Method gpu_function_ptr

tensorflow/stream_executor/gpu/gpu_kernel.h:60–60

Returns the slot that the GpuFunctionHandle is stored within for this object, for the CUDA API which wants to load into a GpuFunctionHandle*.

  // Returns the slot that the GpuFunctionHandle is stored within for this
  // object, for the CUDA API which wants to load into a GpuFunctionHandle*.
  GpuFunctionHandle* gpu_function_ptr() { return &gpu_function_; }

  // CUDA supports setting the preferred cache configuration of a
  // GpuFunctionHandle (more-or-less equivalent to a GpuKernel). We support this