Re:Zen 2 Missives – AMD now delivering efficiencies that are...
by monocasa on Monday August 26, 2019 @01:30PM
(#20801844)
Are their L3 slices MOESI like their L2's are (or at least were). That'd let you have multiple copies in different slices as long as you weren't mutating them.
[ Parent
| on HN ]
Re:Zen 2 Missives – AMD now delivering efficiencies that are...
by dragontamer on Monday August 26, 2019 @01:36PM
(#20801899)
AMD is using MDOEFSI, according to page 15 of: https://www.hotchips.org/wp-content/uploads/hc_archives/hc29...
However, I can't find any information on what MDOEFSI is. I'm assuming:
* Modified * Dirty * Owned * Exclusive * Forwarding * Shared * Invalid
Any information I look up comes up to an NDA-firewall pretty quickly (be it in performance counters, or hardware level documentation). It seems like AMD is highly protective of their coherency algorithm.
> That'd let you have multiple copies in different slices as long as you weren't mutating them.
Seems like the D(irty) state allows multiple copies to be mutated actually. But its still a "multiple copies" methodology. As any particular core comes up to the 8MB (Zen) or 16 MB (Zen2) limit, that's all they get. No way to have a singular dataset with 32MB of cache on Zen or Zen2.