I have wrotten a thread in Linux Forums about a CPU Layer compatibility to answer little-big endian method. I think it would be wonderful to teach the kernel to convert between the two types and ask the layer to do the translations ordering the data in memory or letting them in the original state. If the original data is from the most used format (in terms of Linux Users an lines of code written) to the same it is not converted, else if it is from the last used to the most it is converted.
What do you think about this? Can be made?
P.S.: There would be a lot of code and software repository that could be made in the same way!!!
Best regards Mattia Tristo
On Sat, 2011-12-24 at 16:56 +0100, Mattia Tristo wrote:
I have wrotten a thread in Linux Forums about a CPU Layer compatibility to answer little-big endian method. I think it would be wonderful to teach the kernel to convert between the two types and ask the layer to do the translations ordering the data in memory or letting them in the original state.
It's fairly much impossible to do this unless you know the precise format of the data.
What exactly are you trying to achieve?
Anything for which we have source is unlikely to be so badly-written that it has endianness issues — at least anything we care about.
And anything for which we *don't* have source is probably going to have to run in qemu-i386 anyway.
-
David Woodhouse -
Mattia Tristo