If you're interested in non-speculative accesses to the memory hierarcy you
could try Cachegrind. It will be very fast relative to a MARSS simulation.
I'm not sure if they have multicore support yet.

On 3 August 2012 17:39, avadh patel <[email protected]> wrote:

>
>
> On Thu, Aug 2, 2012 at 9:42 PM, Alireza Haghdoost <[email protected]>wrote:
>
>> Hello
>>
>> I was wondering if it is possible to only simulate memory hierarchy
>> (caches + mem controller) and use emulation mode to execute instructions ?
>> I am not interested in cycle accurate processor statistics. Is it possible
>> to reduce simulation time by turning off cycle accurate OoO processor
>> simulation.
>>
>> Its not possible at this point.  You can use 'atom' core model which is
> about 1.5x faster than OoO.  But keep in mind that memory interactions with
> OoO are much different than in in-order mode because of out-of-order issues
> and issues from wrong path.  It always depends on the type of benchmark but
> still its an important factor to consider behavior of caches, specially L1
> caches.
>
> - Avadh
>
>
>> Thanks
>> Alireza
>>
>> _______________________________________________
>> http://www.marss86.org
>> Marss86-Devel mailing list
>> [email protected]
>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>>
>>
>
> _______________________________________________
> http://www.marss86.org
> Marss86-Devel mailing list
> [email protected]
> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>
>
_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel

Reply via email to