experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Adding rationale for not including uint32 and uint64 as part of TF_CALL_INTEGRAL_TYPES.

Browse Source 
PiperOrigin-RevId: 205165086
This commit is contained in:
Jiri Simsa 2018-07-18 17:22:58 -07:00 committed by TensorFlower Gardener
parent 6bd8f7d241
commit 4a547aa155
1 changed files with 6 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
tensorflow/core/framework/register_types.h
View File

@ -151,6 +151,12 @@ limitations under the License.
// Defines for sets of types.
// TODO(b/111604096): Add uint32 and uint64 to TF_CALL_INTEGRAL_TYPES.
//
// The uint32 and uint64 types were introduced in 10/2017 to be used via XLA and
// thus were not included in TF_CALL_INTEGRAL_TYPES. Including them in
// TF_CALL_INTEGRAL_TYPES should only happen after evaluating the effect on the
// TF binary size and performance.
#define TF_CALL_INTEGRAL_TYPES(m) \
TF_CALL_int64(m) TF_CALL_int32(m) TF_CALL_uint16(m) TF_CALL_int16(m) \
TF_CALL_uint8(m) TF_CALL_int8(m)