[PTLsim-devel] validate option

Komal Jothi
Mon Oct 15 16:57:17 EDT 2007


Hi,

I tried to call the "execute_sequential" function after the
uop.eomcondition is true inside
ooopipe.cpp, as below:

int tmp_status = execute_sequential(ctx, tmp_ptr, 1, 1);
// (tmp_ptr is initialized to null).

The program aborted with a "Exception 14 (page fault)" message when I tried
to execute.

I am not sure how "bbcount" should be passed to the execute_sequential
function from the ooo core.
Can you please help me in this regard?

Komal

On 10/11/07, Matt T. Yourst <yourst at peptidal.com> wrote:
>
> On Thursday 11 October 2007 18:29, Komal Jothi wrote:
> > Hi,
> >
> > Can you please give me the details on how to use the 'validate' option
> in
> > PTLsim?
> >
> > My understanding is that, if a benchmark is run with the -validate
> option,
> > PTLsim would compare the ooo core results with the seq core results,
> during
> > run time.
> >
> > Is there a log file that would clearly indicate the results being
> committed
> > by each core?
> >
> > I could not find much documentation on the -validate parameter in the
> > manual.
> >
>
> This feature isn't implemented in the open source version of PTLsim. We
> use it
> with one of our internally developed core models - it probably shouldn't
> have
> even been included since it isn't actually used anywhere.
>
> You could try to implement this feature for ooocore if you'd like to use
> it.
> The seqcore model has an execute_sequential function like this:
>
> int execute_sequential(Context& ctx, CommitRecord* cmtrec,
>   W64 bbcount, W64 insncount)
>
> The CommitRecord structure records the register state and list of stores
> that
> would be committed if the sequential core was allowed to write to the
> architectural state. When the OOO core commits each x86 instruction in the
> commit stage, you could call execute_sequential for exactly one
> instruction,
> then allow the OOO commit to occur, then compare the architectural state
> against execute_sequential's output. This should be fairly easy to add.
>
>
> - Matt
>
> -------------------------------------------------------
> Matt T. Yourst                    yourst at peptidal.com
> Peptidal Research Inc., Co-Founder and Lead Architect
> -------------------------------------------------------
> _______________________________________________
> PTLsim-devel mailing list
> PTLsim-devel at ptlsim.org
> https://ptlsim.org/mailman/listinfo/ptlsim-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://ptlsim.org/pipermail/ptlsim-devel/attachments/20071015/e4218e35/attachment.html 


More information about the PTLsim-devel mailing list