diff options
| author |  A. Unique TensorFlower <gardener@tensorflow.org> | 2017-10-02 10:18:01 -0700 |
|---|---|---|
| committer | TensorFlower Gardener <gardener@tensorflow.org> | 2017-10-02 10:22:03 -0700 |
| commit | c0644791cfc064d5e4652271e51d826aeccad0c2 (patch) | |
| tree | be0f6fe47be1b464dfe99394c15101298900e859 /tensorflow/core/framework/bfloat16_test.cc | |
| parent | 3c00952c6680d77ee2f10def35fbc7cbd138aea3 (diff) | |
Change bfloat constructor to accept a float to avoid truncation in implicit conversion from non-integer types to uint16_t.
PiperOrigin-RevId: 170717628
Diffstat (limited to 'tensorflow/core/framework/bfloat16_test.cc')
| -rw-r--r-- | tensorflow/core/framework/bfloat16_test.cc | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/tensorflow/core/framework/bfloat16_test.cc b/tensorflow/core/framework/bfloat16_test.cc index 5bd95b806f..af4e6a4411 100644 --- a/tensorflow/core/framework/bfloat16_test.cc +++ b/tensorflow/core/framework/bfloat16_test.cc @@ -23,7 +23,8 @@ namespace { TEST(Bfloat16Test, Simple) { bfloat16 a(12); - EXPECT_EQ(12, a.value); + // Floating point representation of 12: 0x41400000 + EXPECT_EQ(0x4140, a.value); } TEST(Bfloat16Test, Conversion) { |