On Thu, Jun 16, 2011 at 7:34 AM, Carlos José Martinez Gavilán <
[email protected]> wrote:
> Hi,
>
> I still do not understand it well. I want to, for example, a simulation with
> 4 cores ooo with8MB L2 cache associativity and 8, the other parameters are
> indifferent. Could you tellme what commands should I use to compile and
> run correctly and what configuration file should I create.
>
> In default.conf file add a new cache that extends l2_2M_mesi like below:
cache:
l2_8M_mesi:
base: l2_2M_mesi
params:
SIZE: 8M
ASSOC: 8
Now in 'private_L2' machine configuration in default.conf replace
'l2_2M_mesi' with 'l2_8M_mesi'.
My other question is this. Before I use the new configuration I tried with
> this instructionQemu-Q c = 4 and run the simulation (run-stopinsns
> simconfig-100m-kill-after-run).
> The problem is that now with the same instruction and without using the
> config file, the same instruction takes much longer to complete. Why? I do
> something wrong? To use the new version I have only replaced marss required
> folders.
>
> In my simulations I haven't seen any slowdown. If you have compiled with
'debug=1' option then it will be slower then default.
- Avadh
> I'm sorry if my questions are silly and I hope I can help.
> Thanks.
>
> 2011/6/2 avadh patel <[email protected]>
>
>>
>> On Thu, Jun 2, 2011 at 3:38 AM, Carlos José Martinez Gavilán <
>> [email protected]> wrote:
>>
>>> Thanks for this new configuration.
>>>
>>> Can you explain me how to include 4 cores OoO in the configuration?
>>>
>>> Just use configuration 'private L2' and compile with 'c=4' option. If you
>> want 'shared L2' configuration then in 'machine:cores:' remove Atom
>> declaration.
>> The configuration mechanism is flexible in number of cores we simulate, so
>> same configuration can be used for different number of cores.
>>
>> In each 'machine' configuration 'min_context' and 'max_context' set the
>> limit of number of cores you can simulate. If 'max_context' is not specified
>> then user can use any number of cores.
>>
>> - Avadh
>>
>>
>>> Thanks.
>>>
>>>
>>> 2011/5/26 avadh patel <[email protected]>
>>>
>>>>
>>>> On Thu, May 26, 2011 at 3:57 AM, Timothy Hayes <[email protected]> wrote:
>>>>
>>>>> Nice, I think this makes things easier for the user.
>>>>>
>>>>> I was trying to run a test with the core-models branch and the
>>>>> simulation never ended. It terminates without problem on the master
>>>>> branch. I can provide additional details / debugging information, just
>>>>> let me know what you need.
>>>>>
>>>>> Thats interesting. If possible can you provide the binary so I can run
>>>> the tests.
>>>> Also in 'core-models' branch which core model did you test? Atom or OoO?
>>>> If it is OoO model, can you test it with qemu-0.14 branch also?
>>>>
>>>> Thanks for trying out new configuration mechanism.
>>>>
>>>> - Avadh
>>>>
>>>> Cheers
>>>>> Tim
>>>>>
>>>>> On 24 May 2011 21:19, avadh patel <[email protected]> wrote:
>>>>> > Hi All,
>>>>> > For a while we have been working on new and easy way for users to
>>>>> configure
>>>>> > their simulation models without hacking the code. We have finished
>>>>> this
>>>>> > feature
>>>>> > and its now available in 'core-models' branch.
>>>>> > Configuration mechanism is explained in detail over here:
>>>>> > http://marss86.org/index.php/Machine_Configuration
>>>>> > This new configuration mechanism will allow users to configure:
>>>>> > - Cache parameters and type of cache
>>>>> > - Core type (easily allow to design heterogeneous core configuration)
>>>>> > - Connection between cores and caches and interconnects
>>>>> > 'core-models' branch has couple of configuration files available in
>>>>> 'config'
>>>>> > directory.
>>>>> > Please try out this new configuration mechanism and let us know your
>>>>> > comments
>>>>> > or suggestions.
>>>>> > Regards,
>>>>> > Avadh
>>>>> > _______________________________________________
>>>>> > 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
>>>>
>>>>
>>>
>>>
>>> --
>>> Carlos J. Martínez
>>>
>>>
>>
>
>
> --
> Carlos J. Martínez
>
>
_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
