Re: [boost] [Endian] library proposal

Subject: Re: [boost] [Endian] library proposal
From: David Stone (david_at_[hidden])
Date: 2012-04-03 15:36:56

According to a message I received, the ARM CPU uses little-endian 32-bit
integers, and 64-bit integers are created by combining two little-endian
integers in big-endian order, which by my calculations gives us a byte order of
DCBAHGFE. I actually defined PDP-endian to do this byte ordering for 64-bit
words simply because I couldn't forbid it due to dynamic byte ordering concerns.
This page:
http://infocenter.arm.com/help/topic/com.arm.doc.ddi0278b/Cegbbbab.html
suggests that the ARM can actually switch between big-endian, little-endian, and
middle-endian!

However, this brings up a concern of the best way to name things. The PDP-11
didn't have 64-bit integers (unless there's some extension somewhere that I'm
not aware of), so calling this PDP-endian is really a misnomer for 64-bit
integers. The ARM processor is either big-endian or little-endian for 32-bit
access, so it can't even be generalized to some middle-endian version, I would
need something like PDP-endian and ARM-endian.

I'm assuming that all multi-endian (the term bi-endian doesn't apply to the ARM,
which appears to have three endian modes) processors require dynamic endian
determination.