The research side may not be the main culprit, hence I said "compiler engineering" - meaning usable tools delivered to the hands of developers. Though I think t

The research side may not be the main culprit, hence I said "compiler engineering" - meaning usable tools delivered to the hands of developers. Though I think the research side has not focused on this enough either. Regarding the Milk paper specifically, it appears to be about reordering and clustering memory accesses without changes to data layout. This is a fine thing but I think it reinforces my point about lack of attention to memory/data layout optimizations.