Well first thing's first, we have to take a look at the differences in hardware between a standard, run-of-the-mill PC and the PS3:
Modern PC-
CPU: 4-8 cores, 64-bit processors (typically referring to register, address, and ALU bit-width) pushing somewhere around 50-100GFLOPS for double-precision floating point numbers (takes 2 registers on a standard CPU. I'm not sure how the register architecture works on the CELL processors)
Memory: 4-12 gigs of DDR3 RAM
GPU: Somewhere between 600 GFLOPS and 2 TFLOPS
Video Memory: between 512MB and 2GB of very high-bandwidth, high-speed memory
There's plenty of more stats, but those are the ones that matter (or at least I think they are, though others will differ, I'm sure).
PS3-
CPU: CELL Engine Processor - Basically a CPU/GPU hybrid. Manages about 100 GFLOPS, in double precision, "theoretically."
Memory: 256 MB
GPU: NVIDIA RSX 'Reality Synthesizer'- roughly 192 GFLOPS
Video Memory: 256MB (not sure how fast, but probably fairly fast, at least).
Conclusion: PS3s are built for affordability. Any decent, modern PC could easily outperform a PS3 in every sector, so assuming you got past the hardware differences, running a PS3 game wouldn't be that costly to a modern PC. Given that the theoretical CPU power is basically like saying "my algorithm IN THE BEST CASE runs at O(1)" and then ignoring the worst case, which is somewhere closer to O(n^2) (in layman's terms "I can run REALLY FAST... if I've got a 220mph tail-wind...") The more average CPU efficiency is somewhere around 25 GFLOPS, which is honestly rather pathetic, by comparison to even modern CPUs, let alone GPUs.
That all being said, it is true that the CPU's base architecture DOES differ greatly from the modern-day CPU architecture in a number of ways, not least of which is the fact that it runs with a single primary core, and several specialized sub-cores, whereas the majority of standard PC CPU architecture involves duplicating cores that can perform all the necessary functions of a CPU. However, you must keep in mind that even these can vary drastically between various architecture types (the difference between a 4-core and an 8-core CPU is a hell of a lot more complicated than "we shoved more cores in it"). Emulation software and hardware-specific compilers are used for PCs, however, to make a programmer's job much easier, so they have to worry a lot less about architecture support when writing a piece of software, and can focus more on the more important questions, such as "does my software run fast enough, on average?" and "does my software run at all?"
While there is very little CPU-level interfacing built into the PS3's Cell Engine CPU (as in the stuff that makes it a little bit easier for compilers and high-level emulators to manage to support them in a broader sense), all that would really be required would be a layer of emulation (as we have used in the past for emulators). While this is obviously vast oversimplification of the problem, it's about the best answer you're going to get. We will have to wait for someone (or more likely a group of someones) to come up with a feasible way of re-directing the compiled code to the proper channels on PC hardware to help make up for the differences. Given that SONY still uses MIPS, to the best of my knowledge, that should hopefully not take too much longer, if it hasn't already happened.