libyuv r1629 roll for AVX2 optimized HalfFloatPlane and ExtractAlpha.

cc/resources/video_resource_updater_unittest.cc

Index: cc/resources/video_resource_updater_unittest.cc
diff --git a/cc/resources/video_resource_updater_unittest.cc b/cc/resources/video_resource_updater_unittest.cc
index 9b9dcfccf9ab60ef7b9df11c429d745bd2b6ce9f..4ebf31d8af390c0d27c71a023f61893fdcc32354 100644
--- a/cc/resources/video_resource_updater_unittest.cc
+++ b/cc/resources/video_resource_updater_unittest.cc
@@ -548,14 +548,17 @@ TEST_F(VideoResourceUpdaterTest, MakeHalfFloatTest) {
 
     // Multiplier to converting integers to 0.0..1.0 range.
     double multiplier = 1.0 / (num_values - 1);
+    float expected_precision = 0.f;
 
     for (int i = 0; i < num_values; i++) {
       // We expect the result to be within +/- one least-significant bit.
       // Within the range we care about, half-floats values and
       // their representation both sort in the same order, so we
       // can just add one to get the next bigger half-float.
-      float expected_precision =
+      if (i < num_values - 1) {

[hubbe, 2016/10/20 18:55:11]

Why is this change needed?
It seems it will just use the previous expected precision for the last number.
That doesn't seem bad, but why do we need it?

[fbarchard1, 2016/10/20 22:57:34]

Done.  unittest does not need a change.

+        expected_precision =
           FromHalfFloat(half_floats[i] + 1) - FromHalfFloat(half_floats[i]);
+      }
       EXPECT_NEAR(FromHalfFloat(half_floats[i]), integers[i] * multiplier,
                   expected_precision)
           << "i = " << i << " bits = " << bits;