Date: Mon, 16 Feb 2026 14:41:25 +0100
>
> I don't want exact-width types. I want an integral type that's at least N
> bits in width, which currently std::(u)int_leastN_t cannot provide for all
> widths. _BitInt is an interesting idea, however, does it work with
> arbitrary sizes and produces an integral type with at least N bits in width?
> Does it also have a fast version for heavy computational data?
>
> I want to further standardise these 2 cases.
>
Maybe once we have _BitInt, we could add std::bit_int_unpadded_t and
std::bit_int_fast_t traits that give you the next larger _BitInt without
padding, and the next larger _BitInt that is "fast".
I'm not so sure we want such traits for the standard integer types.
> I don't want exact-width types. I want an integral type that's at least N
> bits in width, which currently std::(u)int_leastN_t cannot provide for all
> widths. _BitInt is an interesting idea, however, does it work with
> arbitrary sizes and produces an integral type with at least N bits in width?
> Does it also have a fast version for heavy computational data?
>
> I want to further standardise these 2 cases.
>
Maybe once we have _BitInt, we could add std::bit_int_unpadded_t and
std::bit_int_fast_t traits that give you the next larger _BitInt without
padding, and the next larger _BitInt that is "fast".
I'm not so sure we want such traits for the standard integer types.
Received on 2026-02-16 13:41:38
