diff options
author | Matt Conley <mconley@nvidia.com> | 2018-09-04 16:17:40 -0700 |
---|---|---|
committer | Matt Conley <mconley@nvidia.com> | 2018-09-04 16:17:40 -0700 |
commit | cd6597b8fcd82b51ddb47a297972a1614c2a5d78 (patch) | |
tree | a76e16bd5a03f4957991e56cbf1e89f8884e5df8 | |
parent | fa20b59b920233d35bb8da3fbc3c234c369a8291 (diff) |
Fixed transition typo
-rw-r--r-- | tensorflow/stream_executor/cuda/cuda_gpu_executor.cc | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc b/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc index 458c0e3030..a961e9a6c4 100644 --- a/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc +++ b/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc @@ -493,7 +493,7 @@ void CUDAExecutor::VlogOccupancyInfo(const KernelBase &kernel, // Compute and return maximum blocks per core (occupancy) based on the // device description, some kernel characteristics and the number of threads per // block. If unable to compute occupancy, zero is returned. -int CalculateOccupancy(const DeviceDescription& device_description, +int CUDAExecutor::CalculateOccupancy(const DeviceDescription& device_description, uint64 registers_per_thread, uint64 shared_memory_per_block, const ThreadDim& thread_dims, CUfunction func) { @@ -508,7 +508,7 @@ int CalculateOccupancy(const DeviceDescription& device_description, // Compute and return the suggested thread count to acheive ideal occupancy. // If the provided thread dimensions match this number, zero is returned. -int CompareOccupancy(int* initial_blocks, +int CUDAExecutor::CompareOccupancy(int* initial_blocks, const DeviceDescription& device_description, uint64 registers_per_thread, uint64 shared_memory_per_block, |