Subj : Re: X86S
To   : Nightfox
From : tenser
Date : Thu Apr 25 2024 08:00:51

On 24 Apr 2024 at 12:55p, Nightfox pondered and said...
 
 Ni> I think it's interesting that it happened that way.  I think moving
 Ni> forward while maintaining backward compatibility has been an advantage
 Ni> with x86 processors.

It's been a blessing and a curse.  They've been able to
retain compatibility with truly ancient software, and that's
no small feat.  But the architecture is atrociously complex
as a result.  Most of the time no one has to care, but OS
people do, and there's some seriously annoying vestiges of
bad ideas that you can't get away from (someone explain to
me, please, why I still have to care about the TSS is 64-bit
mode.  It's just a table of stack pointers; stuff 'em in
MSRs).

 Ni> Apple seems to have the opposite strategy, where they have no problem
 Ni> swapping out the processor in their Mac lineup to something entirely
 Ni> different.  They've done that several times in the history of the Mac. 
 Ni> I've done some software development work on an M1 Mac not too long ago,
 Ni> and one of the frustrations was having to use its x86 emulation
 Ni> sometimes to do some builds, as there were some 3rd-party software
 Ni> libraries that were still only supporting x86 and didn't support M1 yet.

By my count, the Mac is on its 4th hardware architecture (68k,
PowerPC, x86, and now ARM).

Their observation, and it's not a bad one, is that it doesn't
matter all _that_ much: once you've got decent binary translation
support in place for the transition, very few people are writing
assembly language directly anymore; you can just recompile for a
new ISA and go with that.  Of course, issues like you described
are irritating for software developers, most that's an edge case.

--- Mystic BBS v1.12 A48 (Linux/64)
 * Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

.