You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The bit operations are allowed only on BitVector and we should remove it from the Integer / ModularInteger types. The frontend will insert appropriate implicit casting as needed. Arithmetic operators are defined for both Integer (wrap around is a runtime error in Debug mode) and ModularInteger (that wrap around). Comparison is also defined for both integers.
ModularInteger is not allowed as a loop variable. Initially the ModularInteger would be of modulo 2^8, 2^16, 2^32, 2^64 only, fast on hardware. Later we can add more (not as fast on hardware obviously).
The two use cases of BitVector (bit fields) and ModularInteger (modular arithmetic) are good use cases, they should not make things complex. We will see if every other use case of "unsigned integer" can be handled by the combination of these two.
The text was updated successfully, but these errors were encountered:
In CPython, we can use https://pypi.org/project/BitVector/ to represent BitVector, and for ModularInteger, we could use the unsigned integers from NumPy or CTypes (possibly wrapped in our own class), or 3rd party packages like mod.
In C, the signed integer overflow is undefined, while unsigned integer overflow wraps around. So ModularInteger wraps around and overflow cannot happen. Integer on the other hand should give a runtime error in Debug mode for overflows, like accessing an array out of bounds.
When interfacing C, the most common usage of unsigned integers is to represent ordinal numbers. In that case, they correspond to ModularInteger, not BitVector. So in BindC we should map ModularInteger to unsigned integer in C.
The bit operations are allowed only on
BitVector
and we should remove it from theInteger
/ModularInteger
types. The frontend will insert appropriate implicit casting as needed. Arithmetic operators are defined for bothInteger
(wrap around is a runtime error in Debug mode) andModularInteger
(that wrap around). Comparison is also defined for both integers.ModularInteger is not allowed as a loop variable. Initially the ModularInteger would be of modulo 2^8, 2^16, 2^32, 2^64 only, fast on hardware. Later we can add more (not as fast on hardware obviously).
We have to be careful to avoid the pitfalls here: https://github.com/lcompilers/lpython/wiki/ASR-Design#unsigned-integers
The two use cases of BitVector (bit fields) and ModularInteger (modular arithmetic) are good use cases, they should not make things complex. We will see if every other use case of "unsigned integer" can be handled by the combination of these two.
The text was updated successfully, but these errors were encountered: