I should probably add that the Consumers of the Ports are initialized
in /mem/slicc/symbols/StateMachine.py
You can do a search for:
for port in self.in_ports:
expr = "%s" % (port.code)
* if expr.find("MYINPORT") >= 0:
code('${{port.code}}.setController(this);')
# Set the queue consumers
code()
for port in self.in_ports:
code('${{port.code}}.setConsumer(this);')
So when the L1requestToL2 port is being initialized you would pass in
a 2nd pointer to MYINPORT as well, so that whenever the L1requestToL2
port wants to schedule an event, it can reschedule MYINPORT.
You would just need to add something like * to determine when your
MYINPORT gets initialized.
Malek
On Thu, Feb 23, 2012 at 9:47 AM, Malek Musleh <[email protected]> wrote:
> Hi Marti,
>
> In that case, you can simply deschedule/ and then reschedule the port.
> Essentially what happens is that the consumers get scheduled on the
> RubyEventQueue (/mem/ruby/eventqueue/RubyEventQueue.*)
>
> So from an external C structure you can access the scheduled consumer
> using the RubyQueue Ptr
>
> g_eventQueue_ptr->
> descheduleEventAbsolute(m_consumer_ptr,
> (Time X);
>
> where deschedule is a Function that I added (may not be part of the
> latest repo?).
>
> void
> +RubyEventQueue::descheduleEventAbsolute(Consumer* consumer, Time timeAbs)
> +{
> + assert(consumer != NULL);
> + if (consumer->alreadyScheduled(timeAbs)) {
> + assert(timeAbs > getTime());
> + consumer->removeScheduledWakeupTime(timeAbs);
> + }
> +}
> +
>
> Then, you can subsequently call the Schedule Consumer Function:
>
> g_eventQueue_ptr->scheduleEventAbsolute(m_consumer_ptr, ready_time);
>
> Since it sounds like you might be needing to reschedule frequently,
> you can just combine those 2 function calls into another function.
>
> I don't think you can do it directly from the *.sm files, so you would
> need to do it from an external C/C++ structure similar to TBE or
> TimerTable (the latter being easier to replicate). Also you would need
> to initialize the consumer ptrs to your port during the
> constructor/init function.
>
> Does that help?
>
> Malek
>
> 2012/2/23 Martí Torrents Lapuerta <[email protected]>:
>>
>> Hi Malek,
>>
>> According to what I understand in the first solution you suggest the boolean
>> is related to a cache entry, what this will force is that if both ports
>> needs to access the same cache line, only one of them will be able to. What
>> I want is that each cycle only one of the two ports issues a request.
>>
>> The second option seems to work better, the problem there is that if I
>> schedule my_InPort in one cycle, where the L1RequestToL2 consumer is not
>> scheduled, but later on, the L1RequestToL2 consumer is scheduled in a cycle
>> where my_InPort is scheduled, both will be executed again in the same cycle.
>> My question here would be: "In this situation it would be possible to
>> reschedule my_InPort consumer to another cycle?" And how? :D If we find out
>> that my issue would be totally solved!
>>
>> Thanks in advance!
>>
>>
>> 2012/2/23 Malek Musleh <[email protected]>
>>>
>>> Hi Marti,
>>>
>>> I would think that you could use a boolean variable that you defined
>>> in the CacheEntry/DirEntry Structure of the L2, and when set to false,
>>> don't do the desired action.
>>>
>>> For example:
>>>
>>> inport(my_port, RequestMsg, latency="") {
>>> peek(my_port, RequestMsg) {
>>>
>>> bool flag := getEntryFlag(cache_entry, address);
>>>
>>> if(flag) {
>>>
>>> trigger(Event: ........);
>>>
>>> }
>>> // else do nothing
>>> o
>>> }
>>> }
>>>
>>> The getEntryFlag Function could be something you define similar to the
>>> isLocalOwnerValid Function.
>>>
>>> But in this case, the port is still scheduled and woken up every
>>> cycle, but you just don't execute anything.
>>>
>>> If you truly don't want the port to be scheduled for wakeup, it is a
>>> little more tricky. Essentially each in_port inherits from the
>>> Consumer Class (/ruby/common/Consumer.hh)
>>>
>>> What you can do, is check to see if the L1RequestToL2 in_port
>>> (consumer) is scheduled for some said cycle using
>>> the bool alreadySchedule(Time) function, and if not, then schedule
>>> your port (consumer) to wake up.
>>>
>>> Does that make sense?
>>>
>>> Malek
>>>
>>> 2012/2/22 Martí Torrents Lapuerta <[email protected]>:
>>> > Hi Nilay!
>>> >
>>> > That seems a good point :D But how can I do that?
>>> >
>>> > Thanks!
>>> >
>>> >
>>> > 2012/2/22 Nilay Vaish <[email protected]>
>>> >>
>>> >> Why can you not use a boolean variable to track whether or not the L1
>>> >> cache sent a request to the L2 cache in this cycle?
>>> >>
>>> >> --
>>> >> Nilay
>>> >>
>>> >>
>>> >> On Wed, 22 Feb 2012, Martí Torrents Lapuerta wrote:
>>> >>
>>> >>> Hy Malek,
>>> >>>
>>> >>> Thanks for your answer, but I think that this parameter (rank) is only
>>> >>> used
>>> >>> when 2 inports are going to access the same address. In my case I want
>>> >>> to
>>> >>> issue only one of the two InPorts per cycle. Do you know any other
>>> >>> way?
>>> >>>
>>> >>> Thanks for your attention.
>>> >>>
>>> >>> 2012/2/22 Malek Musleh <[email protected]>
>>> >>>
>>> >>>> Hi Marti,
>>> >>>>
>>> >>>> I believe you would need to specify the rank of the inport.
>>> >>>>
>>> >>>> For example:
>>> >>>>
>>> >>>> // Response Network
>>> >>>> in_port(responseNetwork_in, ResponseMsg, responseToL1Cache, rank=2)
>>> >>>> {
>>> >>>>
>>> >>>> You would then need to define the priority for each of the in_ports.
>>> >>>>
>>> >>>> Take a look at the MOESI_CMP_Token protocol as reference.
>>> >>>>
>>> >>>> Malek
>>> >>>>
>>> >>>> 2012/2/22 Martí Torrents Lapuerta <[email protected]>:
>>> >>>>>
>>> >>>>> Hello everybody,
>>> >>>>>
>>> >>>>> I am working with the MOESI_CMP_directory-L2cache protocol in Ruby.
>>> >>>>> I
>>> >>>>
>>> >>>> have
>>> >>>>>
>>> >>>>> just added a new input port with some requests that I would like to
>>> >>>>
>>> >>>> inject
>>> >>>>>
>>> >>>>> in L2, however I only want to process this requests, when there is a
>>> >>>>
>>> >>>> cycle
>>> >>>>>
>>> >>>>> where the L1 is not sending any request to the L2. So myInPort must
>>> >>>>> have
>>> >>>>> less priority than the L1requestNetwork_in InPort. I do not know how
>>> >>>>> to
>>> >>>>
>>> >>>> do
>>> >>>>>
>>> >>>>> this, Is there anybody who could help me with this issue?
>>> >>>>>
>>> >>>>> Thanks
>>> >>>>>
>>> >>>>> Marti
>>> >>>>>
>>> >>>>>
>>> >>>>>
>>> >>>>>
>>> >>>>> _______________________________________________
>>> >>>>> gem5-users mailing list
>>> >>>>> [email protected]
>>> >>>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>> >>>>
>>> >>>> _______________________________________________
>>> >>>> gem5-users mailing list
>>> >>>> [email protected]
>>> >>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>> >>>>
>>> >>>
>>> >>>
>>> >>>
>>> >>> --
>>> >>> Martí Torrents Lapuerta
>>> >>> [email protected]
>>> >>> (+34)649 65 10 57
>>> >>
>>> >>
>>> >> _______________________________________________
>>> >> gem5-users mailing list
>>> >> [email protected]
>>> >> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>> >
>>> >
>>> >
>>> >
>>> > --
>>> > Martí Torrents Lapuerta
>>> > [email protected]
>>> > (+34)649 65 10 57
>>> >
>>> > _______________________________________________
>>> > gem5-users mailing list
>>> > [email protected]
>>> > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>> _______________________________________________
>>> gem5-users mailing list
>>> [email protected]
>>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
>>
>>
>>
>>
>> --
>> Martí Torrents Lapuerta
>> [email protected]
>> (+34)649 65 10 57
>>
>> _______________________________________________
>> gem5-users mailing list
>> [email protected]
>> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
