Managing heterogeneous device memory using C++17 memory resources

Abstract Programmers using the C++ programming language are increasingly taught to manage memory implicitly through containers provided by standard library. However, heterogeneous platforms often require explicit allocation and deallocation of memory. This discrepancy in management strategies can be daunting problematic for developers who not already familiar with programming. The C++17 introduces concept resources , which allow user control how library allocate memory; we believe that this addition is a powerful tool towards unification systems best-practice development. In paper, present vecmem allows efficient user-friendly on CUDA, HIP, SYCL devices containers. We investigate design use cases such library, potential performance gains over naive allocation, limitations model.