[PATCH] D142446: [MC] Disable copying and moving of MCInstrDescs the C++20 way

[Ilya Biryukov via Phabricator via llvm-commits]

ilya-biryukov added a comment.

In D142446#4078204 <https://reviews.llvm.org/D142446#4078204>, @dblaikie wrote:

> Alternatively, how bad would it be if aggregate initialization were removed and these objects were initialized with ctor calls? (bad for an -O0 build, no doubt... (how bad?) calling all those ctors to build the tables, but does clang optimize all the ctor calls away to the equivalent of aggregate init when using optimizations?)

The alternative suggestion would be to add some library support, roughly this:

  #include <cstdint>
  
  struct NonCopyableNonMovable {
    NonCopyableNonMovable() = default;
    NonCopyableNonMovable(const NonCopyableNonMovable&) = delete;
    NonCopyableNonMovable(NonCopyableNonMovable&&) = delete;
    NonCopyableNonMovable &operator=(const NonCopyableNonMovable &) = delete;
    NonCopyableNonMovable &operator=(NonCopyableNonMovable &&) = delete;
  };
  
  #if defined(__has_cpp_attribute) && __has_cpp_attribute(no_unique_address)
  #define LLVM_NO_UNIQUE_ADDRESS [[no_unique_address]]
  #elif defined(__has_cpp_attribute) && __has_cpp_attribute(msvc::no_unique_address)
  #define LLVM_NO_UNIQUE_ADDRESS [[msvc::no_unique_address]]
  #else
  #error "WTF"
  #define LLVM_NO_UNIQUE_ADDRESS
  #endif
  
  #if __cplusplus >= 202002L
  // Put as the last member in the aggregate struct.
  #define LLVM_DISABLE_COPY_AND_MOVE(X)\
      LLVM_NO_UNIQUE_ADDRESS NonCopyableNonMovable DisableCopy##X;
  #else
  #define LLVM_DISABLE_COPY_AND_MOVE(X)\
      X(X const&) = delete;\
      X(X&&) = delete;\
      X& operator=(X const&) = delete;\
      X& operator=(X&&) = delete;
  #endif

And use it like this:

  struct Y {
      uint64_t a;
      uint64_t b;
  
      LLVM_DISABLE_COPY_AND_MOVE(Y);
  };
  auto y = Y{1, 2};
  static_assert(sizeof(Y) == 16);

Seem to work in both C++17 and C++20, except in the latest MSVC (looks like a bug, but I'm not entirely certain):
https://gcc.godbolt.org/z/8nsooKz8c
I suspect this could be worked around, but I did not dig any further.

The question is: how useful is it and does it carry its weight?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D142446/new/

https://reviews.llvm.org/D142446