<jeffro256> sech1: quick implementation question: the cache and dataset are byte-for-byte the same after V2 right ? In other words, I don't need to do any special invalidation because of V2, correct ?

<jeffro256> The prefetching is different as I understand it , but that doesn't affect the data itself ?

jeffro256 yes, cache and dataset are the same. You only need to set the RANDOMX_FLAG_V2 for RandomX virtual machines when you switch to v2

cache and dataset are the same

the superscalar hash is not changed (but ofc it gets called more in light mode)

just to be consistent one could put the hardfork threshold (initial one that changes PoW function) on a randomx epoch change

the the changes are all within VM code and parameters

the even

No need to put it on RandomX epoch change. I even added logic to XMRig to not update the dataset if it's a switch to v2 and it's not an epoch change

yeah, if you have the cache/dataset pointers around they can be passed to the VM creation and that'd just work (somehow this message wasn't seen)

Don't even need to create a new vm, see "test_switch" in src/tests.cpp

I mean for monero code itself and verification, not external miners