[turbofan] Frame elision for code stubs

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

Index: src/compiler/arm/code-generator-arm.cc
diff --git a/src/compiler/arm/code-generator-arm.cc b/src/compiler/arm/code-generator-arm.cc
index 474e41dc99c07d6e488bc3edff25a573b56176fe..584871944dea59771ee0b1ef416b91b216c4fdbd 100644
--- a/src/compiler/arm/code-generator-arm.cc
+++ b/src/compiler/arm/code-generator-arm.cc
@@ -244,7 +244,7 @@ class OutOfLineRecordWrite final : public OutOfLineCode {
                                              : OMIT_REMEMBERED_SET;
     SaveFPRegsMode const save_fp_mode =
         frame()->DidAllocateDoubleRegisters() ? kSaveFPRegs : kDontSaveFPRegs;
-    if (!frame()->needs_frame()) {
+    if (!frame_access_state()->access_frame_with_fp()) {

[danno, 2016/03/16 18:18:11]

Here and below, and on other ARM-like platforms, I don't think is is correct.
You can set up frame access with the SP even if you have a frame, this is what
happens in tail calls and there is a bot that builds and runs all of the tests
running with sp-based (not fp-based) frame access. You really need a predicate
on the frame_access_state() called has_frame().

[Mircea Trofin, 2016/03/16 20:20:46]

Ah, yes! I started from Intel and ended up cargo-culting the change from
frame()->needs_frame() over. In  my old prototype, I had this done by passing
down the information about whether the frame is constructed at the beginning of
the function or not. I only did it for arm64, it so happens (it was a
prototype).

Now... in this CL, I'm trying to lose any kind of notion of "was frame created
at beginning of function or not". So in that sense, I would probably want to
capture the frame state entering the OOL code. Can there be more than one code
path entering a OOL, btw? One where there is a frame setup, and one where there
isn't. I'll look into it.

       // We need to save and restore lr if the frame was elided.
       __ Push(lr);
     }
@@ -257,7 +257,7 @@ class OutOfLineRecordWrite final : public OutOfLineCode {
       __ add(scratch1_, object_, Operand(index_));
     }
     __ CallStub(&stub);
-    if (!frame()->needs_frame()) {
+    if (!frame_access_state()->access_frame_with_fp()) {
       __ Pop(lr);
     }
   }
@@ -386,6 +386,9 @@ Condition FlagsConditionToCondition(FlagsCondition condition) {
     DCHECK_EQ(LeaveCC, i.OutputSBit());           \
   } while (0)
 
+void CodeGenerator::AssembleDeconstructFrame() {
+  __ LeaveFrame(StackFrame::MANUAL);
+}
 
 void CodeGenerator::AssembleDeconstructActivationRecord(int stack_param_delta) {
   int sp_slot_delta = TailCallFrameStackSlotDelta(stack_param_delta);
@@ -402,7 +405,7 @@ void CodeGenerator::AssemblePrepareTailCall(int stack_param_delta) {
     __ sub(sp, sp, Operand(-sp_slot_delta * kPointerSize));
     frame_access_state()->IncreaseSPDelta(-sp_slot_delta);
   }
-  if (frame()->needs_frame()) {
+  if (frame_access_state()->access_frame_with_fp()) {
     if (FLAG_enable_embedded_constant_pool) {
       __ ldr(cp, MemOperand(fp, StandardFrameConstants::kConstantPoolOffset));
     }
@@ -580,7 +583,7 @@ void CodeGenerator::AssembleArchInstruction(Instruction* instr) {
       DCHECK_EQ(LeaveCC, i.OutputSBit());
       break;
     case kArchParentFramePointer:
-      if (frame_access_state()->frame()->needs_frame()) {
+      if (frame_access_state()->access_frame_with_fp()) {
         __ ldr(i.OutputRegister(), MemOperand(fp, 0));
       } else {
         __ mov(i.OutputRegister(), fp);
@@ -1199,6 +1202,7 @@ void CodeGenerator::AssembleArchBranch(Instruction* instr, BranchInfo* branch) {
 
 
 void CodeGenerator::AssembleArchJump(RpoNumber target) {
+  AssembleDeconstructFrameBetweenBlocks();
   if (!IsNextInAssemblyOrder(target)) __ b(GetLabel(target));
 }
 
@@ -1259,7 +1263,7 @@ void CodeGenerator::AssembleDeoptimizerCall(
 
 void CodeGenerator::AssemblePrologue() {
   CallDescriptor* descriptor = linkage()->GetIncomingDescriptor();
-  if (frame()->needs_frame()) {
+  if (frame_access_state()->access_frame_with_fp()) {
     if (descriptor->IsCFunctionCall()) {
       if (FLAG_enable_embedded_constant_pool) {
         __ Push(lr, fp, pp);
@@ -1274,10 +1278,7 @@ void CodeGenerator::AssemblePrologue() {
     } else {
       __ StubPrologue(info()->GetOutputStackFrameType());
     }
-  } else {
-    frame()->SetElidedFrameSizeInSlots(0);
   }
-  frame_access_state()->SetFrameAccessToDefault();
 
   int stack_shrink_slots = frame()->GetSpillSlotCount();
   if (info()->is_osr()) {
@@ -1347,15 +1348,16 @@ void CodeGenerator::AssembleReturn() {
   }
 
   if (descriptor->IsCFunctionCall()) {
-    __ LeaveFrame(StackFrame::MANUAL);
-  } else if (frame()->needs_frame()) {
+    AssembleDeconstructFrameWhenLeaving();
+  } else if (frame_access_state()->access_frame_with_fp()) {
     // Canonicalize JSFunction return sites for now.
     if (return_label_.is_bound()) {
+      AssembleDeconstructFrameBetweenBlocks();
       __ b(&return_label_);
       return;
     } else {
       __ bind(&return_label_);
-      __ LeaveFrame(StackFrame::MANUAL);
+      AssembleDeconstructFrameWhenLeaving();
     }
   }
   __ Ret(pop_count);