RE: [Gems-users] Regarding checkpoints

[Mike Marty <mikem@xxxxxxxxxxx>]

> "The memory_transaction_t struct, part of the Simics interface, is
> slightly different for x86 and Sparc.  All the Ruby/Simics interface
> code is in $GEMS_ROOT/ruby/simics"
>
> ok, but ruby does not seem to complain when I load it in enterprise even
> though its unmodified. whould it complain? what happens if I dont change
> the interface code?
>

I don't know.  Never tried it.

>
> > Does anyone have any experience with running enterprise configuration
> > together with ruby?
> >
>
> "Yes.  What do you want to know?"
>
>
> Can I conclude that the absence of i-cache statistics from ruby comes
> from the fact that the -stall option is not supported by x86 targets?
>

What exactly do you mean by "enterprise configuration"?  Is this a Simics
target machine like serengeti?  If so, then I do _not_ have any experience
with running enterprise.

By default, Simics will not feed all instruction fetches into a memory
timing simulator unless you tell it to with the following command:

instruction-fetch-mode instruction-fetch-trace

--Mike