PPT-Orchestrated Scheduling and Prefetching for GPGPUs

Adwait Jog Onur Kayiran Asit Mishra Mahmut Kandemir Onur Mutlu Ravi Iyer Chita Das Multithreading Caching Prefetching Main Memory Improve Replacement Policies Parallelize your code. Ankit Sethia. 1. , Ganesh Dasika. 2. , . Mehrzad. Samadi. 1. ,. Scott Mahlke. 1. 1-University of Michigan. 2-ARM R&D Austin. 1. Introduction. 2. High Throughput – 1 . TeraFlop. . High Energy Efficiency. Tomofumi Yuki . . INRIA . Rennes. Antoine . Morvan. . . ENS . Cachan. Bretagne. Steven . Derrien. . University of Rennes 1. Memory Optimizations. Memory Wall. Adwait Jog. , Onur Kayiran, Asit Mishra, Mahmut Kandemir, Onur Mutlu, Ravi Iyer, Chita Das . Multi-threading. Caching. Prefetching . Main Memory. . Improve. . Replacement . Policies. Parallelize your code!. Enhancement . for Accurate Stream Prefetching. Gang Liu. 1. , . Zhuo. Huang. 1. , . Jih. -Kwon Peir. 1. , . Xudong. Shi. 2. , Lu Peng. 3. 1. University of Florida . 2. Google Inc. 3. Louisiana . State . Seth Pugsley. Predicting the Future. Where have we seen prediction before?. Does it always work?. Prefetching is prediction. Predict which cache line will be used next, and place it in the cache before it is used. Group Report. Ken Ayoob Dean, College of Arts, Humanities & Social Sciences . . (CAHSS). Richard Bruce CAHSS . Office Manager. , Former Psych Department . . Coordinator & University Senator. Ankit Sethia. 1. , Ganesh Dasika. 2. , . Mehrzad. Samadi. 1. ,. Scott Mahlke. 1. 1-University of Michigan. 2-ARM R&D Austin. 1. Introduction. 2. High Throughput – 1 . TeraFlop. . High Energy Efficiency. . Smruti. R. . Sarangi. Contents. Motivation for Prefetching. Simple Schemes. Recent Work. Proactive Instruction Fetching. Return Address Stack Directed Prefetching. Pentium 4 Trace Cache. University of Virginia. April . 26, . 2016. COMPUTER ARCHITECTURE . CS 6354. Prefetching. The content and concept of this course are adapted from CMU ECE 740. AGENDA. Logistics. Review . from last . lecture. Smruti. R. Sarangi. Data Prefetching. Instead of . instructions . let us now . prefetch. . data. Important distinction. Instructions. are fetched into the in-order part of the OOO pipeline. Data. is fetched into the OOO pipeline. . Smruti. R. . Sarangi. Contents. Motivation for Prefetching. Simple Schemes. Recent Work. Proactive Instruction Fetching. Return Address Stack Directed Prefetching. Pentium 4 Trace Cache. Exploiting Unique Opportunities. Ahmad . Lashgar. and . Amirali. . Baniasadi. ECE Department. University of Victoria. 1. Overview. Prefetching Challenges:. Prefetching accuracy. Timeliness. Bandwidth control. Tomofumi Yuki . . INRIA . Rennes. Antoine . Morvan. . . ENS . Cachan. Bretagne. Steven . Derrien. . University of Rennes 1. Memory Optimizations. Memory Wall. Results and Awards. Seth . Pugsley. Thanks!. Big thanks to Hyesoon Kim and the Program Committee. Babak. . Falsafi. Mike . Ferdman. . Aamer. Jaleel. Daniel Jiménez Calvin Lin . Moin. Qureshi.