Function AsGpuTimer

tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:93–96

Given a platform-independent timer datatype, returns the internal CUDA platform implementation pointer.

// Given a platform-independent timer datatype, returns the internal CUDA
// platform implementation pointer.
static GpuTimer* AsGpuTimer(Timer* timer) {
  DCHECK(timer != nullptr);
  return static_cast<GpuTimer*>(timer->implementation());
}

// Given const GPU memory, returns a libcuda device pointer datatype, suitable
// for passing directly to libcuda APIs.