# Ajeya feedback checks ## Abstract - [x] Split in two paragraphs ## Introduction - [x] Add paragraph discussing the techniques proposed by prior work AFTER (?) addressing bottlenecks - [x] BEFORE addressing bottlenecks: "the first requirement for evaluating and improving graph processing is to design a set of representative tools and benchmarks. Beamer et al. propose the most widely used benchmarks suite called GAP." - [x] AFTER bottlenecks: "Researchers have addressed these bottlenecks from two directions---first by proposing a better re-organisation of the graph...one sentence about reordering and one about partitioning...and second by designing techniques that directly address the memory latency bottleneck by designing prefetchers in hardware or software." - [x] We should reference individual sectionsabove. Like Section II covers the terminologyused in the graph processing and the layoutof the graphs. Section III does...Section IVcovers...Then we motivate the need for thisstudy in Section III – Ajeya ## Background - [x] May be the term 'Natural' is not required? ## Motivation - [ ] Define MLP before using the term. - [x] All the acronyms like ROB, LLC, OoO, etc. must be defined for the first time. - [x] UNCLEAR: "(pseudo-)sequential execution in OoO processors" - [x] UNCLEAR: "However, prefetching can be avoided given that data that will be reused is not evicted from the cache"