Hi folks. I have seen a couple series of changes adding support for the PCID feature in x86, as far as I can tell a new feature in the ISA (within the last year?) which lets you set a context ID for mappings.
While I don't think there's anything wrong with adding support for this feature, I'd like to know if we're doing that just to support it (which is valid), or if there's some mechanical problem in gem5 it's supposed to solve. I ask because I remember seeing something talking about how this was necessary for having multiple processes or SMT in x86, and I don't think that's true. This will make those cases perhaps perform better since the TLB won't have to be flushed as aggressively if processes are moved around between contexts (which doesn't usually happen in SE mode), but the CR3 for each (the root page table pointer) should be different since each should have its own set of page tables, and so everybody should see their own view of virtual memory without any extra help. Please fill in any extra context so that we can be sure we're fixing the right problem, and not just side stepping a deeper issue with this. To be clear, having this feature for its own sake is fine, I just want to make sure we know we're doing that if we are! Gabe
_______________________________________________ gem5-dev mailing list -- [email protected] To unsubscribe send an email to [email protected] %(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
