Joel Rainville

You're confusing backward and upward compatibility. That "XP only" program you're talking about will still work in the next version of Windows. That's backward hardware compatibility. Or upward software compatibility if you prefer. Or the other way around. I'm getting confused myself. The fact is, old software runs on new hardware, not the other way around.

Marketing indeed won the war. Guess what that marketing was (and is!) based on? Year after year after year, IBM and Microsoft could come up to the owner of a big company and tell him that he could get faster processing just by replacing the hardware, and that it would run the existing software just fine. And that's basically why you still have an IBM-clone PC running Windows at home. Because the owner of that big company bought into the Microsoft/Intel solution 20 years ago, and kept doing it up until now. If at any point Intel decided to break the chain of backward compatibility (and they would have had very good reasons to do so, I think we all agree on that), things might be different today.

__________________
Time is nature\'s way of keeping everything from happening at once.
http://membres.lycos.fr/jrainville/

Joel Rainville

True, but that's not exactly what Intel and Microsoft are aiming for. They want that brand new Pentium to be backward compatible with last year's Pentium. It will take a while before a processor has enough power to do that in software to allow for a complete reengineering of the x86 architecture.

The Intel x86 hasn't evolved overnight. It's the constant backward compatibility offered year after year that won the war. It's like slowly winning a chess game using only your pawns

__________________
Time is nature\'s way of keeping everything from happening at once.
http://membres.lycos.fr/jrainville/

ThermalRunaway

But any CPU called a "Pentium" should be in some way compatible with another CPU with the same name anyway. Obviously newer Pentium processors may take advantage of new developments in technology, but basic compatability should be there, otherwise you could argue that they shouldn't call it a Pentium anymore. If we go back to Nigel's example of the 68000 series (a fantasic processor in it's day) the various version of this processor are very much compatible with each other, provided you write good code that is. If we take the last of the 68000 series (the 68060) and compare it to the first (the 68000) the 68060 is backwards compatible with it. The 68060 has newer and more improved features that software writers can take advantage of, but a 68060 can quite easily run code which was written for the 68000 because they are the same series of processor. The backwards compatability you're talking about is the leap from 8086 series to the Pentium series of processors. Apparently even these are backwards compatible, and it's this that I feel would be more cheaply accomplished from within software.

If you go back to the 68000 series again, the next CPU up from the 68060 is the PPC series of processors which, because they are not compatible with one another, are given a different series "name".

Brian

Joel Rainville

...

__________________
Time is nature\'s way of keeping everything from happening at once.
http://membres.lycos.fr/jrainville/

Joel Rainville

Indeed, and it is the currently reality. What's your point?

Yes, but someday they decided to stop the chain and start from scratch. That's where you start to lose market share. That's where the best marketing strategy in the world won't save your product. Not in this world anyway.

You're not getting my point. The newer Pentiums don't really care about compatibility with the 8086. It's just that they happen to be compatible with it because the x86 has always been backward compatible, year after year. That's why it doesn't make much sense to do it in software. It's already there in hardware, inherited from years of backward compatibility.

Maybe you're confused by the "Pentium" brand name? The pentium is just the next evolution after the 80486 and was in fact for a while called the 80586 before entering the market. Current Pentiums would be something like 80786s or so. It's the same basic x86 architecture, on serious steroids.

Why do you care about the actual brand names used? We're talking actual architecture and compatibility here, not what the processors are called.

Here's the actual backward compatibility chain. There's no leap between the Pentium and the 8086. The link is indirect, through years of evolution. Get it?

8086 < 80286 < 80386 < 80486 < Pentium < Pentium II < Pentium III < etc...

__________________
Time is nature\'s way of keeping everything from happening at once.
http://membres.lycos.fr/jrainville/

Joel Rainville

In fact, think of each newer processor littely "wrapping" around the previous one. Not just incorporating the previous one, but building on its foundation so that only the bare minimum that needs to evolve does. If you somehow remove the 8086 core, the Pentium doesn't work anymore. It's not just compatible with it for the fun of it. It needs it to run. It needs its 16 bits registers, its addressing scheme, all of its basic functionnality.

__________________
Time is nature\'s way of keeping everything from happening at once.
http://membres.lycos.fr/jrainville/

ThermalRunaway

Hmmm ok, point taken. Mind you, that kind of backwards compatability has obvious disadvantages too. It's undoubtedly hard to make an efficient processor which is heavily based on old technology. I suppose that's why PPC processors are often quoted as achieving the same relative processing power on lower clock speeds.

As it happens I have a small pet-hate for intel processors. I can't really explain why, except that when you learn to program a few different types of CPUs, you tend to develop preferences for some over others. For example, I'm a real fan of the Z80 CPU because I find it easy to program with and it has some excellent features. I also liked the 6502. I tried working with some x86 processors for a small while and I just hated everything about them, I didn't find them as easy to program as processors I'd tried before. They just seem to have horrible little quirks about them that I don't like.

One advantage of programming in C is that you're largely above all those low-level differences. You just let the compiler do all the hard work

Brian