[Ti] Will Apple drop its OS?

[Chris Olson]

On Apr 6, 2006, at 11:53 PM, Glenn L. Austin wrote:

> On the other hand, x86 has a direct "load relative to PC"  
> instruction which doesn't flush the instruction pipeline -- so  
> MachO is much more efficient.

Ah!  Now you're beginning to understand.  Mac OS X contains an  
abomination of a RISC kernel with a CISC ABI that tries to built on  
the perceived advantages of a microkernel and capitalize on the speed  
of a monolithic kernel.  Didn't happen.  Every commercial and open  
source Unix or Unix-like kernel in existence kicks XNU's a$$.  On x86  
it really gets its a$$ kicked.  So Apple tried to fix the abomination  
with GCC performing -mdynamic-no-pic optimizations, which hard-codes  
the data addresses in the code, so the result is roughly equivalent  
to the CFM ABI in OS 9.

A runtime ABI does not a RISC or CISC kernel make.  And neither did I  
make any assertions that OS X was even *designed* correctly.  Merely  
that XNU is a RISC kernel ported to x86.

> I was at Apple during the NeXT integration.  And you were where...?

Writing BSD kernel code.  Hiring Jordan didn't fix your design  
abomination like you'd planned, did it?
-- 
Chris

-------------------------
PGP Key:  http://astcomm.net/~chris/PGP_Public_Key/
-------------------------