Nothing wrong with answer (1) - it's the way I've done it in the past.
Zealots would advocate you use a union rather than a cast, but we've had
that particular religious war.
I thought the early Alphas were even worse - they didn't have bytes. So
string handling all had to be emulated with shifts and masks and nastiness
like that.
I think your answer is pretty acceptable though - stuff 'em. No mass-market
architecture is going to make that mistake again in a hurry.
Tom Forsyth - Muckyfoot bloke and Microsoft MVP.
This email is the product of your deranged imagination,
and does not in any way imply existence of the author.
> -----Original Message-----
> From: Brian Hook [mailto:bri...@py...]
> Sent: 08 November 2002 18:58
> To: gam...@li...
> Subject: [GD-General] Some portability questions
>
>
> Okay, these are fun ones. Let's do the easy one first.
>
> 1.) How do you portably and safely byte swap a floating point value?
>
> Disregarding the whole "some architectures aren't IEEE and/or
> don't have
> 32-bit floats" issue, it seems like the right way to do this safely is
> to cast the bits to an integer value, swap those, and store out as
> integer. To load a floating point value, you perform the inverse, and
> pray that you didn't load some invalid thing that will generate an FP
> exception on some systems.
>
> I hadn't really thought about this until I started working on POSH
> (Portable Open Source Harness).
>
> 2.) How do you cleanly handle architectures where, say, all data types
> are either 1-byte or 8-bytes. The Cray and early 21064
> Alphas were like
> this (actually, I don't think early Alphas even had atomic <8-byte
> read/writes).
>
> Answer: it's a Cray, dude. Shut up. If I can't figure out a clean,
> safe way of handling this, I may just punt on those architectures.
>
> Brian
|
