Hello everyone
In DVFS in real systems, there are 2 parameters that come into picture
during voltage-frequency transition.
1. "Core dead time" during which core does not execute anything and is
the time taken to actually change the frequency.
2. "voltage change time" during which the core is executing instructions.
I have done some quick digging into the DVFS implementation in gem5. It
seems there is one parameter "transition_latency" associated with dvfs
handler. When a perlevel (V,f) change request is initiated, the handler
schedules an event that changes the perflevel (V,F) after
"transition_latency". I think this models only parameter (2) above
(voltage change time) and the parameter (1) above (core dead time) is
not modelled precisely.
Please comment if this observation of mine seems accurate/inaccurate!
--
Regards
Lokesh Jindal
UW-Madison
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
