cuda pinned memory allocator for std::vector

cudaPinnedMemoryAllocator.cu

#include <cuda_runtime.h>

// how to use?
// std::vector<float, CudaHostAllocator<float>> vec;


template <typename T>
class CudaHostAllocator {
  public:
    using value_type = T;
    CudaHostAllocator() = default;

    template <class U> constexpr CudaHostAllocator(
      const CudaHostAllocator<U>&) noexcept {}

    T* allocate(std::size_t n) {
      T* ptr = nullptr;
      cudaError_t err = cudaMallocHost((void**)&ptr, n*sizeof(T));
      if (err != cudaSuccess) {
        throw std::bad_alloc();
      }
      return ptr;
    }

    void deallocate(T* ptr, std::size_t) noexcept {
      cudaFreeHost(ptr);
    }
};

template <typename T, typename U>
bool operator==(const CudaHostAllocator<T>&, const CudaHostAllocator<U>&) {
  return true;
}

template <typename T, typename U>
bool operator!=(const CudaHostAllocator<T>&, const CudaHostAllocator<U>&) {
  return false;
}