Replies to post #26764 on Advanced Micro Devices Inc (AMD)

[sgolds]

HailMary, I agree that there will be no new architectures for a long time. I differ from your list of survivors because I would add Power and put a big question mark near IPF:

x86 - long live the king!
IPF - ??? - can it survive beyond the decade?
Power - IBM seems to be positioning it nicely.
SPARC - Nahhh.

Maybe over time x86 will be able to drop most of the legacy ugliness and become more useful.

More useful? I guess you must mean 'continued useful'. Any case, have a look at the 64-bit design of AMD64. It is designed to cut off the legacy stuff over time. There are two major modes, and each one has two sub-modes:

Long Mode supports 64-bit Mode and Compatibility Mode. 64-Bit Mode uses new 64-bit instructions only. Compatibility Mode supports 32-bit and 16-bit Protected Modes applications under a 64-bit OS (underlying control structures are 64-bit, so it is kind of a Protected Mode emulation).

Legacy Mode supports the stuff that (IMHO) is partitioned to be jettisoned later: Protected Mode, Virtual-8086 Mode and Real Mode.

I expect that it won't be long before we see a version of Opteron that leaves off Legacy Mode. After all, who needs it on a server? If you run a 64-bit OS then each application is being run either directly in Long Mode, or in Compatibility Mode to emulate the current environment. Under that 64-bit OS, Legacy Mode isn't even available!

The upgrade path is very nicely done. Anyone who thinks of this scheme as an extension of x86-32 is fooling himself: It is a way of replacing x86-32 with a whole new scheme in an upgradable manner. New register structure, new opcodes, everything changes.

Edit: I've been wondering about all the new cores that AMD is releasing on their roadmap. Maybe some of the high-end cores will not have Legacy Mode at all.

[chipguy]

Maybe over time x86 will be able to drop most of the legacy ugliness and become more useful. It is already going in the right direction. I'm sure we haven't seen the last of the extensions and evolution.

I agree, there is certainly a lot of room for improvement left
in x86. Keeping making better, more logical, and streamlined
until it looks exactly like Alpha and then there will be virtually
nothing left to do. :-P

However, improvements that ease the plight of x86 compiler
writers and x86 asm programmers polishing this 25+ year old
turd for more performance do not remove one gram of burden
from the multi-ton millstone around the neck of the poor souls
who have to design x86 chips to support *every* layer of x86
accretion all the way down to 16 bit segmented madness and
x87 stack stupidity. In the x86 world baggage is forever.