Int8, Int16, Int32, Int64 vs Int1, Int2, Int4, Int8

[alextretyak]

What way of the designation of the size of Int/Float do you prefer? Bits or bytes?

 

[alextretyak]

Bytes are used in PostgreSQL and in Fortran 77.
Bits are besides being used much more often, provide an opportunity to define arbitrary precision integers, e.g. Int2 or Int4 for 2 or 4 bits like in Object Vision, or even Int3 for 3-bits integer (which translates into C bit fields: int i : 3).

Also, there can be [UInt1] type, which represents a bit array {like std::vector<bool>} in contrast to [Bool] type (which represents an array of Bools).

 

[farter]

tree-like keyword is innovative! also have some idea here.
so far best imo is like rust, U8 I32, but if one'd think they look ugly under global namespace, what about putting those under T., like T.U8?
also to note some, T.I8x16 for simd, T.F_1_8_7??? for some custom format like bfloat16 (since you mentioned fortran, haha)

 

[alextretyak]

Float4 can be [confused/]associated with 4D vector from HLSL and daScript. [This is an argument for Float32 rather than Float4.]