3 Jun
2013
3 Jun
'13
7:16 p.m.
On 06/03/2013 08:06 PM, David Riley wrote:
On Jun 3, 2013, at 5:24 PM, Dave McGuire <mcguire
at neurotica.com> wrote:
I thought it was because the 386 didn't have some of the atomic opcodes
needed to make mutexes efficient (cmpxchg, in particular) so they had to
create a bunch of "rewritten at boot time" code chunks depending if
those opcodes were present.
On 06/03/2013 03:48 PM, Liam Proven wrote:
Ahh, then why was it so important for
Linux to specifically de-support only the 80386?
I know there's at least one
byte-swapping instruction that appeared
in the 486. Beyond that, I assume it has more to do with the MMU
and other associated builtins (cache controller, maybe?) than with
actual instructions. When 486
and Pentium came out, they were treated as fast 386s, and seemed
to stay that way for a very long time.
Well, they are. The instruction set
didn't change from the 80386 in
1985 until the Pentium-MMX in about 1996.
- Dave