Comment by Arnavion | Hacker News Reader

Comment by Arnavion | Hacker News Reader

Parent Full thread

>I suspect that Casey is using C++ so he may (unfortunately) have multiple translation units in his program.

Yes, the video uses C++.

Obviously if one compiles a library then the compiler has no way of knowing that other subclasses of `shape_base` do not exist. My point is that when compiling a binary as they are doing for their video, the compiler knows that there are no other subclasses that it needs to cater to.

It might require LTO explicitly, of course. At the very least godbolt doesn't devirtualize without LTO [1], but godbolt itself breaks if I enable LTO [2] and I CBA to test locally right now.

[1]: https://gcc.godbolt.org/z/498nKEzhK

[2]: https://gcc.godbolt.org/z/WqhP3fzxK

Correct me if I'm wrong. Even if the compiler devirtualizes the classes, you still have the memory cost of storing the vtable pointer in each of the object instances (8 bytes for each instance), which means you need to do more fetches from memory. Does CPU prefetching negate the cost of these additional memory lookups?

The enum to tell what shape an object is costs something, probably 4 bytes. Storing the 8 byte vtable constant isn't so bad considering.

looking at a value is faster than dereferencing a pointer—why is this controversial?

Even when the target is an app, said app might be loading DLLs at runtime which may contain other subclasses.

The fundamental problem here is that C++ doesn't have a good abstraction to represent visibility of public types, since any other translation unit - even across the DLL boundary! - can re-declare the type and then derive from it. The only way to constrain visibility is to use anonymous namespaces, and that only works if the type can be confined to a single unit (that C++ compilers seem to ignore the optimization opportunities here in practice perhaps indicates just how rare this actually is).

To fix 2, add '-fno-rtti' to the args: https://gcc.godbolt.org/z/zMdfx71qE Unfortunately, gcc does not seem to devirtualize in your example

Reply on news.ycombinator.com