aboutsummaryrefslogtreecommitdiffhomepage
path: root/tensorflow/core/kernels/reduction_ops_gpu_bool.cu.cc
diff options
context:
space:
mode:
authorGravatar A. Unique TensorFlower <gardener@tensorflow.org>2017-09-13 14:32:25 -0700
committerGravatar TensorFlower Gardener <gardener@tensorflow.org>2017-09-13 14:36:05 -0700
commitd6f9d6109474a9162ef4d99520a2d4ef0becfb14 (patch)
tree697b9ef4b78b4ab570d7f1a074219844813cc9ed /tensorflow/core/kernels/reduction_ops_gpu_bool.cu.cc
parentf445958edbca3ad292c9ed8c9de0c7e047b1d2bd (diff)
Switch the softmax to use the new deterministic reductions on the GPU,
results in a speed up of 10-40x on the existing ImageNet benchmarks and 2-3x on the newly added transformer benchmarks. Update the benchmark to also run on the GPU. Remove duplicate cpu tests. PiperOrigin-RevId: 168596693
Diffstat (limited to 'tensorflow/core/kernels/reduction_ops_gpu_bool.cu.cc')
-rw-r--r--tensorflow/core/kernels/reduction_ops_gpu_bool.cu.cc2
1 files changed, 1 insertions, 1 deletions
diff --git a/tensorflow/core/kernels/reduction_ops_gpu_bool.cu.cc b/tensorflow/core/kernels/reduction_ops_gpu_bool.cu.cc
index 3e7a33ba3f..79ec1d59df 100644
--- a/tensorflow/core/kernels/reduction_ops_gpu_bool.cu.cc
+++ b/tensorflow/core/kernels/reduction_ops_gpu_bool.cu.cc
@@ -17,7 +17,7 @@ limitations under the License.
#define EIGEN_USE_GPU
-#include "tensorflow/core/kernels/reduction_ops_gpu_kernels.h"
+#include "tensorflow/core/kernels/reduction_gpu_kernels.cu.h"
namespace tensorflow {
namespace functor {