[llvm-branch-commits] [flang] [mlir] [mlir][Transforms] Support 1:N mappings in `ConversionValueMapping` (PR #116524)

[Markus Böck via llvm-branch-commits]

================
@@ -1192,51 +1256,28 @@ LogicalResult ConversionPatternRewriterImpl::remapValues(
       });
       return failure();
     }
-
     // If a type is converted to 0 types, there is nothing to do.
     if (legalTypes.empty()) {
       remapped.push_back({});
       continue;
     }
 
-    if (legalTypes.size() != 1) {
-      // TODO: This is a 1:N conversion. The conversion value mapping does not
-      // store such materializations yet. If the types of the most recently
-      // mapped values do not match, build a target materialization.
-      ValueRange unpackedRange(unpacked);
-      if (TypeRange(unpackedRange) == legalTypes) {
-        remapped.push_back(std::move(unpacked));
-        continue;
-      }
-
-      // Insert a target materialization if the current pattern expects
-      // different legalized types.
-      ValueRange targetMat = buildUnresolvedMaterialization(
-          MaterializationKind::Target, computeInsertPoint(repl), operandLoc,
-          /*valueToMap=*/Value(), /*inputs=*/unpacked,
-          /*outputType=*/legalTypes, /*originalType=*/origType,
-          currentTypeConverter);
-      remapped.push_back(targetMat);
+    ValueVector repl = mapping.lookupOrDefault({operand}, legalTypes);
+    if (!repl.empty() && TypeRange(repl) == legalTypes) {
+      // Mapped values have the correct type or there is an existing
+      // materialization. Or the opreand is not mapped at all and has the
+      // correct type.
+      remapped.push_back(repl);
----------------
zero9178 wrote:

```suggestion
      remapped.push_back(std::move(repl));
```

https://github.com/llvm/llvm-project/pull/116524