Dave,
Multifacet does not distribute its workloads due to complications with
proprietary software. Others have been successful in reproducing
similar workloads independently, however, and some may be found online.
(Note apache is bundled with many *nix distributions)
The SPLASH suite website: http://www-flash.stanford.edu/apps/SPLASH/
The preferred method is usually to compile the workloads on a real
machine and then move them to the target. This is to prevent very long
*simulated* compile-times. Your choice of static vs. dynamic linkage
should reflect the libraries available on your simulated target.
Finally, since GEMS uses Simics to model *full-system* simulation, by
default GEMS will collect statistics from combined User/OS execution.
However, if you want to collect data from user-space only (assuming you
are NOT using Opal), I recommend you modify Ruby's commands.C (I
believe the relevant function there is operate_memory() ) to check
memory references (memory_transaction_t / v9_memory_transaction_t) for
their "priv" bit. (See the Simics Programming Guide, Simics Target
Guide: Serengeti, and Simics Reference Manual). If the bit is set,
simply return from the appropriate function immediatially instead of
calling any other Ruby functionality -- this will artificially make all
OS functionality invisible to Ruby. Be sure to return "zero", to
prevent the calling Simics processor from stalling.
Good luck.
Regards,
Dan
Dave wrote:
Hello,
I have a few questions on the workloads explained at
http://www.cs.wisc.edu/gems/doc/wiki/moin.cgi/gen-scripts. The
workloads are not distributed in the GEMS release, but where can we
obtain them? We need to compile the workloads on a real Sun machine and
then simulate them, right? Are shared libraries okay or do we need to
statically link the programs? How can we isolate the program simulation
if we don't want to get metrics related with the OS?
Thank you for help.
Dave
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
|
|