Remove SMI length check from TF_BUILTIN(ArrayIncludes, CodeStubAssembler)

src/builtins/builtins-array.cc

Index: src/builtins/builtins-array.cc
diff --git a/src/builtins/builtins-array.cc b/src/builtins/builtins-array.cc
index e8ee8640c665b16a380fb3f3a2c65eabca1c8929..7b4e7f540c635393dfb1f75e674b7956c17ca1f8 100644
--- a/src/builtins/builtins-array.cc
+++ b/src/builtins/builtins-array.cc
@@ -1702,11 +1702,12 @@ TF_BUILTIN(ArrayIncludes, CodeStubAssembler) {
 
   Bind(&init_len);
   {
-    // Handle case where JSArray length is not an Smi in the runtime
-    Node* len = LoadObjectField(array, JSArray::kLengthOffset);
-    GotoIfNot(TaggedIsSmi(len), &call_runtime);
+    // JSArray length is always an Smi for fast arrays.
+    CSA_ASSERT(this,
+               TaggedIsSmi(LoadObjectField(array, JSArray::kLengthOffset)));
+    Node* len = LoadAndUntagObjectField(array, JSArray::kLengthOffset);
 
-    len_var.Bind(SmiToWord(len));
+    len_var.Bind(len);

[caitp, 2017/02/25 18:03:06]

Do we still need `len_var`? It never changes at any point in the algorithm, I
think

[vabr (Chromium), 2017/02/25 18:35:48]

So that I learn more:
(1) Changing would mean that there is a len_var.Bind() call somewhere after this
line?
(2) If I remove len_var, then I should keep the Node* pointer (len) and replace
all len_var.value() calls with it, correct?

[caitp, 2017/02/25 18:40:12]

There's another Bind() call before this line, but IIRC the intent was just to
make the compiler happy having a bound value for len in the call_runtime block
(which doesn't use it). So, as far as I can tell (based on a quick look), this
is the only Bind that does anything useful (e.g. changes the value
meaningfully).
Yeah, I think that would be better.

 
     GotoIf(IsUndefined(start_from), &select_loop);