I'm genuinely not sure. Mills certainly speculate instruction just like the next guy but they handle memory access faults differently. Instead of stopping the

I'm genuinely not sure. Mills certainly speculate instruction just like the next guy but they handle memory access faults differently. Instead of stopping the world on a bad read they just tag the data as bad. With speculative execution and stop the world you have to supress stopping the world until the speculation is resolved or risk interrupting good programs because your branch predictor made a mistake. But there's no reason for a Mill to suppress tagging. Loading from invalid data should just short circuit to producing more invalid data instead of actually producing a load so I don't see why there should be data ex-filtration from arbitrary memory locations as in Spectre. And they're certainly not intensely optimizing everything for every last drop of performance in the way Intel did to suffer from Meltdown. I wouldn't be surprised if there was some sort of Spectre-esque speculative attack you could run against current Mill designs to get some data out but probably not from arbitrary memory locations? But all the above is pure speculation.