The amount of memory per board in Simics affects what valid physical
address ranges are, but *Simics provides no memory timing whatsoever*.
Hence, Ruby's timing results are valid regardless of where memory
resides on Simics's "boards" in absentia of NUMA-aware memory allocation.
I cannot comment on how to properly set up a bagle machine -- we use
sarek locally. I do know that Simics is rather fussy about what
constitutes a "valid" configuration.
Regards,
Dan
Lei Yang wrote:
Dear list,
I was trying to set up a CMP system with 4 processors. I first created
a checkpoint with bagle using bagle-4p.simics. Then looking at that
file, I found their set up a bit weird:
# set up 4 processors with 256MB
@boards = [[0, 2, 256], [2, 2, 0]]
# the rest is common for all bagle machines
run-command-file "bagle-common.simics"
It seems that the 4 cores are on two boards and 256MB memory is
created only on the first board. In my ruby configuration, I set the
number of processors to 4, and number of processor on each chip to 1.
The memory size is 4GB. I'm wondering, does the different Simics set
up affect Ruby result? Or is it true that as along as the number of
processors is set correctly, the rest of memory system does not matter
because Ruby just receives instructions from the processors?
I'd appreciate it if you could provide a Simics set up example if the
above isn't correct.
Thanks in advance!
Lei
------------------------------------------------------------------------
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.
--
http://www.cs.wisc.edu/~gibson [esc]:wq!
|