Maybe it should be separated from the simple atomic CPU again? That
would have made sense back when they were mostly the same, but they
might be different enough now that it just adds to the confusion. I'm
not advocating that necessarily, but it sounds like it could help.
Gabe
Quoting Steve Reinhardt <[email protected]>:
It could use some cleanup. The idea has always been that it's the
simplest CPU that can drive the timing memory system, so there's no
guarantee it will be simple in an absolute sense. But I'm sure it
could be simpler if we worked at it. I think the combination of
dealing with unaligned accesses and split-phase translations has
definitely hit it hard.
Steve
On Sat, Jun 5, 2010 at 1:45 PM, Ali Saidi <[email protected]> wrote:
My comment was more that we might want to re-evaluate the timing
simple cpu as it's left the simple bit a while ago. Many of the
fault == NoFault checks don't matter with the translation state
change since it's always going to return no fault.
I was hoping to get a discussion going about what if anything we should do.
Ali
On Jun 4, 2010, at 5:29 PM, nathan binkert wrote:
On a somewhat related note, it seems like the timing simple cpu control
flow has gotten pretty messy with the addition of the translation state
code. To figure out what is going on you really need a large white board
and a lot of time to map out the call tree and what happens on faults and
the like.
Ali,
If it is correct, can you commit the patch? If you're not the right
person, Gabe?
Nate
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev
