Short answer: give me one more day to come up with the benchmarks and the answers.
long answer:
I will back myself with benchmarks tomorrow. I couldn't do it today because I spent too much time trying to debug my code using visual studio 2026 but then I realized that the errors that I was debugging were from a bug in visual studio 2026 regarding template metaprogramming. Like the issue with visual studio 2026 is that the AI is misleading(ruins your code) but even worse: sometimes your code compilers and sometimes doesn't not, even while the code always compiles in Godbolt. All I can say now is that the issue is that heterogeneous lists are used like tuples are used, and its not better to use reflectors or complex code to index tuples using runtime indexes, like in my case, Godbolt is throwing "tuple constructor not found":
https://godbolt.org/z/ee8o8fzfh . I will spend tomorrow trying to fix it but relying on users to implement everything is frustrating. Like code in my pdf file works sometimes on visual studio 2026 and sometimes doesn't, but always works on Godbolt, so the point is that users want a standard interface so that they don't have to implement and debug it themselves. Not everyone is willing to spend their whole day reinventing the wheel. Most people don't care if there is an bug in the compiler or something that causes issues. Again, I am sorry that I haven't finished the benchmark code yet but I am working on it. I wont use C++ to benchmark but rather some specialized tool, the benchmark code is to take into all specializations possible, in fact I will. I couldn't find enough time to do it today because I was reading the bash manual, and I thought that this code(in the way I envisioned) would be extremely easy to write, but when I sat late, I realized that I was in for the trenches.