Quoting Steve Reinhardt <[email protected]>:
On Monday, September 26, 2011, Gabe Black wrote:
My plan is for this to only be a temporary thing while the code still
expects full system to be a global on/off setting. This will help to
keep track of the first stage of the transition, making sure FULL_SYSTEM
isn't used for conditional compilation. Ultimately, I expect SE vs. FS
modes to be mostly reflected in properties of the system objects and
what's plugged into them, and then that extending into the decoders
which will return instructions which behave correctly for whichever the
current mode is.
I agree with this part (as we already covered in the other thread), but I
still think we should name this variable properly now, so maybe all we have
to do in the future is stick a 'system->' in front of it.
I have no problem renaming it, I just don't think it's important
enough to put together a new change since I essentially just renamed
it to make bookkeeping easier. If you want to change it, feel free to
make it whatever you like. It just needs to not show up if I grep for
FULL_SYSTEM.
At that point the distinction between SE and FS will be
mostly conceptual, and no global constant, variable, or function will be
necessary. I wouldn't mind it if it wasn't even a specific setting on
System objects or the decoders.
You lost me here... seems like the "SE vs FS" bit will still need to be
stored somewhere, and that location will still need a name...
There should be no SE vs. FS bit, just like there's no UP vs MP bit.
It should just be an emergent property of how the simulation is
configured, ie. a process and system calls emulated, or an OS and not.
Then you can have fuzzy lines like processes with devices, OSes with
emulated BIOS calls, etc.
Gabe
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
