Don't strength reduce divisions by 1 or 2 as they can often be...

src/parser.cc

Index: src/parser.cc
===================================================================
--- src/parser.cc	(revision 2211)
+++ src/parser.cc	(working copy)
@@ -2653,10 +2653,15 @@
         double y_val = y->AsLiteral()->handle()->Number();
         int64_t y_int = static_cast<int64_t>(y_val);
         // There are rounding issues with this optimization, but they don't
-        // apply if the number to be divided with has a reciprocal that can
-        // be precisely represented as a floating point number.  This is
-        // the case if the number is an integer power of 2.
-        if (static_cast<double>(y_int) == y_val && IsPowerOf2(y_int)) {
+        // apply if the number to be divided with has a reciprocal that can be
+        // precisely represented as a floating point number.  This is the case
+        // if the number is an integer power of 2.  Negative integer powers of
+        // 2 work too, but for -2, -1, 1 and 2 we don't do the strength
+        // reduction because the inlined optimistic idiv has a reasonable
+        // chance of succeeding by producing a Smi answer with no remainder.
+        if (static_cast<double>(y_int) == y_val &&
+            (IsPowerOf2(y_int) || IsPowerOf2(-y_int)) &&
+            (y_int > 2 || y_int < -2)) {
           y = NewNumberLiteral(1 / y_val);
           op = Token::MUL;
         }