[Mlir-commits] [mlir] [mlir][LLVM] `FuncToLLVM`: Add 1:N type conversion support (PR #153823)

[Tobias Gysi]

================
@@ -565,34 +567,60 @@ struct CallOpInterfaceLowering : public ConvertOpToLLVMPattern<CallOpType> {
         static_cast<int32_t>(promoted.size()), 0};
     newOp.getProperties().op_bundle_sizes = rewriter.getDenseI32ArrayAttr({});
 
-    SmallVector<Value, 4> results;
-    if (numResults < 2) {
-      // If < 2 results, packing did not do anything and we can just return.
-      results.append(newOp.result_begin(), newOp.result_end());
-    } else {
-      // Otherwise, it had been converted to an operation producing a structure.
-      // Extract individual results from the structure and return them as list.
-      results.reserve(numResults);
-      for (unsigned i = 0; i < numResults; ++i) {
-        results.push_back(LLVM::ExtractValueOp::create(
-            rewriter, callOp.getLoc(), newOp->getResult(0), i));
+    // Helper function that extracts an individual result from the return value
+    // of the new call op. llvm.call ops support only 0 or 1 result. In case of
+    // 2 or more results, the results are packed into a structure.
----------------
gysit wrote:

```suggestion
    // 2 or more results, the results are packed into a structure.
```
maybe it is worth explaining that the type conversion can also lead to multiple results? I assume that is the reason why we cannot keep using numResults as before.

https://github.com/llvm/llvm-project/pull/153823