Memory coherence in the age of multicores

As we enter an era of exascale multicores, the question of efficiently supporting a shared memory model has become of paramount importance. On the one hand, programmers demand the convenience of coherent shared memory; on the other, growing core counts place higher demands on the memory subsystem and increasing on-chip distances mean that interconnect delays are becoming a significant part of memory access latencies. In this article, we first review the traditional techniques for providing a shared memory abstraction at the hardware level in multicore systems. We describe two new schemes that guarantee coherent shared memory without the complexity and overheads of a cache coherence protocol, namely execution migration and library cache coherence. We compare these approaches using an analytical model based on average memory latency, and give intuition for the strengths and weaknesses of each. Finally, we describe hybrid schemes that combine the strengths of different schemes.

[1]  Anoop Gupta,et al.  Reducing Memory and Traffic Requirements for Scalable Directory-Based Cache Coherence Schemes , 1990, ICPP.

[2]  Jaehyuk Huh,et al.  Exploiting ILP, TLP, and DLP with the polymorphous TRIPS architecture , 2003, ISCA '03.

[3]  Mainak Chaudhuri.  PageNUCA: Selected policies for page-grain locality management in large shared chip-multiprocessor caches , 2009, 2009 IEEE 15th International Symposium on High Performance Computer Architecture.

[4]  Anoop Gupta,et al.  The SPLASH-2 programs: characterization and methodological considerations , 1995, ISCA.

[5]  D. Banks,et al.  Assembly and Packaging , 2006 .

[6]  Henry Hoffmann,et al.  On-Chip Interconnection Architecture of the Tile Processor , 2007, IEEE Micro.

[7]  Doug Burger,et al.  An adaptive, non-uniform cache structure for wire-delay dominated on-chip caches , 2002, ASPLOS X.

[8]  Babak Falsafi,et al.  Selective, accurate, and timely self-invalidation using last-touch prediction , 2000, ISCA '00.

[9]  Srinivas Devadas,et al.  Brief announcement: distributed shared memory based on computation migration , 2011, SPAA '11.

[10]  David J. Lilja,et al.  So many states, so little time: verifying memory coherence in the Cray X1 , 2003, Proceedings International Parallel and Distributed Processing Symposium.

[11]  Sang Lyul Min,et al.  Design and Analysis of a Scalable Cache Coherence Scheme Based on Clocks and Timestamps , 1992, IEEE Trans. Parallel Distributed Syst..

[12]  Srinivas Devadas,et al.  Scalable, accurate multicore simulation in the 1000-core era , 2011, (IEEE ISPASS) IEEE INTERNATIONAL SYMPOSIUM ON PERFORMANCE ANALYSIS OF SYSTEMS AND SOFTWARE.

[13]  Rami G. Melhem,et al.  A timestamp-based selective invalidation scheme for multiprocessor cache coherence , 1996, Proceedings of the 1996 ICPP Workshop on Challenges for Parallel Processing.

[14]  Angela C. Sodan.  Message-passing and shared-data programming models - wish vs. reality , 2005, 19th International Symposium on High Performance Computing Systems and Applications (HPCS'05).

[15]  S. Borkar,et al.  An 80-Tile Sub-100-W TeraFLOPS Processor in 65-nm CMOS , 2008, IEEE Journal of Solid-State Circuits.

[16]  Srinivas Devadas,et al.  Deadlock-free fine-grained thread migration , 2011, Proceedings of the Fifth ACM/IEEE International Symposium.

[17]  Mark Horowitz,et al.  An evaluation of directory schemes for cache coherence , 1998, ISCA '98.

[18]  Babak Falsafi,et al.  Reactive NUCA: near-optimal block placement and replication in distributed caches , 2009, ISCA '09.

[19]  Niladrish Chatterjee,et al.  Micro-pages: increasing DRAM efficiency with locality-aware data placement , 2010, ASPLOS 2010.

[20]  David Wentzlaff,et al.  Processor: A 64-Core SoC with Mesh Interconnect , 2010 .

[21]  Stefan Rusu,et al.  A 45nm 8-core enterprise Xeon ® processor , 2009 .

[22]  Vivek Sarkar,et al.  Baring It All to Software: Raw Machines , 1997, Computer.

[23]  Coniferous softwood.  GENERAL TERMS , 2003 .

[24]  David A. Wood,et al.  Dynamic self-invalidation: reducing coherence overhead in shared-memory multiprocessors , 1995, Proceedings 22nd Annual International Symposium on Computer Architecture.

[25]  Pierre Michaud.  Exploiting the cache capacity of a single-chip multi-core processor with execution migration , 2004, 10th International Symposium on High Performance Computer Architecture (HPCA'04).

[26]  Sangyeun Cho,et al.  Managing Distributed, Shared L2 Caches through OS-Level Page Allocation , 2006, 2006 39th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO'06).

[27]  Omer Khan,et al.  System-level Optimizations for Memory Access in the Execution Migration Machine ( EM 2 ) , 2011 .

[28]  Michael D. Noakes,et al.  The J-machine multicomputer: an architectural evaluation , 1993, ISCA '93.

[29]  Krste Asanovic,et al.  Victim replication: maximizing capacity while hiding wire delay in tiled chip multiprocessors , 2005, 32nd International Symposium on Computer Architecture (ISCA'05).

[30]  Anoop Gupta,et al.  Operating system support for improving data locality on CC-NUMA compute servers , 1996, ASPLOS VII.

[31]  Rajeev Balasubramonian,et al.  Dynamic hardware-assisted software-controlled page placement to manage capacity allocation and sharing within large caches , 2009, 2009 IEEE 15th International Symposium on High Performance Computer Architecture.

[32]  Marcelo Cintra,et al.  An OS-based alternative to full hardware coherence on tiled CMPs , 2008, 2008 IEEE 14th International Symposium on High Performance Computer Architecture.

[33]  David W. Nellans,et al.  Micro-pages: increasing DRAM efficiency with locality-aware data placement , 2010, ASPLOS XV.

[34]  Omer Khan,et al.  EM2: A Scalable Shared-Memory Multicore Architecture , 2010 .

[35]  Mahmut T. Kandemir,et al.  A novel migration-based NUCA design for Chip Multiprocessors , 2008, 2008 SC - International Conference for High Performance Computing, Networking, Storage and Analysis.