Ok, that seems to make sense, being also a simpler approach, but where exactly it should be done? Objects are instantiated from python scripts, so I imagine you are thinking about a parameter for one of those scripts? Also, TLBs are instantiated from the CPU (itb and dtb), but I want to collect the TLBs of each cpu on the System. CPUs are instantiated in m5/simulate.py along other objects descending from Root, but I'm not sure how to retrieve the TLBs from there. I'm not really used to the Python/C++ mixture in the config, and grep search where the objects are instantiated is a bit hard. So it is not really clear to me how to proceed.

And if I finally manage to create that list, if it is a parameter on a script as mentioned, how can I make it accesible from the tlb.cc class?

Thank you for your response,
Albert.

El 12/11/12 21:32, Nilay Vaish escribió:
On Mon, 12 Nov 2012, Albert wrote:

Hello there!

I am kind of a newbie with gem5 so this may be a silly question. I am trying to implement a method capable of looking up into the TLBs of the other CPUs in the system in order to resolve a TLB miss. In order to do so, I should be able to communicate with other processors' TLBs when an entry is not found in the local TLB.

The system would be a parallel application running on a FS on an ARM multi-processor system. It seems the src/arch/arm/tlb.xx uses the ThreadContext to retrieve the correct TLB for that thread. I understand that various ThreadContext may be pointing to the same process that is being executed in parallel in various processors. So I guess I should find other ThreadContext in the simulation pointing to the same process in order to obtain their TLBs?

I am still trying to find a suitable approach to do so (to communicate with other TLBs). If someone can point out a proper starting point it would be highly appreciated.


Why not just create a list of all the tlbs when the tlb objects are instantiated?

--
Nilay


_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to