there be an error emitted about failure here? (I don't know where the errors are reported from or how we ensure at least an error is reported) | |
| mlir/lib/IR/BuiltinDialectBytecode.cpp | ||
|---|---|---|
| 379 | All of the API provided by the DialectBytecodeReader/DialectBytecodeWriter call into the main bytecode API, which will emit errors on failure (so we don't emit here to avoid double-emit). | |
Should there be an error emitted about failure here? (I don't know where the errors are reported from or how we ensure at least an error is reported)