Modeling a shared prefetcher between multiple cores

[Akilesh Kannan ee18b122]

Hi,

I am trying to model a global prefetcher for TLBs, which is shared between multiple cores. How do I implement this in sniper? I'd like to know how to implement it for caches and then try to extend it to TLBs from there. I took a look at the cache_cntlr.cc file in parametric_dram_directory_msi folder where the prefetcher related functions — trainPrefetcher(), Prefetch(), doPrefetch() — are declared, but I am not sure how to modify them.

Any help in this regard would be highly beneficial to me.

Thanks,

Akilesh