Might it be possible to extend the proposal “Integer overflow arithmetic” (below), enabling throwing exceptions for overflows in the case the CPU supports it? The purpose is to enable constructing a type with performance in the absence of overflow approaching that of the built-in integral types, switching to higher precision when there is, and it is not possible to get high performance by conditional checks within the C++ language.

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2024/p3161r2.html
https://lists.isocpp.org/std-proposals/2024/02/9079.php