From d0574f6b25ab01052e093ab92612520a7e4ada8d Mon Sep 17 00:00:00 2001 From: Matt Conley Date: Thu, 6 Sep 2018 08:22:37 -0700 Subject: Fixed clang formatting --- tensorflow/stream_executor/cuda/cuda_gpu_executor.cc | 17 +++++++++-------- tensorflow/stream_executor/cuda/cuda_gpu_executor.h | 12 ++++++------ 2 files changed, 15 insertions(+), 14 deletions(-) (limited to 'tensorflow/stream_executor') 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, -- cgit v1.2.3