I'll have to defer to theMy understanding is that OoOE makes everything run better. Even the most highly optimised serial code will run faster. You can't optimise highly enough to negate all benefits of the CPU being able to re-orer its workloads on the fly.
proper devs on this. My understanding is based more on the principles going back to really simple CPUs. I suppose it's possible on modern architectures with multiple functional blocks to run operations in parallel OoO. However, the transcoding process itself is going to tie up some of that, and I don't think Jaguar has a lot of functional blocks to use. Aren't they comparable PPC vs Jag?