More on this:
Simics in "fast mode" is now VERY fast relative to simics with a timing
model (like Ruby and/or Opal). Read: no longer a 15x difference. This
applies to Simics 2.2.19 in "fast mode" and 3.0.X in default (non-stall)
mode.
Are you starting ocean from the beginning? It might be wise to only run
Ruby for the "parallel" section of opal -- a MAGIC_BREAKPOINT inserted
right after the thread_fork() calls would be an ideal way to signal the
start of the section.
Regards,
Dan
Daniele Bordes wrote:
Hi all. I am trying to run some simulations of Ocean Benchmark over a
Solaris/Sparc simulated machine (8 processors, 1GB of Ram) using Simics + Ruby.
The configuration of the simulated machine was taken from donut directory.
The host machine (over which Simics runs) is an Intel XEON 3.40 Ghz
with 4GB of RAM. The problem is that 18 hours where sufficient only to
obtain the first two lines of Multigrid Output (total is 13 lines) of
Ocean, which was launched with 66x66 grid dimension and only one
thread.
Instead, using only Simics without Ruby, the simulation of Ocean
finishes in less than three minutes.
Well, did anybody simulate Ocean using Simics+Ruby? How did it take to
complete the simulation?
_______________________________________________
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.
|
![[Computer Systems Lab]](https://www-auth.cs.wisc.edu/pics/csl_logo.gif){kind=logo}