Originally posted by AppleRISC
You haven't "shown" anything either. Quit this derogatory nonsense.
Given that the Cell requires quite the underlying architectural change in terms of hardware design as well as operating system magic to deal with parallelism, threading, etc, then yes, the OS would likely require significant work. However, you continue to fail to present any reason why a run-of-the-mill Mac OS X application could not run without a recompile if a standard PowerPC instruction set were available, as early Cell specs seem to indicate. If you wan't to inform me exactly why that wouldn't work, please do, otherwise cut out the baseless posturing.
This is entirely meaningless. What does this have to do with running PPC binaries without recompiles? Sony would not be talking about building workstations based on Cell if it weren't a viable desktop architecture.
Which would be Apple's internal problem and not a transition issue.
No, it wouldn't. That would only include programs that are specifically trying to take advantage of parallelism/threading. There's no reason why single-threaded or even traditionally multi-threaded applications couldn't run on the PPE.
And what is that supposed to mean? I should discard the article because some anonymous guy on the AI forum said he doesn't "know anyone who is impressed" with it?
A traditional G5 processor could serve as the core as well.
Who said anything to that effect? How do you know what Apple's research efforts in other directions have been? Moreover, my entire comment was predicated on the assumption that if we were to be actually in the process of a Cell transition, these issues would have been resolved somehow. In which case, no, a Cell transition would not be "more disruptive." Or do you really think Apple would float some half-baked, slow alpha-level prototype crap into the market?
You continue to fail to mention why current code wouldn't be able to run on this technology, if it were properly factored for desktop use.
So that's what this entire rant and arrogant posturing has been about? About something I never even said? All I've said is this has potential, and that if the technology were currently in a state where it were actually a viable option for Macs that it would not be "more disruptive" than an x86 move. And I continue to maintain that. Adding VMX to the G4 didn't stop software from working, utilizing the GPU for image/video processing didn't stop software from working, and there's no reason to believe that a Cell processor with a PPC core couldn't run current software. Nothing would force a developer to refactor everything into cells, in fact that wouldn't even make sense for all software.
Those who say something is going to work are the ones responsible for showing how. If you said instead that you hope it were possible, that's different.
"quite the architectural change" is the point I was making. I'm not sure how after admitting that, you can go on to say "standard PowerPC instruction set".
While the PPE is compliant, it does not support re-entry code. You cannot simply ignore that. Without supporting that ( something that all modern processors support), a processor would continually drop instructions as it flushed its cache an inordinate amount of times. If the pipeline is short enough for this not to happen, that would be fine. But slow. The key is the "two issue, IN ORDER," part. All PPC's, as well as all other cpu's anywhere, support (and rely upon) out of order instructions.
The PPE can handle game code which is not that complex, because it can hand off the difficult parts of the program to the spe's.
But the part it's handing off are graphically oriented. For the most part, the spe's work as dsp's, even though they can do more.
If Sony and IBM want to make a workstation out of this other than the one they have made to use to program games for the PS3, they have to come up with a new OS and a set of API's that are written specifically for the Cell.
Apple was able to get OS X working on x86 because despite their differences, they are more alike than different.
The Cell is more different than alike.
You can't after all that (even in your rebuttal) say that it's "Apple's internal problem and not a transition issue." and dismiss it. That's the problem I'm talking about here. It's a big problem, and Apple, if they did think about it, no doubt agreed.
Why is there no reason that multi threaded apps wouldn't run? This only supports two threads. Some apps have more than that. Some now have several per cpu. The OS hands those threads off from one processor, when present, to the other depending on how busy it is.
Now we are getting to some more interesting things. If we used a G5 core, then we wouldn't have a Cell, because that concept is in direct contradiction to the idea of using a complex cpu like the G5. The concept was to strip the PPE down to its essentials so that the power usage would better fit a game console. The G5 surely doesn't do that.
Another problem is that the Cell uses Rambus technology for its memory and another Rambus technology for its I/O. Neither have any counterpart in Apple's equipment. They are both vastly different. The way the memory is implemented isn't used in the PC world either where they have flirted with Rambus. The entire memory architecture in Apple's OS and any program that runs on it would have to be completely reworked.
Insofar as Apple's research goes, we know that Next ran on x86. I suppose you could say that they had other things going at the same time, and you could be right. But the Cell is too new. They haven't even gone to silicon yet. Just a few prototypes. The PS3 itself won't be out until late spring or early summer.
How much time do you think it would take Apple to get their OS working on it? A year? Two? The OS works now on x86, and it will still take between one and two years to get machines and 3rd party software out the door in viable numbers. How many of these developers would want to make this switch?
And what about support chips? There aren't any out yet. We don't know what form these workstation will take. The industry is skeptical that they will be a success because no one is going to write to a whole new set of API's.
I haven't "failed" to show why current programs won't be able to run. You don't want to listen. you can't go by any one article. Especially as it goes against pretty much everything else being written.
Again, I'm not saying it's impossible, just that it's so difficult that it's not practical unless IBM decides to make it easy, which we know they won't.
I'm sorry about the last sentence. It wasn't directed specifically towards you. I should have made that clear. But if you have been following the posts on the threads, you would see what I mean.
I tried to present enough information to be useful. If you want to debate it further that's fine.
Otherwise, personal truce.