logo

SCIENCE CHINA Information Sciences, Volume 61, Issue 1: 012105(2018) https://doi.org/10.1007/s11432-016-0408-1

Toward multi-programmed workloads with different memory footprints: a self-adaptive last level cache scheduling scheme

More info
  • ReceivedJun 15, 2016
  • AcceptedSep 21, 2016
  • PublishedJul 14, 2017

Abstract

With the emerging of 3D-stacking technology, the dynamic random-access memoryDRAM) can be stacked on chips to architect the DRAM last level cacheLLC). Compared with static random-access memorySRAM), DRAM is larger but slower. In the existing research papers, a lot of work has been devoted to improving the workload performance using SRAM and stacked DRAM together, ranging from SRAM structure improvement, to optimizing cache tag and data access. Instead, little attention has been paid to designing an LLC scheduling scheme for multi-programmed workloads with different memory footprints. Motivated by this, we propose a self-adaptive LLC scheduling scheme, which allows us to utilize SRAM and 3D-stacked DRAM efficiently, achieving better workload performance. This scheduling scheme employs (1) an evaluation unit, which is used to probe and evaluate the cache information during the process of programs being executed; and łinebreak (2) an implementation unit, which is used to self-adaptively choose SRAM or DRAM. To make the scheduling scheme work correctly, we develop a data migration policy. We conduct extensive experiments to evaluate the performance of our proposed scheme. Experimental results show that our method can improve the multi-programmed workload performance by up to 30% compared with the state-of-the-art methods.


Acknowledgment

This work was supported by National Basic Research Program of China (973 Program) (Grant No. 2015CB352403), National Natural Science Foundation of China (Grant Nos. 61261160502, 61272099, 61303012, 61572323, 61628208), Scientific Innovation Act of STCSM (Grant No. 13511504200), EU FP7 CLIMBER Project (Grant No. PIRSES-GA-2012-318939), and CCF-Tencent Open Fund. We would like to acknowledge the anonymous reviewers for their careful work and instructive suggestions. Also, we thank Dr. Zhi-Jie Wang for his warm help and advices.


References

[1] Chou C, Jaleel A, Qureshi M K. BEAR: techniques for mitigating bandwidth bloat in gigascale DRAM caches. ACM SIGARCH Comput Arch News, 2016, 43: 198--210. Google Scholar

[2] Hudec B, Hsu C W, Wang I T, et al. 3D resistive ram cell design for high-density storage class memory---a review. Sci China Inf Sci, 2016, 59: 061403. Google Scholar

[3] Lun Z Y, Du G, Zhao K, et al. A two-dimensional simulation method for investigating charge transport behavior in 3-D charge trapping memory. Sci China Inf Sci, 2016, 59: 122403. Google Scholar

[4] Lee Y, Kim J, Jang H, et al. A fully associative, tagless DRAM cache. In: Proceedings of ACM/IEEE International Symposium on Computer Architecture, Portland, 2015. 211--222. Google Scholar

[5] Hameed F, Bauer L, Henkel J. Adaptive cache management for a combined SRAM and DRAM cache hierarchy for multi-cores. In: Proceedings of Design, Automation and Test in Europe, Grenoble, 2013. 77--82. Google Scholar

[6] Hundal R, Oklobdzija V G. Determination of optimal sizes for a first and second level SRAM-DRAM on-chip cache combination. In: Proceedings of IEEE International Conference on Computer Design: VLSI in Computers and Processors, Cambridge, 1994. 60--64. Google Scholar

[7] Hameed F, Bauer L, Henkel J. Reducing latency in an SRAM/DRAM cache hierarchy via a novel tag-cache architecture. In: Proceedings of Design Automation Conference, San Francisco, 2014. 1--6. Google Scholar

[8] Huang C C, Nagarajan V. ATCache: reducing DRAM cache latency via a small SRAM tag cache. In: Proceedings of International Conference on Parallel Architectures and Compilation, Edmonton, 2014. 51--60. Google Scholar

[9] Qureshi M K, Loh G H. Fundamental latency trade-off in architecting DRAM caches: outperforming impractical SRAM-tags with a simple and practical design. In: Proceedings of IEEE/ACM International Symposium on Microarchitecture, Vancouver, 2012. 235--246. Google Scholar

[10] Elhelw A S, El-Moursy A, Fahmy H A H. Adaptive time-based least memory intensive scheduling. In: Proceedings of IEEE 9th International Symposium on Embedded Multicore/Manycore Systems-on-Chip, Turin, 2015. 167--174. Google Scholar

[11] Elhelw A S, Moursy A E, Fahmy H A H. Time-based least memory intensive scheduling. In: Proceedings of IEEE 8th International Symposium on Embedded Multicore/Manycore Systems-on-Chip, Aizu-Wakamatsu, 2014. 311--318. Google Scholar

[12] Chen Q, Zheng L, Guo M. DWS: demand-aware work-stealing in multi-programmed multi-core architectures. In: Proceedings of International Workshop on Programming Models and Applications on Multicores and Manycores, Orlando, 2014. 131. Google Scholar

[13] Chen Q, Zheng L, Guo M. Adaptive demand-aware work-stealing in multi-programmed multi-core architectures. J Concurr Comput Prac Exp, 2016, 28: 455--471. Google Scholar

[14] Roscoe B, Herlev M, Liu C. Auto-tuning multi-programmed workload on the SCC. In: Proceedings of International Green Computing Conference, Arlington VA, 2013. 1--5. Google Scholar

[15] Castellana V G, Ferrandi F. Abstract: speeding-up memory intensive applications through adaptive hardware accelerators. In: Proceedings of SC Companion: High Performance Computing, Networking Storage and Analysis, Salt Lake City, 2012. 1415--1416. Google Scholar

[16] Huang C, Ravi S, Raghunathan A, et al. Synthesis of heterogeneous distributed architectures for memory-intensive applications. In: Proceedings of International Conference on Computer Aided Design, San Jose, 2003. 46--53. Google Scholar

[17] Huang C, Ravi S, Raghunathan A, et al. Generation of heterogeneous distributed architectures for memory-intensive applications through high-level synthesis. IEEE Trans Very Large Scale Int Syst, 2007, 15: 1191--1204. Google Scholar

[18] Athanasaki E, Anastopoulos N, Kourtis K, et al. Exploring the performance limits of simultaneous multithreading for memory intensive applications. J Supercomp, 2008, 44: 64--97. Google Scholar

[19] Chun K C, Jain P, Kim C H. Logic-compatible embedded DRAM design for memory intensive low power systems. In: Proceedings of IEEE International Symposium on Circuits and Systems, Paris, 2010. 277--280. Google Scholar

[20] Yi W, Tang Y, Wang G, et al. A case study of SWIM: optimization of memory intensive application on GPGPU. In: Proceedings of International Symposium on Parallel Architectures, Algorithms and Programming, Dalian, 2010. 123--129. Google Scholar

[21] Qin X, Jiang H, Zhu Y, et al. A Feedback control mechanism for balancing I/O-and memory-intensive applications on clusters. Scal Comput Prac Exp, 2005, 6: 95--107. Google Scholar

[22] Qin X, Jiang H, Zhu Y, et al. Dynamic load balancing for I/O- and memory-intensive workload in clusters using a feedback control mechanism. In: Proceedings of International Euro-Par Conference, Klagenfurt, 2003. 224--229. Google Scholar

[23] Jaleel A, Nuzman J, Moga A, et al. High performing cache hierarchies for server workloads: relaxing inclusion to capture the latency benefits of exclusive caches. In: Proceedings of IEEE International Symposium on High Performance Computer Architecture, Burlingame, 2015. 343--353. Google Scholar

[24] Xiao N, Zhao Y J, Liu F, et al. Dual queues cache replacement algorithm based on sequentiality detection. Sci China Inf Sci, 2012, 55: 191--199. Google Scholar

[25] Akin B, Franchetti F, Hoe J C. Data reorganization in memory using 3D-stacked DRAM. In: Proceedings of ACM/IEEE International Symposium on Computer Architecture, Portland, 2015. 131--143. Google Scholar

[26] Jevdjic D, Volos S, Falsafi B. Die-stacked DRAM caches for servers: hit ratio, latency, or bandwidth? have it all with footprint cache. In: Proceedings of ACM/IEEE International Symposium on Computer Architecture, Tel-Aviv, 2013. 404--415. Google Scholar

[27] Oskin M, Loh G H. A software-managed approach to die-stacked DRAM. In: Proceedings of International Conference on Parallel Architecture and Compilation, San Francisco, 2015. 188--200. Google Scholar

[28] Loh G H, Hill M D. Supporting very large DRAM caches with compound-access scheduling and MissMap. IEEE Micro, 2012, 32: 70--78. Google Scholar

[29] Loh G H, Hill M D. Efficiently enabling conventional block sizes for very large die-stacked DRAM caches. In: Proceedings of IEEE/ACM International Symposium on Microarchitecture, Porto Alegre, 2011. 454--464. Google Scholar

[30] Dong H W, Seong N H, Lee H H S. Pragmatic integration of an SRAM row cache in heterogeneous 3-D DRAM architecture Using TSV. IEEE Trans Very Large Scale Int Syst, 2013, 21: 1--13. Google Scholar

[31] Chen X, Xu C, Dick R P, et al. Performance and power modeling in a multi-programmed multi-core environment. In: Proceedings of Design Automation Conference, Anaheim, 2010. 813--818. Google Scholar

[32] Suo G, Yang X. System level speedup oriented cache partitioning for multi-programmed systems. In: Proceedings of IFIP International Conference on Network and Parallel Computing, Gold Coast, 2009. 204--210. Google Scholar

[33] Kirovski D, Lee C, Potkonjak M, et al. Application-driven synthesis of memory-intensive systems-on-chip. IEEE Trans Comp-Aided Des Int Circ Syst, 1999, 18: 1316--1326. Google Scholar

[34] Sim J, Loh G H, Sridharan V, et al. A configurable and strong RAS solution for die-stacked DRAM caches. IEEE Micro, 2014, 34: 80--90. Google Scholar

[35] Chou C, Jaleel A, Qureshi M K. CAMEO: a two-level memory organization with capacity of main memory and flexibility of hardware-managed cache. In: Proceedings of IEEE/ACM International Symposium on Microarchitecture, Cambridge, 2014. 1--12. Google Scholar

[36] Ou J, Patton M, Moore M D, et al. A penalty aware memory allocation scheme for key-value cach. In: Proceedings of International Conference on Parallel Processing, Beijing, 2015. 530--539. Google Scholar

[37] Hennessy J L, Patterson D A. Computer Architecture: a Quantitative Approach. 5th ed. Waltham: Morgan Kaufmann, 2012. 72--96. Google Scholar

[38] Sim J, Loh G H, Kim H, et al. A mostly-clean DRAM cache for effective hit speculation and self-balancing dispatch. In: Proceedings of IEEE/ACM International Symposium on Microarchitecture, Vancouver, 2012. 247--257. Google Scholar

[39] Begum R, Hempstead M. Power-agility metrics: measuring dynamic characteristics of energy proportionality. In: Proceedings of IEEE International Conference on Computer Design, New York, 2015. 643--650. Google Scholar

Copyright 2019 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1