Differences

This shows you the differences between two versions of the page.

Link to this comparison view

seminars:seminar_12_10_02 [2017/09/20 22:02] (current)
Line 1: Line 1:
 +======The Evicted-Address Filter: A Unified Mechanism to Address Both Cache Pollution and Thrashing======
  
 +Tuesday Oct. 2, 2012\\
 +Hamerschlag Hall 1107\\
 +4:30pm\\
 +
 +{{:​seminars:​vivek.jpg|}}
 +
 +**Vivek Seshadri (CMU)**\\
 +
 +=====Abstract=====
 +Off-chip main memory has long been a bottleneck for system per-
 +formance. With increasing memory pressure due to multiple on-
 +chip cores, effective cache utilization is important. In a system with
 +limited cache space, we would ideally like to prevent 1) cache pol-
 +lution, i.e., blocks with low reuse evicting blocks with high reuse
 +from the cache, and 2) cache thrashing, i.e., blocks with high reuse
 +evicting each other from the cache.
 +
 +In this talk, we present a simple mechanism to predict the
 +reuse behavior of missed cache blocks in a manner that mitigates
 +both pollution and thrashing. Our mechanism tracks the addresses
 +of recently evicted blocks in a structure called the Evicted-Address
 +Filter (EAF). Missed blocks whose addresses are present in the EAF
 +are predicted to have high reuse and all other blocks are predicted to
 +have low reuse. The key observation behind this prediction scheme
 +is that if a block with high reuse is prematurely evicted from the
 +cache, it will be accessed soon after eviction. We show that an EAF-
 +implementation using a Bloom filter, which is cleared periodically,​
 +naturally mitigates the thrashing problem by ensuring that only a
 +portion of a thrashing working set is retained in the cache, while
 +incurring low storage cost and implementation complexity.
 +
 +We compare our EAF-based mechanism to five state-of-the-art
 +mechanisms that address cache pollution or thrashing, and show
 +that it provides significant performance improvements for a wide
 +variety of workloads and system configurations.
 +\\
 +
 +=====Bio=====
 +Vivek Seshadri is a fourth year PhD student in the Computer Science
 +Department. He is advised by Prof. Todd Mowry and Prof. Onur Mutlu.
 +His research interests are in improving the performance of memory
 +subsystems. Vivek received his B.Tech degree in Computer Science from
 +Indian Institute of Technology Madras in 2009.
 +\\
 +\\
 +**[[seminars| Back to the seminar page]]**