| Lord N said: The PS2 architecture was considered a nightmare by programmers. Some might even say that it was worse than the PS3, so that wasn't the reason they went multiplatform. That can be blamed on the PS3's high launch price and subsequent poor sales as developers knew that they wouldn't be able to sell enough games had they remained exclusive to the PS3. |
The problem is a bit more complicated, than you think. The PS-2 was more or less in parts an embedded system, that you had to use. This was its disadvantage. But once the programmers understood the concepts they were able to use this system as it was intended. On the other hand you can work with the PS-3 without any special care for the SPUs.
While the basic architecture of the PS-2 and the Xbox was too different to really reuse the same code-base on multiple plattforms, the differences between the PC the Xbox 360 and the PS-3 are not so deep. Furthermore: all of this plattforms work with HD graphics. If yozu have these graphics for one plattform you can modify the system cheaply and reuse these resources.
But while the differences between Little Endian and Big Endian architectures can be handled by the compiler in most of the cases, the SPUs can't use and access the same librabries as the PPU or each of the different cores in the Xbox 360 or the PC. Instead you have to dive much more into the architecture of the SPU, but while you have a very detailed access to the SPU, you have no such control about the interprocess communication inside the Cell itself and the rest of the memory systems, instead they are optimised for data throughput, not for time or priority. The Cell was not developed for real time applications with priority issues. If you have a SPU with critical results these results can be delayed by less important results!
Furthermore: The Cell was a complete new architecture. Without such a daring move the processor would have been cheaper as well as the design and construction of the complete PS-3.
