Boost logo

Boost :

From: Yashin Evgeny (evgeny.yashin_at_[hidden])
Date: 2006-04-17 00:07:23


  Hello.
  Fixed point computation appears too often on an embedded systems, but it
seems that implement bug free class too complex (i found a lots of
implementations in internet but all of them have 2 issues - either they have
an computations bugs and precision loss either they not template). That's
why it will be good if in Boost appear standart fixed point library.

  I assume, that interface should look like these:

template< typename T, int N >
class fixed_point { .. };

where T - set for us bitness of the variable, and signed/unsigned flag, and
N - a number of bits for fractional part of number.

of course, for T - best to use types like: uint32, int32 and so on.. or
change interface on something like:

template< bool signed, int number_all_bits, int number_fract_bits >
class fixed_point { .. };


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk