I apologize Jaws, but you called my statement "contraddictory" in that thread and I have a right to ask you to prove my contraddiction and to dispute the proofs that come forward (if it is possible to do so) or admit defeat and conceede that my argument was contraddictng itself (if this indeed happened).
What I see is that you want to have the GPU being CELL based.
I also suggested that the CELL CPU's ALUs consisting of S|APUs be complemented by a CELL based GPU who's ALUs consisted of SALC/SALPs. The VS work done on a CELL CPU and the PS work done on the CELL GPU.
How would that GPU be CELL based ? The SALCs/SALPs now are modified to support the CELL ISA ?
Would it really matter if APUs doing all shading ops (vertex and pixel) be replaced by say, vertex shading via APUs and pixel shading via these SALPs in the PixelEngine? Indeed, the GPU maybe without APUs and replaced by these SALPs entirely...
Then it is hardly CELL based, isn't it ?
In that case you could not run the same threads on the GPU as the SALPs/SALCs are not APUs, do not share their ISA and reconfiguring them to also to on-the-fly decoding of apulets and all that is required to mimic the APUs (DMA access functionality, use of LS, etc...).
The fact that Apulets can migrate over networks and be processed by any CELL based processor is possible because all those devices have units which have the same basic ISA, the SPUs/APUs's ISA, which was made clear IMHO by Suzuoki's patent. If you start adding layers of emulation to enforce compatibility with the APU ISA then you are defeating the point of CELL as stated in that very patent: the loss of efficiency due to complex software interfaces between CPUs with different ISAs that are networked together and are running the same software with say a Virtual Machine like the JVM.
Are you expecting the SALCs/SALPs to be modified to support the SPU/APU's ISA ?
IMHO, the point of SALCs was to connect serial ALUs in programmable pipelines (their structure would be flexible) to perform specific functions.
You could implement the execution logic of the APUs with SALCs: you could micro-code the entire SPU/APU ISA and have the SALCs do the work, but would that be more optimal than have the SPUs/APUs themselves do the work ?
If you used SALCs/SALPs in the GPU they would do a better job as programmable Pixel Engines and programmable TMUs.
I've supplied Hofstee's slides and video on his presentation...that is more known than my conclusion OR YOUR conclusion...
But his slides and his video do not mandate the logical step you make to generate your second premise which is a conclusion of your first premise.
Premise 1: If Hofstee was talking about two-way comms between CPU<=>GPU,
Premise 2 (a conclusion based on the validity of premise 1): then these VS CELL threads should run on the GPU also, no?
Replace the "should" with a "could".
Let's see the two arguments:
1.)
Premise 1: Hofstee was talking about two-way comms between CPU<=>GPU (accepted assumption = VS threads are running on the CPU).
Premise 2 (a conclusion based on the validity of premise 1): then these VS CELL threads should run on the GPU also, no?
---------
Conclusion: The GPU is CELL based.
2.)
Premise 1: If Hofstee was talking about two-way comms between CPU<=>GPU,
Premise 2 (a conclusion based on the validity of premise 1): then these VS CELL threads could run on the GPU also, no?
---------
Conclusion: This could mean that there is a point making a GPU that is CELL based to go along with a CELL based CPU, that is load balancing.
Your original conclusion was way too strong for its premises: "This should imply the GPU is CELL based" this basically goes to say that a non CELL based GPU makes no sense accroding to what Hofstee said.
Let's go back to your original post:
Premise 1: if you're gonna have vertex shading on the CELL CPU, then they're gonna be VS CELL threads, (aka software Cells).
Premise 1a: Hofstee was talking about two-way comms between CPU<=>GPU
---------------------
Conclusion 1 (or Premise 2 for Conclusion 2): these VS CELL threads run on the GPU also.
Conclusion 2: The GPU is CELL based.
Not only this commits a non sequitur (if you take, as I did while making of your post a series of Premises and Conclusions, the "should run on the GPU" as "the VS threads must be running on the GPU to make any sense" instead of a could run or "it would make some sense for them to run on the GPU too, for load balancing") going from the Premise 1 and Premise 1a to Conclusion 1 (the premises, which I am taking as true) do not necessarily support that conclusion (The GPU could not be CELL based and that conclusion would follwo from Premise 1 and Premise 1a), but you do beg the question going to conclusion 2 as conclusion 1, now your Premise 2, is not IMHO standing on solid ground thus you have a premise which is not necessarily better known than your Conclusion.
You used this logic to prove that my saying that VS could be done on the CPU while PS was done by a non CELL based GPU was contraddictory.
You did not say that we do not know which GPU we will have: CELL based or non CELL based, they both work...
You said:
Panajev2001a wrote:
....
With this said, I still say that IMHO the PlayStation 3 GPU is not CELL based, it does not have the SPUs/APUs.
....
It would seem to me a not bad idea to assign all the Vertex Shading work to the CELL based CPU:
....
To me the above two statements seem to contradict each other.. Confused
Can you please be patient with my stubborn head and go through the steps that show my logical contraddiction ?
Why would dev generate geometry on the fly insted offline (+++ qualitity) and use all those power to other things ( AI...) 
