diff options
author | Matt Conley <mconley@nvidia.com> | 2018-09-06 08:22:37 -0700 |
---|---|---|
committer | Matt Conley <mconley@nvidia.com> | 2018-09-06 08:22:37 -0700 |
commit | d0574f6b25ab01052e093ab92612520a7e4ada8d (patch) | |
tree | 89ba73228de937a3ff8ba0128833cc566c89ab4c /tensorflow/stream_executor | |
parent | 475b7715f16ad0f94fa9986a0eefc1b2cf2044bd (diff) |
Fixed clang formatting
Diffstat (limited to 'tensorflow/stream_executor')
-rw-r--r-- | tensorflow/stream_executor/cuda/cuda_gpu_executor.cc | 17 | ||||
-rw-r--r-- | tensorflow/stream_executor/cuda/cuda_gpu_executor.h | 12 |
2 files changed, 15 insertions, 14 deletions
diff --git a/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc b/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc index ce2f1ce3ae..ef84d01a94 100644 --- a/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc +++ b/tensorflow/stream_executor/cuda/cuda_gpu_executor.cc @@ -493,10 +493,10 @@ 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 CUDAExecutor::CalculateOccupancy(const DeviceDescription& device_description, - uint64 registers_per_thread, - uint64 shared_memory_per_block, - const ThreadDim& thread_dims, CUfunction func) { +int CUDAExecutor::CalculateOccupancy( + const DeviceDescription& device_description, uint64 registers_per_thread, + uint64 shared_memory_per_block, const ThreadDim& thread_dims, + CUfunction func) { int suggested_blocks = 0; int suggested_threads = 0; CUresult err = @@ -509,10 +509,11 @@ int CUDAExecutor::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 CUDAExecutor::CompareOccupancy(int* initial_blocks, - const DeviceDescription& device_description, - uint64 registers_per_thread, - uint64 shared_memory_per_block, - const ThreadDim& thread_dims, CUfunction func) { + const DeviceDescription& device_description, + uint64 registers_per_thread, + uint64 shared_memory_per_block, + const ThreadDim& thread_dims, + CUfunction func) { int suggested_blocks = 0; int suggested_threads = 0; CUresult err = diff --git a/tensorflow/stream_executor/cuda/cuda_gpu_executor.h b/tensorflow/stream_executor/cuda/cuda_gpu_executor.h index e8ebbc3220..1481dcc19a 100644 --- a/tensorflow/stream_executor/cuda/cuda_gpu_executor.h +++ b/tensorflow/stream_executor/cuda/cuda_gpu_executor.h @@ -71,16 +71,16 @@ class CUDAExecutor : public internal::StreamExecutorInterface { const KernelArgsArrayBase &args) override; int CalculateOccupancy(const DeviceDescription& device_description, + uint64 registers_per_thread, + uint64 shared_memory_per_block, + const ThreadDim& thread_dims, CUfunction func); + + int CompareOccupancy(int* initial_blocks, + const DeviceDescription& device_description, uint64 registers_per_thread, uint64 shared_memory_per_block, const ThreadDim& thread_dims, CUfunction func); - int CompareOccupancy(int* initial_blocks, - const DeviceDescription& device_description, - uint64 registers_per_thread, - uint64 shared_memory_per_block, - const ThreadDim& thread_dims, CUfunction func); - void *Allocate(uint64 size) override; void *AllocateSubBuffer(DeviceMemoryBase *mem, uint64 offset_bytes, |