4 ms·
> get to do a dataless read-for-ownership, so they actually move 1/3 less data around than any other implementation when they miss L2. Can't you achieve same u
by vardump 9y ago
> get to do a dataless read-for-ownership, so they actually move 1/3 less data around than any other implementation when they miss L2.
Can't you achieve same using cache line sized non-temporal stores?
- stephencanon 9y agoThere are no cacheline-size store instructions pre AVX-512 (on Ivybridge, the widest store is 32B, and even that cracks to two 16B store uops; it's a real 32B store on Haswell+). Also the NT stores suffer from a bunch of other hazards that rep/movs avoids.
- gpderetta 9y agoNon temporal stores shouldn't go to the cache at all though. They go to write combining buffers instead.
- stephencanon 9y agoShouldn't, but do. They RFO into the LLC.