Hyper
Mandy You should not have to unless you are running Mr Jones' applications, setup, and configuration. And I would doubt even then. No one else seems to be...
Mandy Although I am a fairly standard Linux user (I use Fedora-3 with KDE 90% of the time, Windows-XP 9% of the time, and an ancient Mac 1% of...
I would have called that fine-grained time slicing, where each given piece works on a uop for one virtual processor then a uop from maybe the other. (see below) Those pieces that are time sliced, either in fine or coarse granularity, are the key parts of the machine.
I recall the article stating that some portion of the scheduling does alternate. Maybe that's the input side of the scheduling queues. Maybe the output side of the queues let them mingle. (I've can't afford as detailed an analysis of the detailed article content this evening.)
Maybe it would have been more accurate for me to have said each piece of the major, key parts of the machine time slice in a fine or coarse grained manner.
True, from a high-level view. From a low-level view, each piece is time slicing independently of the other pieces.
I think you have discovered the basis of the disagreement. The time slicing (as I was using the term) is done in a distributed manner. Some pieces, such as the instruction decoders, slice in a large block and switch only once every several cycles. Other pieces switch every cycle or so, with very small pieces (as small as a single register or buffer) switching as a unit.
Does the wrap up the discussion?
Robert Riches (Yes, that is one of my email addresses.)