[turbofan]: Implement tail calls with more callee than caller parameters

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

 // Copyright 2014 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/arm/macro-assembler-arm.h"
 #include "src/compiler/code-generator-impl.h"
 #include "src/compiler/gap-resolver.h"
 #include "src/compiler/node-matchers.h"
@@ skipping 338 matching lines @@
       __ cmp(offset, i.InputImmediate(1));        \
     }                                             \
     auto value = i.InputRegister(2);              \
     __ asm_instr(value, i.InputOffset(3), lo);    \
     DCHECK_EQ(LeaveCC, i.OutputSBit());           \
   } while (0)
 
 
 void CodeGenerator::AssembleDeconstructActivationRecord(int stack_param_delta) {
   CallDescriptor* descriptor = linkage()->GetIncomingDescriptor();
-  int stack_slots = frame()->GetSpillSlotCount();
-  if (descriptor->IsJSFunctionCall() || stack_slots > 0) {
-    __ LeaveFrame(StackFrame::MANUAL);
-  }
-  if (stack_param_delta < 0) {
-    int offset = -stack_param_delta * kPointerSize;
-    __ add(sp, sp, Operand(offset));
+  int spill_slots = frame()->GetSpillSlotCount();
+  bool has_frame = descriptor->IsJSFunctionCall() || spill_slots > 0;
+  if (has_frame) {
+    if (stack_param_delta != 0) {

[Jarin, 2015/11/20 10:56:09]

Is the special casing for the stack_parameter_delta == 0 necessary?

[danno, 2015/11/20 14:43:23]

No, but the code is more compact. For simplicity I'll remove the special case.

+      int total_discarded_slots = frame()->GetTotalFrameSlotCount();
+      // Leave the PC and saved frame pointer on the stack.
+      total_discarded_slots -=
+          StandardFrameConstants::kFixedFrameSizeFromFp / kPointerSize;
+      // Discard only slots that won't be used by new parameters.
+      total_discarded_slots -= stack_param_delta;
+      if (total_discarded_slots > 0) {
+        __ add(sp, sp, Operand(total_discarded_slots * kPointerSize));
+      }
+    } else {
+      __ mov(sp, fp);
+    }
+    if (FLAG_enable_embedded_constant_pool) {
+      __ ldm(ia_w, sp, pp.bit() | fp.bit() | lr.bit());
+    } else {
+      __ ldm(ia_w, sp, fp.bit() | lr.bit());
+    }
   }
 }
 
+
+void CodeGenerator::AssemblePrepareTailCall(int stack_param_delta) {
+  if (stack_param_delta > 0) {
+    int total_discarded_slots = frame()->GetTotalFrameSlotCount();
+    // Leave the PC and saved frame pointer on the stack.
+    total_discarded_slots -=
+        StandardFrameConstants::kFixedFrameSizeFromFp / kPointerSize;
+    // Discard only slots that won't be used by new parameters.
+    total_discarded_slots -= stack_param_delta;

[Jarin, 2015/11/20 10:56:09]

The calculation of total_discarded_slots is repeated in the function above and
in (all?) other platforms. Could you extract it to a platform-independent
function?

[danno, 2015/11/20 14:43:23]

Done.

+    if (total_discarded_slots < 0) {
+      __ sub(sp, sp, Operand(-total_discarded_slots * kPointerSize));
+    }
+  }
+}
+
 
 // Assembles an instruction after register allocation, producing machine code.
 void CodeGenerator::AssembleArchInstruction(Instruction* instr) {
   ArmOperandConverter i(this, instr);
 
   masm()->MaybeCheckConstPool();
 
   switch (ArchOpcodeField::decode(instr->opcode())) {
     case kArchCallCodeObject: {
       EnsureSpaceForLazyDeopt();
@@ skipping 56 matching lines @@
     case kArchLazyBailout: {
       EnsureSpaceForLazyDeopt();
       RecordCallPosition(instr);
       break;
     }
     case kArchPrepareCallCFunction: {
       int const num_parameters = MiscField::decode(instr->opcode());
       __ PrepareCallCFunction(num_parameters, kScratchReg);
       break;
     }
+    case kArchPrepareTailCall:
+      AssemblePrepareTailCall(i.InputInt32(instr->InputCount() - 1));
+      break;
     case kArchCallCFunction: {
       int const num_parameters = MiscField::decode(instr->opcode());
       if (instr->InputAt(0)->IsImmediate()) {
         ExternalReference ref = i.InputExternalReference(0);
         __ CallCFunction(ref, num_parameters);
       } else {
         Register func = i.InputRegister(0);
         __ CallCFunction(func, num_parameters);
       }
       break;
@@ skipping 881 matching lines @@
       padding_size -= v8::internal::Assembler::kInstrSize;
     }
   }
 }
 
 #undef __
 
 }  // namespace compiler
 }  // namespace internal
 }  // namespace v8