On 11.07.2013 13:39, Nilay Vaish wrote: 

> On Sun, 7 Jul 2013,
Nilay Vaish wrote:
> 
>> On Fri, 5 Jul 2013, Ali Saidi wrote: 
>> 
>>>
Now lets say we're using the code that Andreas Sandberg committed ~6
months ago that lets you migrate execution between gem5 CPU models and a
KVM virtual CPU. If you drain the system and stop in the middle of an
instruction, there is no way the KVM virtual CPU is going to be able to
pick up where you left off.
>> Well, I would not call that a concrete
example. And, in an x86 system, it seems there is a dependence between
the boot cpu (cpu0) and others (cpu1). This means that if you are going
to wait for cpu1 to get out of the Idle state while cpu0 has already
drained, you will have a deadlock for sure.
> 
> Are we done with the
discussion here, or is there more to it?
> 
> --
> Nilay
>
_______________________________________________
> gem5-dev mailing
list
> [email protected]
> http://m5sim.org/mailman/listinfo/gem5-dev
[1]

I think that is a concrete example. We can't use the KVM switching
code if "drained" in a CPU means you can be in the middle of a microcode
sequence. Can you explain your situation more? I thought we determined
there was no way into the Idle state? 

Thanks, 

Ali 




Links:
------
[1] http://m5sim.org/mailman/listinfo/gem5-dev
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to