on 14.04.2010 at 19:50 Domagoj Saric wrote :
Considering it is an ancient trick and that any compiler targeting an OS as widespread as Windows has to 'support' it (e.g. BITMAPINFO struct) it is probably safe to assume as 'safe'/portable...
OTOH, it would be great if you could separate the core 'big int'/math logic from the storage/allocation logic, for example have: - a base 'math handling' class with functions that all take the actual location and size of the number as parameters
a very good point this way it would be almost trivial to implement an arbitrary precision number as well as fixed precision numbers (e.g. 'xint::fixed<128>') oh, you already wrote that:
- a wrapping template class(es) that can be configured with policies whether to use/work with fixed sized buffers/numbers (thus no memory allocation, thus no exception handling code, thus maximally lean code) or dynamically sized buffers (with or without SBOs, deep or shallow copies, reference counting etc etc...)...
and a policy regulating whether to use ordinary arithmetic or modulo one would be vital e.g. a natural expectation is that a fixed 128 bit integer does modulo (2^128) arithmetic -- Pavel P.S. if you notice a grammar mistake or weird phrasing in my message please point it out