Thank you, Ali. It seems that this is the reason, because I always found
that the two instruction when they are memory instructions (I am only
interested in this type) have exactly the same access address. I think I
need to consider this issue in my statistics. Thanks.

On Sun, Nov 14, 2010 at 8:15 PM, Ali Saidi <[email protected]> wrote:

> You would have to provide us with at Exec trace of it happening to be sure,
> but the only reason that comes to mind is if the instruction takes a fault
> or is a syscall, the atomic cpu might execute the faulting instruction and
> the first instruction of the fault handler in the same cycle.
>
> Ali
>
> On Nov 14, 2010, at 12:24 AM, Veydan Wu wrote:
>
> > Hi, I am using the atomic CPU in alpha FS mode. Because I need to do some
> statistics based on instruction count, two instructions in the same "tick"
> makes some instruction count be 0. Could someone tell me why it occurs given
> that the "width" parameter is "1" by default? thanks.
> >
> > --
> > weidan
> > _______________________________________________
> > m5-users mailing list
> > [email protected]
> > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>
> _______________________________________________
> m5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>



-- 
Regards,

Weidan Wu
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to