Remove obsolete V8_INFINITY macro.

src/arm64/simulator-arm64.cc

 // Copyright 2013 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 <stdlib.h>
 #include <cmath>
 #include <cstdarg>
 #include "src/v8.h"
 
 #if V8_TARGET_ARCH_ARM64
@@ skipping 2487 matching lines @@
 template <class T, int ebits, int mbits>
 static T FPRound(int64_t sign, int64_t exponent, uint64_t mantissa,
                  FPRounding round_mode) {
   DCHECK((sign == 0) || (sign == 1));
 
   // Only the FPTieEven rounding mode is implemented.
   DCHECK(round_mode == FPTieEven);
   USE(round_mode);
 
   // Rounding can promote subnormals to normals, and normals to infinities. For
-  // example, a double with exponent 127 (FLT_MAX_EXP) would appear to be
+  // example, a double with exponent 127 (std::numeric_limits<float>::max()_EXP)

[Sven Panne, 2014/12/19 06:50:16]

:-)

+  // would appear to be
   // encodable as a float, but rounding based on the low-order mantissa bits
   // could make it overflow. With ties-to-even rounding, this value would become
   // an infinity.
 
   // ---- Rounding Method ----
   //
   // The exponent is irrelevant in the rounding operation, so we treat the
   // lowest-order bit that will fit into the result ('onebit') as having
   // the value '1'. Similarly, the highest-order bit that won't fit into
   // the result ('halfbit') has the value '0.5'. The 'point' sits between
@@ skipping 1311 matching lines @@
 
   delete[] format;
 }
 
 
 #endif  // USE_SIMULATOR
 
 } }  // namespace v8::internal
 
 #endif  // V8_TARGET_ARCH_ARM64