CUDA - Heterogeneous Parallel Processing 笔记

Class 2.2 CUDA Device Memory Management API functions

   cudaMalloc()

  -- Allocates object in the device global memory

  -- Two parameters

      Address of a pointer to the allocated object

      Size of allocated object  in terms of bytes

  cudaFree()

    -- Frees object from device global memory

 cudaMemcpy()

   -- memory data transfer

  -- requires four parameters

  -- Transfer to device is asynchronous