InvestorsHub Logo
icon url

8-/

12/18/03 1:46 PM

#21111 RE: sgolds #21108

sgolds Looking at the Windows task manager, I see that I currently have 390 threads, so it would take a lot of space in the core to make this efficient on a server (which can have many more threads).

But, of those 390 threads, how many are actually using CPU time?

I'd guess the vast majority are just hanging around waiting to do something.

Just pointers in a 'things to do' list.

8-/
icon url

HailMary

12/18/03 2:13 PM

#21119 RE: sgolds #21108

I guess this means that current generation processors still push all the registers for a thread context shift. Too bad.

You may not realize what you are saying, but Hyperthreading is built on what you are describing. It allows 2 threads to have complete state on the processor without having to do a context dump to memory when switching. P4 hyperthreading is reusing the OoO (a.k.a microarchitecture) registers for this feature so you can get instruction level parallelism between 2 threads in this manner.

You can see for yourself how much Hyperthreading buys by looking at various reviews. In some cases it is very good to have this, but overall is not a huge win. Same thread out of order execution is a much bigger thing.
icon url

Petz

12/18/03 2:46 PM

#21126 RE: sgolds #21108

Hey, sgolds (or anyother oldtimer), remember STM 14,12,12(13)?

Petz