From: boost-users-bounces@lists.boost.org [mailto:boost-users-bounces@lists.boost.org] On Behalf Of Murilo Adriano Vasconcelos
Sent: Tuesday, July 20, 2010 3:39 PM
To: boost-users@lists.boost.org
Subject: Re: [Boost-users] About naming of integer log in base 2

Hi,

Maybe, it depends on the relative cost of two: the reason for unchecked_factorial is that low-order factorials are often used inside the inner loop of a computation, where anything more expensive than a simple table lookup is out of the question (and the check can easily be moved outside the loop anyway).  Does the same thing apply here.... or is the cost of the check trivial compared to the bit manipulations?  I don't know the answers to these BTW, just some questions you might want to think about ;-)

Thank you John! 

I thought better and followed Paul's recommendations and implemented math::policies on ilog2() [0], and also changed the count_leading_zeros(T value) to return the size in bits of T when value is equal to 0 (all bits are leading 0's).

So I've added two ifs, but think that the design of the functions are better.

[0] - https://svn.boost.org/trac/boost/browser/sandbox/SOC/2010/bits_and_ints/boost/integer/ilog2.hpp?rev=64198

Looks more promising, but you have a typo in the error message ;-)

retunrning -1",

and maybe you need to document some examples of using policies that don’t bother to check or at least do so quietly (your original idea), just quietly return -1, produce a message and return -1, or message and then throw an exception?

But maybe it’s a TODO ;-)

Paul

---

Paul A. Bristow

Prizet Farmhouse

Kendal, UK   LA8 8AB

+44 1539 561830, mobile +44 7714330204

pbristow@hetp.u-net.com