Re: [Gems-users] Gems Questions

["Fuad Tabba" <fuad@xxxxxxxxxxxxxxxxx>]

Thanks Jayaram, that was informative.

A (maybe trivial) question; how do I call SIMICS_remove_timing_model() from inside a simics script?

Cheers,
/Fuad

On Thu, Jul 10, 2008 at 11:08 AM, Jayaram Bobba <bobba@xxxxxxxxxxx> wrote:

Fuad Tabba wrote:
> Hi,
>
> Was wondering if anyone could help me with any of the following:-
>
> - Are SIMICS_BEGIN_BARRIER or SIMICS_END_BARRIER important for a
> simulation to run properly? Or are they just used for giving better
> tracing information. By inspecting commands.C it seems to me that
> they're only used for visualization, but without them LogTM acts up!

They are just used for collecting some stats/visualization. It is
strange that they impact your simulations.

> - I found that disabling interrupts on the running processors (using
> psrset -f) helps with the simulation running smoother.  Is there good
> reason why I should *not* be doing that?
>

It might be ok to disable interrupts while you are running
microbenchmarks and small programs. For larger workloads and workloads
with system activity, the OS might freak out if some interrupts (like
TLB shootdowns) are not handled in a timely manner.

> - I couldn't find a difference between using processor_bind or
> pset_bind. Looking at the man pages they seem to do the same thing. I
> prefer processor_bind because it doesn't require root access, so I can
> use it on the real sparc machine I run the software benchmarks on
> without having to go through the hassle of asking for the right
> privileges. Is there a reason why creating cabinets and binding them
> with pset_bind is preferable?

http://developers.sun.com/solaris/articles/solaris_processor.html
I think pset_bind gives stronger guarantees about non-interference from
OS threads...

>
> - Is there a way to unload ruby? Or at least speed it up (while
> sacrificing statistics or something)? The reason I'm asking is that
> after running the benchmarks I normally like to run some sanity checks
> and collect statistics from the data structures. This step takes place
> on only one processor, and I don't have any need for the ruby features
> at that point. Currently this step takes a lot of time (almost 25%).
> Is there a way I could speed it up?

https://lists.cs.wisc.edu/archive/gems-users/2006-September/msg00008.shtml

Jayaram
_______________________________________________
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.