Fix #239: Improve _Invoke_traits throughput

[lozinska]

Description

Checklist

Be sure you've read README.md and understand the scope of this repo.

If you're unsure about a box, leave it unchecked. A maintainer will help you.

• Identifiers in product code changes are properly _Ugly as per
  https://eel.is/c++draft/lex.name#3.1 or there are no product code changes.
• The STL builds successfully and all tests have passed (must be manually
  verified by an STL maintainer before automated testing is enabled on GitHub,
  leave this unchecked for initial submission).
• These changes introduce no known ABI breaks (adding members, renaming
  members, adding virtual functions, changing whether a type is an aggregate
  or trivially copyable, etc.).
• These changes were written from scratch using only this repository,
  the C++ Working Draft (including any cited standards), other WG21 papers
  (excluding reference implementations outside of proposed standard wording),
  and LWG issues as reference material. If they were derived from a project
  that's already listed in NOTICE.txt, that's fine, but please mention it.
  If they were derived from any other project (including Boost and libc++,
  which are not yet listed in NOTICE.txt), you must mention it here,
  so we can determine whether the license is compatible and what else needs
  to be done.

[BillyONeal]

Does this make it more difficult to if constexpr-ize invoke by virtue of exposing its implementation details in more places?

[StephanTLavavej]

Does this make it more difficult to if constexpr-ize invoke by virtue of exposing its implementation details in more places?

Maybe, but we can always undo or rework this later.

While this change successfully built and passed all of the STL's tests, it failed in the compiler's test suite, with the internal-only "checked" build of the compiler (which contains additional assertions to verify compiler state). The failure looks like:

996233_2.cpp(4): fatal error C1075: '{': no matching token found
D:\Contest\xlgmztty.fyl\binaries\x86chk\inc\type_traits(1733): Expectation failed: EXPECTATION_UNREACHABLE, file d:\agent\_work\2\s\src\vctools\Compiler\CxxFE\sl\p1\c\p0gettok.c, line 11957

This points to _Invoke_traits, so I'm confident that it was caused by this change. As far as I can tell, nothing is wrong with the change itself, it's simply triggering a pre-existing bug in the compiler. I'll need to reduce this to a minimal test case for the compiler team to investigate and fix; in the meantime, this PR is blocked, so I've marked it accordingly. Thanks for your patience.

[CaseyCarter]

The compiler folk have mentioned in the past that MSVC doesn't memoize the results of overload resolution but does memoize alias template expansions. I suspect we could improve throughput here by adding an alias template for decltype(_Invoker<_Types...>::_Call(_STD declval<_Types>()...)) a la:

template <class... _Types>
using _Meow = decltype(_Invoker<_Types...>::_Call(_STD declval<_Types>()...));

and replacing the two occurrences of decltype(_Invoker<_Types...>::_Call(_STD declval<_Types>()...)) here with _Meow<_Types...>.

[StephanTLavavej]

Thanks to Xiang Fan, the compiler bug has been fixed by Microsoft-internal MSVC-PR-222035 which will be available in VS 2019 16.6 (presumably Preview 1); we need to wait for that release, as we currently don't have a faster way to consume unreleased compilers.

[StephanTLavavej]

Unfortunately, we merged #585 which requires this PR to be updated.

[StephanTLavavej]

I'm testing an update.

[StephanTLavavej]

I went ahead and merged this since it passed our Microsoft-internal build and tests. (We checked in a GitHub test harness today, but we're still working through sporadic failures and long queue times.)

Thanks for the throughput improvement! Compile times are a top customer concern, so every bit helps.