[turbofan] Fewer gap moves for tail calls

src/compiler/x64/code-generator-x64.cc

 // Copyright 2013 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "src/compiler/code-generator.h"
 
 #include "src/ast/scopes.h"
 #include "src/compiler/code-generator-impl.h"
 #include "src/compiler/gap-resolver.h"
 #include "src/compiler/node-matchers.h"
@@ skipping 560 matching lines @@
       ASSEMBLE_CHECKED_STORE_INTEGER_IMPL(asm_instr, Immediate); \
     }                                                            \
   } while (false)
 
 
 void CodeGenerator::AssembleDeconstructActivationRecord(int stack_param_delta) {
   int sp_slot_delta = TailCallFrameStackSlotDelta(stack_param_delta);
   if (sp_slot_delta > 0) {
     __ addq(rsp, Immediate(sp_slot_delta * kPointerSize));
   }
-  if (frame()->needs_frame()) {
-    __ popq(rbp);
-  }
   frame_access_state()->SetFrameAccessToDefault();
 }
 
 
 void CodeGenerator::AssemblePrepareTailCall(int stack_param_delta) {
   int sp_slot_delta = TailCallFrameStackSlotDelta(stack_param_delta);
   if (sp_slot_delta < 0) {
     __ subq(rsp, Immediate(-sp_slot_delta * kPointerSize));
     frame_access_state()->IncreaseSPDelta(-sp_slot_delta);
   }
+  if (frame()->needs_frame()) {
+    __ movq(rbp, MemOperand(rbp, 0));
+  }
   frame_access_state()->SetFrameAccessToSP();
 }
 
 
 // Assembles an instruction after register allocation, producing machine code.
 void CodeGenerator::AssembleArchInstruction(Instruction* instr) {
   X64OperandConverter i(this, instr);
 
   switch (ArchOpcodeField::decode(instr->opcode())) {
     case kArchCallCodeObject: {
@@ skipping 1443 matching lines @@
     int padding_size = last_lazy_deopt_pc_ + space_needed - current_pc;
     __ Nop(padding_size);
   }
 }
 
 #undef __
 
 }  // namespace compiler
 }  // namespace internal
 }  // namespace v8