diff --git a/src/hotspot/cpu/arm/interp_masm_arm.cpp b/src/hotspot/cpu/arm/interp_masm_arm.cpp
index 481c3d09e58..635acd781f9 100644
--- a/src/hotspot/cpu/arm/interp_masm_arm.cpp
+++ b/src/hotspot/cpu/arm/interp_masm_arm.cpp
@@ -303,15 +303,19 @@ void InterpreterMacroAssembler::load_field_entry(Register cache, Register index,
 }
 
 void InterpreterMacroAssembler::load_method_entry(Register cache, Register index, int bcp_offset) {
+  assert_different_registers(cache, index);
+
   // Get index out of bytecode pointer
   get_index_at_bcp(index, bcp_offset, cache /* as tmp */, sizeof(u2));
+
+  // sizeof(ResolvedMethodEntry) is not a power of 2 on Arm, so can't use shift
   mov(cache, sizeof(ResolvedMethodEntry));
   mul(index, index, cache); // Scale the index to be the entry index * sizeof(ResolvedMethodEntry)
 
   // load constant pool cache pointer
   ldr(cache, Address(FP, frame::interpreter_frame_cache_offset * wordSize));
   // Get address of method entries array
-  ldr(cache, Address(cache, ConstantPoolCache::method_entries_offset()));
+  ldr(cache, Address(cache, in_bytes(ConstantPoolCache::method_entries_offset())));
   add(cache, cache, Array<ResolvedMethodEntry>::base_offset_in_bytes());
   add(cache, cache, index);
 }
diff --git a/src/hotspot/cpu/arm/templateInterpreterGenerator_arm.cpp b/src/hotspot/cpu/arm/templateInterpreterGenerator_arm.cpp
index ba9a3fd7a9b..efaf78ee568 100644
--- a/src/hotspot/cpu/arm/templateInterpreterGenerator_arm.cpp
+++ b/src/hotspot/cpu/arm/templateInterpreterGenerator_arm.cpp
@@ -370,17 +370,16 @@ address TemplateInterpreterGenerator::generate_return_entry_for(TosState state,
   if (index_size == sizeof(u4)) {
     __ load_resolved_indy_entry(Rcache, Rindex);
     __ ldrh(Rcache, Address(Rcache, in_bytes(ResolvedIndyEntry::num_parameters_offset())));
-    __ check_stack_top();
-    __ add(Rstack_top, Rstack_top, AsmOperand(Rcache, lsl, Interpreter::logStackElementSize));
   } else {
     // Pop N words from the stack
     assert(index_size == sizeof(u2), "Can only be u2");
     __ load_method_entry(Rcache, Rindex);
-    __ ldrh(Rcache, Address(Rcache, in_bytes(ResolvedIndyEntry::num_parameters_offset())));
-    __ check_stack_top();
-    __ add(Rstack_top, Rstack_top, AsmOperand(Rcache, lsl, Interpreter::logStackElementSize));
+    __ ldrh(Rcache, Address(Rcache, in_bytes(ResolvedMethodEntry::num_parameters_offset())));
   }
 
+  __ check_stack_top();
+  __ add(Rstack_top, Rstack_top, AsmOperand(Rcache, lsl, Interpreter::logStackElementSize));
+
   __ convert_retval_to_tos(state);
 
   __ check_and_handle_popframe();
diff --git a/src/hotspot/cpu/arm/templateTable_arm.cpp b/src/hotspot/cpu/arm/templateTable_arm.cpp
index e478f08c977..55a0120e7c7 100644
--- a/src/hotspot/cpu/arm/templateTable_arm.cpp
+++ b/src/hotspot/cpu/arm/templateTable_arm.cpp
@@ -3666,15 +3666,15 @@ void TemplateTable::prepare_invoke(Register Rcache, Register recv) {
   // load receiver if needed (after extra argument is pushed so parameter size is correct)
   if (load_receiver) {
     __ ldrh(recv, Address(Rcache, in_bytes(ResolvedMethodEntry::num_parameters_offset())));
-    Address recv_addr = __ receiver_argument_address(Rstack_top, Rtemp, recv);
-    __ ldr(recv, recv_addr);
+    __ add(recv, Rstack_top, AsmOperand(recv, lsl, Interpreter::logStackElementSize));
+    __ ldr(recv, Address(recv, -Interpreter::stackElementSize));
     __ verify_oop(recv);
   }
 
   // load return address
   { const address table = (address) Interpreter::invoke_return_entry_table_for(code);
-    __ mov_slow(Rtemp, table);
-    __ ldr(LR, Address::indexed_ptr(Rtemp, ret_type));
+    __ mov_slow(LR, table);
+    __ ldr(LR, Address::indexed_ptr(LR, ret_type));
   }
 }
 
@@ -3744,10 +3744,13 @@ void TemplateTable::invokevirtual(int byte_no) {
 void TemplateTable::invokespecial(int byte_no) {
   transition(vtos, vtos);
   assert(byte_no == f1_byte, "use this argument");
+
   const Register Rrecv  = R2_tmp;
-  load_resolved_method_entry_special_or_static(R2_tmp,  // ResolvedMethodEntry*
+  const Register Rflags = R3_tmp;
+
+  load_resolved_method_entry_special_or_static(Rrecv,  // ResolvedMethodEntry*
                                                Rmethod, // Method*
-                                               R3_tmp); // Flags
+                                               Rflags); // Flags
   prepare_invoke(Rrecv, Rrecv);
   __ verify_oop(Rrecv);
   __ null_check(Rrecv, Rtemp);
@@ -3760,12 +3763,16 @@ void TemplateTable::invokespecial(int byte_no) {
 void TemplateTable::invokestatic(int byte_no) {
   transition(vtos, vtos);
   assert(byte_no == f1_byte, "use this argument");
-  load_resolved_method_entry_special_or_static(R2_tmp,  // ResolvedMethodEntry*
+
+  const Register Rrecv  = R2_tmp;
+  const Register Rflags = R3_tmp;
+
+  load_resolved_method_entry_special_or_static(Rrecv,  // ResolvedMethodEntry*
                                                Rmethod, // Method*
-                                               R3_tmp); // Flags
-  prepare_invoke(R2_tmp, R2_tmp);
+                                               Rflags); // Flags
+  prepare_invoke(Rrecv, Rrecv);
   // do the call
-  __ profile_call(R2_tmp);
+  __ profile_call(Rrecv);
   __ jump_from_interpreted(Rmethod);
 }
 
@@ -3788,10 +3795,10 @@ void TemplateTable::invokeinterface(int byte_no) {
   const Register Rflags  = R3_tmp;
   const Register Rklass  = R2_tmp; // Note! Same register with Rrecv
 
-  load_resolved_method_entry_interface(R2_tmp,  // ResolvedMethodEntry*
-                                       R1_tmp,  // Klass*
+  load_resolved_method_entry_interface(Rrecv,   // ResolvedMethodEntry*
+                                       Rinterf, // Klass*
                                        Rmethod, // Method* or itable/vtable index
-                                       R3_tmp); // Flags
+                                       Rflags); // Flags
   prepare_invoke(Rrecv, Rrecv);
 
   // First check for Object case, then private interface method,