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

Define missing EIGEN_USE_THREADS in Tensorflow kernels.

Browse Source 
Including Eigen/Tensor header with/without EIGEN_USE_THREADS define in different translation units (read kernels) is an ODR violation and leads to undefined behavior, and performance degradation: default eigen executor does no vectorization and does not use threads.

PiperOrigin-RevId: 277082344
Change-Id: I55e079979c2271d190ad0d1ee26dc713f4153262
This commit is contained in:
Eugene Zhulenev 2019-10-28 09:45:12 -07:00 committed by TensorFlower Gardener
parent 09a394544e
commit bbde233007
4 changed files with 13 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
tensorflow/core/kernels/data/optional_ops.cc
View File

@ -12,6 +12,9 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#define EIGEN_USE_THREADS
#include "tensorflow/core/kernels/data/optional_ops.h"
#include "tensorflow/core/common_runtime/dma_helper.h"

11
tensorflow/core/kernels/list_kernels.cc
View File

@ -13,23 +13,24 @@ See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#include <limits>
#include "tensorflow/core/framework/allocator.h"
#define EIGEN_USE_THREADS
#if GOOGLE_CUDA || TENSORFLOW_USE_ROCM
#define EIGEN_USE_GPU
#endif // GOOGLE_CUDA || TENSORFLOW_USE_ROCM
#include "tensorflow/core/kernels/list_kernels.h"
#include <limits>
#include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
#include "tensorflow/core/framework/allocator.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/register_types.h"
#include "tensorflow/core/framework/tensor_types.h"
#include "tensorflow/core/framework/variant.h"
#include "tensorflow/core/framework/variant_op_registry.h"
#include "tensorflow/core/kernels/concat_lib.h"
#include "tensorflow/core/kernels/list_kernels.h"
#include "tensorflow/core/lib/core/coding.h"
#include "tensorflow/core/lib/core/errors.h"
#include "tensorflow/core/util/util.h"

2
tensorflow/core/kernels/nextafter_op.cc
View File

@ -13,6 +13,8 @@ See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#define EIGEN_USE_THREADS
#include "tensorflow/core/kernels/nextafter_op.h"
#include "tensorflow/core/kernels/cwise_ops_common.h"

2
tensorflow/core/kernels/ragged_tensor_to_tensor_op.cc
View File

@ -13,6 +13,8 @@ See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#define EIGEN_USE_THREADS
#include <stddef.h>
#include <algorithm>