diff --git a/src/wasm/baseline/liftoff-assembler.cc b/src/wasm/baseline/liftoff-assembler.cc
index 0c2dad6cb2..6c00b9a374 100644
--- a/src/wasm/baseline/liftoff-assembler.cc
+++ b/src/wasm/baseline/liftoff-assembler.cc
@@ -11,6 +11,7 @@
 #include "src/compiler/wasm-compiler.h"
 #include "src/macro-assembler-inl.h"
 #include "src/wasm/function-body-decoder-impl.h"
+#include "src/wasm/wasm-linkage.h"
 #include "src/wasm/wasm-opcodes.h"
 
 namespace v8 {
@@ -620,6 +621,22 @@ void LiftoffAssembler::ParallelRegisterMove(
   }
 }
 
+void LiftoffAssembler::MoveToReturnRegisters(FunctionSig* sig) {
+  // We do not support multi-value yet.
+  DCHECK_EQ(1, sig->return_count());
+  ValueType return_type = sig->GetReturn(0);
+  StackTransferRecipe stack_transfers(this);
+  LiftoffRegister return_reg =
+      needs_reg_pair(return_type)
+          ? LiftoffRegister::ForPair(kGpReturnRegisters[0],
+                                     kGpReturnRegisters[1])
+          : reg_class_for(return_type) == kGpReg
+                ? LiftoffRegister(kGpReturnRegisters[0])
+                : LiftoffRegister(kFpReturnRegisters[0]);
+  stack_transfers.LoadIntoRegister(return_reg, cache_state_.stack_state.back(),
+                                   cache_state_.stack_height() - 1);
+}
+
 #ifdef ENABLE_SLOW_DCHECKS
 bool LiftoffAssembler::ValidateCacheState() const {
   uint32_t register_use_count[kAfterMaxLiftoffRegCode] = {0};
diff --git a/src/wasm/baseline/liftoff-assembler.h b/src/wasm/baseline/liftoff-assembler.h
index 6476f0c22d..64a2d8c25b 100644
--- a/src/wasm/baseline/liftoff-assembler.h
+++ b/src/wasm/baseline/liftoff-assembler.h
@@ -338,6 +338,8 @@ class LiftoffAssembler : public TurboAssembler {
   };
   void ParallelRegisterMove(Vector<ParallelRegisterMoveTuple>);
 
+  void MoveToReturnRegisters(FunctionSig*);
+
 #ifdef ENABLE_SLOW_DCHECKS
   // Validate that the register use counts reflect the state of the cache.
   bool ValidateCacheState() const;
diff --git a/src/wasm/baseline/liftoff-compiler.cc b/src/wasm/baseline/liftoff-compiler.cc
index 49c5ab19d4..26484ac9e4 100644
--- a/src/wasm/baseline/liftoff-compiler.cc
+++ b/src/wasm/baseline/liftoff-compiler.cc
@@ -1075,25 +1075,17 @@ class LiftoffCompiler {
     __ cache_state()->stack_state.pop_back();
   }
 
-  void DoReturn(FullDecoder* decoder, Vector<Value> values, bool implicit) {
+  void DoReturn(FullDecoder* decoder, Vector<Value> /*values*/, bool implicit) {
     if (implicit) {
       DCHECK_EQ(1, decoder->control_depth());
       Control* func_block = decoder->control_at(0);
       __ bind(func_block->label.get());
       __ cache_state()->Steal(func_block->label_state);
+      TraceCacheState(decoder);
     }
-    if (!values.is_empty()) {
-      if (values.size() > 1) return unsupported(decoder, "multi-return");
-      LiftoffRegister reg = __ PopToRegister();
-      LiftoffRegister return_reg =
-          kNeedI64RegPair && values[0].type == kWasmI64
-              ? LiftoffRegister::ForPair(kGpReturnRegisters[0],
-                                         kGpReturnRegisters[1])
-              : reg_class_for(values[0].type) == kGpReg
-                    ? LiftoffRegister(kGpReturnRegisters[0])
-                    : LiftoffRegister(kFpReturnRegisters[0]);
-      if (reg != return_reg) __ Move(return_reg, reg, values[0].type);
-    }
+    size_t num_returns = decoder->sig_->return_count();
+    if (num_returns > 1) return unsupported(decoder, "multi-return");
+    if (num_returns > 0) __ MoveToReturnRegisters(decoder->sig_);
     __ LeaveFrame(StackFrame::WASM_COMPILED);
     __ DropStackSlotsAndRet(
         static_cast<uint32_t>(descriptor_->StackParameterCount()));