diff --git a/src/dynarmic/backend/arm64/a32_address_space.cpp b/src/dynarmic/backend/arm64/a32_address_space.cpp
index a8f89a83..686a45d5 100644
--- a/src/dynarmic/backend/arm64/a32_address_space.cpp
+++ b/src/dynarmic/backend/arm64/a32_address_space.cpp
@@ -166,6 +166,27 @@ void A32AddressSpace::EmitPrelude() {
 
     code.BR(X0);
 
+    prelude_info.step_code = code.ptr<PreludeInfo::RunCodeFuncType>();
+    ABI_PushRegisters(code, ABI_CALLEE_SAVE | (1 << 30), sizeof(StackLayout));
+
+    code.MOV(Xstate, X1);
+    code.MOV(Xhalt, X2);
+
+    code.LDR(Wscratch0, Xstate, offsetof(A32JitState, upper_location_descriptor));
+    code.AND(Wscratch0, Wscratch0, 0xffff0000);
+    code.MRS(Xscratch1, oaknut::SystemReg::FPCR);
+    code.STR(Wscratch1, SP, offsetof(StackLayout, save_host_fpcr));
+    code.MSR(oaknut::SystemReg::FPCR, Xscratch0);
+
+    oaknut::Label step_hr_loop;
+    code.l(step_hr_loop);
+    code.LDAXR(Wscratch0, Xhalt);
+    code.ORR(Wscratch0, Wscratch0, static_cast<u32>(HaltReason::Step));
+    code.STLXR(Wscratch1, Wscratch0, Xhalt);
+    code.CBNZ(Wscratch1, step_hr_loop);
+
+    code.BR(X0);
+
     prelude_info.return_from_run_code = code.ptr<void*>();
 
     code.LDR(Wscratch0, SP, offsetof(StackLayout, save_host_fpcr));
diff --git a/src/dynarmic/backend/arm64/a32_core.h b/src/dynarmic/backend/arm64/a32_core.h
index 10ec3d7b..86f6f284 100644
--- a/src/dynarmic/backend/arm64/a32_core.h
+++ b/src/dynarmic/backend/arm64/a32_core.h
@@ -19,6 +19,12 @@ public:
         const auto entry_point = process.GetOrEmit(location_descriptor);
         return process.prelude_info.run_code(entry_point, &thread_ctx, halt_reason);
     }
+
+    HaltReason Step(A32AddressSpace& process, A32JitState& thread_ctx, volatile u32* halt_reason) {
+        const auto location_descriptor = A32::LocationDescriptor{thread_ctx.GetLocationDescriptor()}.SetSingleStepping(true);
+        const auto entry_point = process.GetOrEmit(location_descriptor);
+        return process.prelude_info.step_code(entry_point, &thread_ctx, halt_reason);
+    }
 };
 
 }  // namespace Dynarmic::Backend::Arm64
diff --git a/src/dynarmic/backend/arm64/a32_interface.cpp b/src/dynarmic/backend/arm64/a32_interface.cpp
index 74ed9cdc..69e65abb 100644
--- a/src/dynarmic/backend/arm64/a32_interface.cpp
+++ b/src/dynarmic/backend/arm64/a32_interface.cpp
@@ -79,11 +79,11 @@ struct Jit::Impl final {
             jit_interface->is_executing = false;
         };
 
-        ASSERT_FALSE("Unimplemented");
+        HaltReason hr = core.Step(current_address_space, current_state, &halt_reason);
 
         PerformRequestedCacheInvalidation();
 
-        return HaltReason{};
+        return hr;
     }
 
     void ClearCache() {
diff --git a/src/dynarmic/backend/arm64/emit_arm64_a32.cpp b/src/dynarmic/backend/arm64/emit_arm64_a32.cpp
index 5d0ce01e..415e9b55 100644
--- a/src/dynarmic/backend/arm64/emit_arm64_a32.cpp
+++ b/src/dynarmic/backend/arm64/emit_arm64_a32.cpp
@@ -61,8 +61,8 @@ void EmitSetUpperLocationDescriptor(oaknut::CodeGenerator& code, EmitContext& ct
 void EmitA32Terminal(oaknut::CodeGenerator& code, EmitContext& ctx, IR::Term::LinkBlock terminal, IR::LocationDescriptor initial_location, bool) {
     EmitSetUpperLocationDescriptor(code, ctx, terminal.next, initial_location);
 
-    code.MOV(Xscratch0, terminal.next.Value());
-    code.STUR(Xscratch0, Xstate, offsetof(A32JitState, regs) + sizeof(u32) * 15);
+    code.MOV(Wscratch0, A32::LocationDescriptor{terminal.next}.PC());
+    code.STR(Wscratch0, Xstate, offsetof(A32JitState, regs) + sizeof(u32) * 15);
     EmitRelocation(code, ctx, LinkTarget::ReturnFromRunCode);
 
     // TODO: Implement LinkBlock optimization
@@ -71,8 +71,8 @@ void EmitA32Terminal(oaknut::CodeGenerator& code, EmitContext& ctx, IR::Term::Li
 void EmitA32Terminal(oaknut::CodeGenerator& code, EmitContext& ctx, IR::Term::LinkBlockFast terminal, IR::LocationDescriptor initial_location, bool) {
     EmitSetUpperLocationDescriptor(code, ctx, terminal.next, initial_location);
 
-    code.MOV(Xscratch0, terminal.next.Value());
-    code.STUR(Xscratch0, Xstate, offsetof(A32JitState, regs) + sizeof(u32) * 15);
+    code.MOV(Wscratch0, A32::LocationDescriptor{terminal.next}.PC());
+    code.STR(Wscratch0, Xstate, offsetof(A32JitState, regs) + sizeof(u32) * 15);
     EmitRelocation(code, ctx, LinkTarget::ReturnFromRunCode);
 
     // TODO: Implement LinkBlockFast optimization