Re: 68k and coldfire and future
On Mon, Feb 26, 2007 at 01:39:28AM -0800, Brian Morris wrote:
> what i understood from reading about coldfire:
> it is 10x faster than 68060 (not fast by today's standards,
> but still capable if not unreasonable about demand from it)
> it is problematic in some floating point areas
> -- not that much faster there (think ppc604*), more in the integer part
> -- it does not really support the traditional 68k method of using trap,
> which is severely deprecated in all the rest of the computing world
> (* but with faster memory and bus and other fresher connectivities)
> -->>> 68k has to change some to accomodate CF.
Sounds approximately correct, yeah.
> it is very efficient, suitable for a handheld computer.
> (a real step from the old apple Newton, anyway)
> i believe that handhelds will be the hot product in the next few
> years. we don't expect them to be so fast. but us Linux people
> would like something like a fair respectable OS. it could require
> some packages need build on a desktop or/and with cross/dist
> could someone explain issue with TLS,
> ? it is needed because SSL is not up to GNU FSW standards ?
No, not *that* TLS :)
"TLS" for "Thread Local Storage". I don't know the intricate details
myself (yet), but it's some idiom supported by 2.6 kernels, which allows
(amongst other things) the dynamic loader to choose a library at
runtime, so that it can use a ColdFire-optimized library on a ColdFire
processor, and a 68040-optimized library on a 68040 processor.
See the "libc6-686" package on the i386 port for an example of how this
works in practice.
TLS is about more than dynamic linking, BTW; newer threading
implementations in glibc require it in order to function.
<Lo-lan-do> Home is where you have to wash the dishes.
-- #debian-devel, Freenode, 2004-09-22