|
@Liran_Alon | |||||
|
Surprised by interesting numbers in Intel Optimisation Guide section 2.5.5.2 L1 DCache. Turns out there are only 4 DTLB entires for 1GB pages! Crazy! So mapping all guests memory as 1GB pages in EPT may be less efficient? Worth benchmarking! @_msw_ @Karim_Allah @rsinghal1 (1/2) pic.twitter.com/wmlyFJBp8z
|
||||||
|
||||||
|
Liran Alon
@Liran_Alon
|
17. pro |
|
I also wonder how these entries are separated between "translation entries" and "paging-structure-cache entries". I didn't find this information specified anywhere. I do hope these numbers don't include the latter and that there are additional entries for that somewhere. :) (2/2)
|
||
|
|
||
|
Paolo Bonzini
@fagiolinux
|
17. pro |
|
4GB is still more than what the 2MB DTLB entries can provide though. Also EPT page size can make TLB entries bigger but the size for a combined translation depends anyway on the page size specified by the guest. So you never know but it would be very surprising.
|
||
|
|
||
|
Liran Alon
@Liran_Alon
|
17. pro |
|
Right. But note the 1GB TLB entries can hold guest-physical translations while the 4KB/2MB TLB entries hold combined translations. Leading to sufficient combined translations & avoiding EPT walk on combined translation miss. So it's probably OK and my first tweet is misleading :P
|
||
|
|
||