Re: [Boost-users] [Review] UUID library (mini-)review starts today, November 23rd

Subject: Re: [Boost-users] [Review] UUID library (mini-)review starts today, November 23rd
From: Scott McMurray (me22.ca+boost_at_[hidden])
Date: 2008-12-09 12:40:11

On Tue, Dec 9, 2008 at 12:07, Andy Tompkins <atompkins_at_[hidden]> wrote:
>
> I understand your desire.  I would like to put in mutating iterators.  I
> wrote
> in my last post, should the uuid library require CHAR_BITS == 8, or can
> it
> relax and only require that CHAR_BITS % 8 == 0.  That is the uuid
> library only
> needs 8 bit bytes.  I guess the difference will be in implementating the
> iterators.
>

I think it's reasonable for the uuid class itself to work iff
boost/cstdint.hpp provides a uint8_t, and for the generators to work
if boost/cstdint.hpp provides a uint32_t.  (The latter is mostly for
the hash, but requiring it for the random too is convenient).

I don't see the added implementation complexity to support unusual
architectures as being worth it.  For example, even the C compiler for
the last mcu I used (one of the weaker MSP430s) provided both, despite
it being about 500 (or 65000, depending) times slower than and having
one millionth the ram of my laptop.

That said, I recall a discussion on the LLVM list a bit back about an
architecture with 96-bit registers separated into four 24-bit
channels.  Is anyone running Boost on such things?