[Mlir-commits] [mlir] [MLIR] Enable caching of type conversion in the presence of context-aware conversion (PR #158072)

[Matthias Springer]

================
@@ -285,9 +285,12 @@ conversions. A context-unaware conversion function converts a `Type` into a
 `Type`. A context-aware conversion function converts a `Value` into a type. The
 latter allows users to customize type conversion rules based on the IR.
 
-Note: When there is at least one context-aware type conversion function, the
-result of type conversions can no longer be cached, which can increase
-compilation time. Use this feature with caution!
+Note: context-aware type conversion function impact the ability of the
+framework to cache the conversin result: only context-free type conversions
+added after a context-aware type conversion can be cached (conversions are
----------------
matthias-springer wrote:

Maybe worth noting that caching is also possible if there is no context-aware type conversion at all.

https://github.com/llvm/llvm-project/pull/158072