PC General Discussion Thread

[E6600]

I like their idea of non-symmetrical cores.

[Power666]

Non-symmetrical cores is usually a taboo in the PC space. Operating systems have to be adapted to properly load balance between them.

If that chip uses different instruction sets per core type, don't expect it to pick up support unless the auxiliary cores are abstracted through an industry standard API.

[chedabob]

It sounds like a good idea on paper, but I bet it'll turn out to be another Cyrix.

[E6600]

I want to know how they'll figure out how much processing does a process need, and to which core it needs to be directed.

[chedabob]

Process priorities in the assembly code?

[E6600]

That's one way, but you can never know the future, in other words you can never know how much clock cycles the process will take.

[Power666]

Originally Posted by E6600

That's one way, but you can never know the future, in other words you can never know how much clock cycles the process will take.

Sure you can. Cycle accurate simulators do exist.

[E6600]

Originally Posted by Power666

Sure you can. Cycle accurate simulators do exist.

Hmmm.. I'm sure there was something in the OS course I took that talked about we can never know the future and it applies to the resources a process will need in its lifetime.

[Power666]

Cycle accurate simulators don't predict the future, they record the present state of the running machine (actually a virtual machine).

[E6600]

What does recording the present state of the running machine has to do with knowing how much CPU time a process will need (which core to assign the process to)?