diff --git a/compiler/Android.mk b/compiler/Android.mk
index 84176a1..70c7e52 100644
--- a/compiler/Android.mk
+++ b/compiler/Android.mk
@@ -279,7 +279,11 @@
   LOCAL_ADDITIONAL_DEPENDENCIES := art/build/Android.common_build.mk
   LOCAL_ADDITIONAL_DEPENDENCIES += $(LOCAL_PATH)/Android.mk
   # Vixl assembly support for ARM64 targets.
-  LOCAL_SHARED_LIBRARIES += libvixl
+  ifeq ($$(art_ndebug_or_debug),debug)
+    LOCAL_SHARED_LIBRARIES += libvixld
+  else
+    LOCAL_SHARED_LIBRARIES += libvixl
+  endif
   ifeq ($$(art_target_or_host),target)
     # For atrace.
     LOCAL_SHARED_LIBRARIES += libcutils
diff --git a/compiler/oat_writer.cc b/compiler/oat_writer.cc
index a57f892..8a7abb4 100644
--- a/compiler/oat_writer.cc
+++ b/compiler/oat_writer.cc
@@ -366,6 +366,8 @@
     Offset offset(mirror::ArtMethod::EntryPointFromQuickCompiledCodeOffset(
         kArm64PointerSize).Int32Value());
     assembler.JumpTo(ManagedRegister(arm64::X0), offset, ManagedRegister(arm64::IP0));
+    // Ensure we emit the literal pool.
+    assembler.EmitSlowPaths();
     std::vector<uint8_t> thunk_code(assembler.CodeSize());
     MemoryRegion code(thunk_code.data(), thunk_code.size());
     assembler.FinalizeInstructions(code);
diff --git a/compiler/optimizing/code_generator.cc b/compiler/optimizing/code_generator.cc
index 6cacd4f..e581af2 100644
--- a/compiler/optimizing/code_generator.cc
+++ b/compiler/optimizing/code_generator.cc
@@ -71,11 +71,7 @@
     }
   }
   GenerateSlowPaths();
-
-  size_t code_size = GetAssembler()->CodeSize();
-  uint8_t* buffer = allocator->Allocate(code_size);
-  MemoryRegion code(buffer, code_size);
-  GetAssembler()->FinalizeInstructions(code);
+  Finalize(allocator);
 }
 
 void CodeGenerator::CompileOptimized(CodeAllocator* allocator) {
@@ -97,9 +93,13 @@
     }
   }
   GenerateSlowPaths();
+  Finalize(allocator);
+}
 
+void CodeGenerator::Finalize(CodeAllocator* allocator) {
   size_t code_size = GetAssembler()->CodeSize();
   uint8_t* buffer = allocator->Allocate(code_size);
+
   MemoryRegion code(buffer, code_size);
   GetAssembler()->FinalizeInstructions(code);
 }
diff --git a/compiler/optimizing/code_generator.h b/compiler/optimizing/code_generator.h
index 321a31f..4c0d3ea 100644
--- a/compiler/optimizing/code_generator.h
+++ b/compiler/optimizing/code_generator.h
@@ -90,6 +90,7 @@
   }
 
   virtual void Initialize() = 0;
+  virtual void Finalize(CodeAllocator* allocator);
   virtual void GenerateFrameEntry() = 0;
   virtual void GenerateFrameExit() = 0;
   virtual void Bind(HBasicBlock* block) = 0;
diff --git a/compiler/optimizing/code_generator_arm64.cc b/compiler/optimizing/code_generator_arm64.cc
index 0fc4307..8d43a5d 100644
--- a/compiler/optimizing/code_generator_arm64.cc
+++ b/compiler/optimizing/code_generator_arm64.cc
@@ -479,6 +479,12 @@
 #undef __
 #define __ GetVIXLAssembler()->
 
+void CodeGeneratorARM64::Finalize(CodeAllocator* allocator) {
+  // Ensure we emit the literal pool.
+  __ FinalizeCode();
+  CodeGenerator::Finalize(allocator);
+}
+
 void CodeGeneratorARM64::GenerateFrameEntry() {
   // TODO: Add proper support for the stack overflow check.
   UseScratchRegisterScope temps(GetVIXLAssembler());
diff --git a/compiler/optimizing/code_generator_arm64.h b/compiler/optimizing/code_generator_arm64.h
index a40f27f..236a04d 100644
--- a/compiler/optimizing/code_generator_arm64.h
+++ b/compiler/optimizing/code_generator_arm64.h
@@ -232,6 +232,8 @@
     }
   }
 
+  void Finalize(CodeAllocator* allocator) OVERRIDE;
+
   // Code generation helpers.
   void MoveConstant(vixl::CPURegister destination, HConstant* constant);
   void MoveHelper(Location destination, Location source, Primitive::Type type);
diff --git a/compiler/trampolines/trampoline_compiler.cc b/compiler/trampolines/trampoline_compiler.cc
index 733b58f..cb07ffa 100644
--- a/compiler/trampolines/trampoline_compiler.cc
+++ b/compiler/trampolines/trampoline_compiler.cc
@@ -83,6 +83,7 @@
       break;
   }
 
+  assembler->EmitSlowPaths();
   size_t cs = assembler->CodeSize();
   std::unique_ptr<std::vector<uint8_t>> entry_stub(new std::vector<uint8_t>(cs));
   MemoryRegion code(&(*entry_stub)[0], entry_stub->size());
diff --git a/compiler/utils/arm64/assembler_arm64.cc b/compiler/utils/arm64/assembler_arm64.cc
index 390f2ea..21014c8 100644
--- a/compiler/utils/arm64/assembler_arm64.cc
+++ b/compiler/utils/arm64/assembler_arm64.cc
@@ -329,12 +329,12 @@
     if (dst.IsXRegister()) {
       if (size == 4) {
         CHECK(src.IsWRegister());
-        ___ Mov(reg_x(dst.AsXRegister()), reg_w(src.AsWRegister()));
+        ___ Mov(reg_w(dst.AsOverlappingWRegister()), reg_w(src.AsWRegister()));
       } else {
         if (src.IsXRegister()) {
           ___ Mov(reg_x(dst.AsXRegister()), reg_x(src.AsXRegister()));
         } else {
-          ___ Mov(reg_x(dst.AsXRegister()), reg_w(src.AsWRegister()));
+          ___ Mov(reg_x(dst.AsXRegister()), reg_x(src.AsOverlappingXRegister()));
         }
       }
     } else if (dst.IsWRegister()) {
@@ -484,9 +484,9 @@
   CHECK(size == 1 || size == 2) << size;
   CHECK(reg.IsWRegister()) << reg;
   if (size == 1) {
-    ___ sxtb(reg_w(reg.AsWRegister()), reg_w(reg.AsWRegister()));
+    ___ Sxtb(reg_w(reg.AsWRegister()), reg_w(reg.AsWRegister()));
   } else {
-    ___ sxth(reg_w(reg.AsWRegister()), reg_w(reg.AsWRegister()));
+    ___ Sxth(reg_w(reg.AsWRegister()), reg_w(reg.AsWRegister()));
   }
 }
 
@@ -495,9 +495,9 @@
   CHECK(size == 1 || size == 2) << size;
   CHECK(reg.IsWRegister()) << reg;
   if (size == 1) {
-    ___ uxtb(reg_w(reg.AsWRegister()), reg_w(reg.AsWRegister()));
+    ___ Uxtb(reg_w(reg.AsWRegister()), reg_w(reg.AsWRegister()));
   } else {
-    ___ uxth(reg_w(reg.AsWRegister()), reg_w(reg.AsWRegister()));
+    ___ Uxth(reg_w(reg.AsWRegister()), reg_w(reg.AsWRegister()));
   }
 }