Thank you for the reply! Interesting about parallel misses handling, thanks! One worry is that this tends to compound other effects -- say, non-prefetch-frien

Thank you for the reply! Interesting about parallel misses handling, thanks! One worry is that this tends to compound other effects -- say, non-prefetch-friendly access combined with TLB misses resulting in increasingly expensive slowdowns (as in the continuous-vs.-random array access example in the paper).