[Mlir-commits] [mlir] [mlir][ArmSME] Use ArmSMETypeConverter for all VectorToLLVM patterns (PR #65261)

[Markus Böck]

================
@@ -83,21 +83,26 @@ void LowerVectorToLLVMPass::runOnOperation() {
   // Convert to the LLVM IR dialect.
   LowerToLLVMOptions options(&getContext());
   options.useOpaquePointers = useOpaquePointers;
-  LLVMTypeConverter converter(&getContext(), options);
+
+  LLVMTypeConverter *converter;
+  if (armSME)
+    converter = new arm_sme::ArmSMETypeConverter(&getContext(), options);
+  else
+    converter = new LLVMTypeConverter(&getContext(), options);
----------------
zero9178 wrote:

This is a memory leak. You should at the very least use `std::unique_ptr<LLVMTypeConverter>` here and use `make_unique` in the logic.

https://github.com/llvm/llvm-project/pull/65261