5 ms·

For stack allocated scalars sure. What about stuff on the heap or arrays and structs?

• comex 8y ago
  Generally speaking, if a memcpy's size argument is a constant expression equal to the size of a native integer type, it will be transformed to a native load from the source pointer followed by a store to the destination pointer. After that, the load or store operation (or both) may itself be optimized away if the pointer points to a local variable. Thus, for example, given something like: void write32(int *ptr, int val) { memcpy(ptr, &val, 4); } …the load is removed entirely, and the store is kept but becomes a native store: write32: mov DWORD PTR [rdi], esi ret Godbolt link: https://godbolt.org/g/HYv24F https://godbolt.org/g/HYv24F