Check  ISCA tutorial. slide  84 and 85 show the flow chart of coherent
cache in marss
http://cloud.github.com/downloads/avadhpatel/marss/Marss_ISCA_2012_tutorial.pdf


On Sun, Feb 9, 2014 at 10:57 PM, alireza nazari <[email protected]>wrote:

> Hello,
>
> I have a question about handling cache coherency protocol.
> 1. As I understand MarssX86 is able to handle coherency in either
> snooping(only MESI) or directory(only MOESI). Right?
>
> 2. In this code:
>
>
>
> void MESILogic::handle_local_hit(CacheQueueEntry *queueEntry)
> {
> switch(oldState) {
> ...
>
> case MESI_SHARED:
>
>             if(type == MEMORY_OP_WRITE) {
>                 if(controller->is_lowest_private()) {
>                     queueEntry->line->state = MESI_MODIFIED;
>                     newState                = MESI_MODIFIED;
>                     queueEntry->sendTo      = queueEntry->sender;
>
>
>                     controller->send_evict_to_lower(queueEntry);
>                     controller->wait_interconnect_cb(queueEntry);
>                 } else {
>
> ...
>
> This core should make sure that the other cores' copy is invalidated and then 
> write the data in cache. I assume this happens in wait_interconnect_cb() 
> (correct?)but I can not understand how. How it broadcast the message? How 
> this core understands that all the other cores already invalidated their 
> copy? is there any cb function or reply message?
>
>
> also when we send eviction to lower(send_evict_to_lower), is there any cb to 
> understand in higher cache that eviction is already done?
>
> I am trying to count the number of cycles that it takes to evict lower caches 
> and it takes to invalidate the other cores shared data.
>
>
>
> Thank you
>  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

Reply via email to