Make new space iterable when transitioning double array to objects

src/arm/codegen-arm.cc

Index: src/arm/codegen-arm.cc
diff --git a/src/arm/codegen-arm.cc b/src/arm/codegen-arm.cc
index cfc9dfec4dc6a3e1d905a3182ceb4635887a9596..be32ed4418eccc6742498e604b144166166f7b91 100644
--- a/src/arm/codegen-arm.cc
+++ b/src/arm/codegen-arm.cc
@@ -562,7 +562,7 @@ void ElementsTransitionGenerator::GenerateDoubleToObject(
   //  -- r3    : target map, scratch for subsequent call
   //  -- r4    : scratch (elements)
   // -----------------------------------
-  Label entry, loop, convert_hole, gc_required, only_change_map;
+  Label entry, loop, convert_hole, gc_required, gc_cleanup, only_change_map;
 
   if (mode == TRACK_ALLOCATION_SITE) {
     __ JumpIfJSArrayHasAllocationMemento(r2, r4, fail);
@@ -604,6 +604,15 @@ void ElementsTransitionGenerator::GenerateDoubleToObject(
   // r9: heap number map
   __ b(&entry);
 
+  __ bind(&gc_cleanup);
+#ifdef VERIFY_HEAP
+  // Make sure new space is iterable if we are verifying the heap.
+  __ mov(r0, Operand(Smi::FromInt(0)));

[Hannes Payer (out of office), 2014/04/08 14:19:27]

In allocation folding, we store the one_pointer_filler_map to uninitialized
memory when we verify the heap. Either do that to stay in sync with allocation
folding implementation or install a free_space_map with the given size. I
actually would prefer option B. Then you do not have to loop over the array.

[Jarin, 2014/04/08 21:04:16]

Replaced with one_pointer_filler_map (although I am still not completely sure
what is the advantage over Smi 0 - the array is dead, so any valid value should
be fine).

As for option B - we would have to calculate the size again and that seems to be
more error prone than just using the counter in r3 to finish the initialization
of the array. Performance does not matter here. The case is rare and we are
verifying heap - that dominates the running time anyway.

+  __ str(r0, MemOperand(r3, 4, PostIndex));
+  __ cmp(r3, r5);
+  __ b(lt, &gc_cleanup);
+#endif
+
   // Call into runtime if GC is required.
   __ bind(&gc_required);
   __ Pop(r3, r2, r1, r0);
@@ -618,7 +627,7 @@ void ElementsTransitionGenerator::GenerateDoubleToObject(
   __ b(eq, &convert_hole);
 
   // Non-hole double, copy value into a heap number.
-  __ AllocateHeapNumber(r2, r0, lr, r9, &gc_required);
+  __ AllocateHeapNumber(r2, r0, lr, r9, &gc_cleanup);
   // r2: new heap number
   __ ldr(r0, MemOperand(r4, 12, NegOffset));
   __ Strd(r0, r1, FieldMemOperand(r2, HeapNumber::kValueOffset));