Analysis of a Chip Multiprocessor Using Scientific Applications

Slides:



Advertisements
Similar presentations
Electrical and Computer Engineering UAH System Level Optical Interconnect Optical Fiber Computer Interconnect: The Simultaneous Multiprocessor Exchange.
Advertisements

QuT: A Low-Power Optical Network-on-chip
A Novel 3D Layer-Multiplexed On-Chip Network
0 Arun Rodrigues, Scott Hemmert, Dave Resnick: Sandia National Lab (ABQ) Keren Bergman: Columbia University Bruce Jacob: U. Maryland John Shalf, Paul Hargrove:
CCNoC: On-Chip Interconnects for Cache-Coherent Manycore Server Chips CiprianSeiculescu Stavros Volos Naser Khosro Pour Babak Falsafi Giovanni De Micheli.
L2 to Off-Chip Memory Interconnects for CMPs Presented by Allen Lee CS258 Spring 2008 May 14, 2008.
Firefly: Illuminating Future Network-on-Chip with Nanophotonics Yan Pan, Prabhat Kumar, John Kim †, Gokhan Memik, Yu Zhang, Alok Choudhary EECS Department.
IP I/O Memory Hard Disk Single Core IP I/O Memory Hard Disk IP Bus Multi-Core IP R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R Networks.
Hybrid Electric/Photonic Networks for Scientific Applications on Tiled CMPs Ankit Jain, Shoaib Kamil, Marghoob Mohiyuddin CS258 Final Presentation Prof.
Dispersion Measurements Lecture-3. Dispersion Measurements Measurement of Intermodal Dispersion The most common method for measuring multimode fiber bandwidth.
Performance and Power Efficient On-Chip Communication Using Adaptive Virtual Point-to-Point Connections M. Modarressi, H. Sarbazi-Azad, and A. Tavakkol.
COLUMBIA UNIVERSITY Interconnects Jim Tomkins: “Exascale System Interconnect Requirements” Jeff Vetter: “IAA Interconnect Workshop Recap and HPC Application.
RFAD LAB, YONSEI University IEEE TRANSACTIONS ON COMPUTERS, VOL. 57, NO. 9, SEPTEMBER 2008 Photonic Networks-on-Chip for Future Generations of Chip Multiprocessors.
Photonic Networks on Chip Yiğit Kültür CMPE 511 – Computer Architecture Term Paper Presentation 27/11/2008.
ROBERT HENDRY, GILBERT HENDRY, KEREN BERGMAN LIGHTWAVE RESEARCH LAB COLUMBIA UNIVERSITY HPEC 2011 TDM Photonic Network using Deposited Materials.
On-Chip Networks and Testing
Report Advisor: Dr. Vishwani D. Agrawal Report Committee: Dr. Shiwen Mao and Dr. Jitendra Tugnait Survey of Wireless Network-on-Chip Systems Master’s Project.
International Symposium on Low Power Electronics and Design NoC Frequency Scaling with Flexible- Pipeline Routers Pingqiang Zhou, Jieming Yin, Antonia.
Waveguide High-Speed Circuits and Systems Laboratory B.M.Yu High-Speed Circuits and Systems Laboratory 1.
Déjà Vu Switching for Multiplane NoCs NOCS’12 University of Pittsburgh Ahmed Abousamra Rami MelhemAlex Jones.
Design and Evaluation of Hierarchical Rings with Deflection Routing Rachata Ausavarungnirun, Chris Fallin, Xiangyao Yu, ​ Kevin Chang, Greg Nazario, Reetuparna.
A 1.25-Gb/s Digitally-Controlled Dual-Loop Clock and Data Recovery Circuit with Improved Effective Phase Resolution Chang-Kyung Seong 1), Seung-Woo Lee.
Performance and Energy Comparison of Electrical and Hybrid Photonic Networks for CMPs Ankit Jain, Shoaib Kamil, Marghoob Mohiyuddin, John Shalf, John Kubiatowicz.
Silicon Nanophotonic Network-On-Chip Using TDM Arbitration
Rev PA1 1 Performance energy trade-offs with Silicon Photonics Sébastien Rumley, Robert Hendry, Dessislava Nikolova, Keren Bergman.
Enabling System-Level Modeling of Variation-Induced Faults in Networks-on-Chips Konstantinos Aisopos (Princeton, MIT) Chia-Hsin Owen Chen (MIT) Li-Shiuan.
Performance and Energy Comparison of Electrical and Hybrid Photonic Networks for CMPs Ankit Jain, Shoaib Kamil, Marghoob Mohiyuddin, John Shalf, John Kubiatowicz.
Yu Cai Ken Mai Onur Mutlu
HPEC 2007, Lexington, MA18-20 September, 2007 On-Chip Photonic Communications for High Performance Multi-Core Processors Keren Bergman, Luca Carloni, Columbia.
Assaf Shacham, Keren Bergman, Luca P. Carloni Presented for HPCAN Session by: Millad Ghane NOCS’07.
1 November 11, 2015 A Massively Parallel, Hybrid Dataflow/von Neumann Architecture Yoav Etsion November 11, 2015.
Hybrid Optoelectric On-chip Interconnect Networks Yong-jin Kwon 1.
Virtual-Channel Flow Control William J. Dally
Simultaneous Multi-Layer Access Improving 3D-Stacked Memory Bandwidth at Low Cost Donghyuk Lee, Saugata Ghose, Gennady Pekhimenko, Samira Khan, Onur Mutlu.
Optical Computing. Objectives  Definition of Optical computing  Advantages of Optical computing  Optical components for binary digital computer  Misconceptions,
Network On Chip Cache Coherency Final presentation – Part A Students: Zemer Tzach Kalifon Ethan Kalifon Ethan Instructor: Walter Isaschar Instructor: Walter.
Building manycore processor-to-DRAM networks using monolithic silicon photonics Ajay Joshi †, Christopher Batten †, Vladimir Stojanović †, Krste Asanović.
HAT: Heterogeneous Adaptive Throttling for On-Chip Networks Kevin Kai-Wei Chang Rachata Ausavarungnirun Chris Fallin Onur Mutlu.
UNIVERSITY OF WATERLOO Nortel Networks Institute University of Waterloo.
M AESTRO : Orchestrating Predictive Resource Management in Future Multicore Systems Sangyeun Cho, Socrates Demetriades Computer Science Department University.
Runtime Reconfigurable Network-on- chips for FPGA-based systems Mugdha Puranik Department of Electrical and Computer Engineering
Network-on-Chip Paradigm Erman Doğan. OUTLINE SoC Communication Basics  Bus Architecture  Pros, Cons and Alternatives NoC  Why NoC?  Components 
Process Variation Aware Crosstalk Mitigation for DWDM based Photonic NoC Architectures. By Ishan Thakkar
Ring Resonator Gyroscope
FlexiBuffer: Reducing Leakage Power in On-Chip Network Routers
Discussion today Over the next two class periods we will be designing a ring resonator based 2-channel wavelength division multiplexing (WDM) optical link.
Lecture 23: Interconnection Networks
System On Chip.
ESE532: System-on-a-Chip Architecture
3Boston University ECE Dept.;
Optical Amplifier.
Pablo Abad, Pablo Prieto, Valentin Puente, Jose-Angel Gregorio
Exploring Concentration and Channel Slicing in On-chip Network Router
Azeddien M. Sllame, Amani Hasan Abdelkader
Router and Routing Designs for 3D Optical
OpenSMART: Single-cycle Multi-hop NoC Generator in BSV and Chisel
Gilbert Hendry Johnnie Chan, Daniel Brunina,
William Stallings Data and Computer Communications
The Role of Light in High Speed Digital Design
Gwangsun Kim Niladrish Chatterjee Arm, Inc. NVIDIA Mike O’Connor
Rahul Boyapati. , Jiayi Huang
Discussion today Over the next two class periods we will be designing a ring resonator based 2-channel wavelength division multiplexing (WDM) optical link.
Israel Cidon, Ran Ginosar and Avinoam Kolodny
Subject Name: Digital Switching Systems Subject Code:10EC82 Prepared By: Aparna.P, Farha Kowser Department: Electronics and Communication Date:
Leveraging Optical Technology in Future Bus-based Chip Multiprocessors
Integrated Optical Wavelength Converters and Routers for Robust Wavelength-Agile Analog/ Digital Optical Networks Daniel J. Blumenthal (PI), John E. Bowers,
Optical communications & networking - an Overview
Fiber Optic Transmission
Exploring Chip to Chip Photonic Networks
Multiprocessors and Multi-computers
Presentation transcript:

Analysis of a Chip Multiprocessor Using Scientific Applications Gilbert Hendry Shoaib Kamil Marghoob Mohiyuddin Ankit Jain Leonid Oliker John Kubiatowicz John Shalf Aleksandr Biberman Johnnie Chan Benjamin G. Lee Luca P. Carloni Keren Bergman

Motivation CMPs of the future = 3D stacking Lots of data on chip Photonics offers key advantages Network layer Memory layers Multi-core processor layer International Symposium on Networks-on-Chip 11/11/2018

Why Photonics? OPTICS: ELECTRONICS: Photonics changes the rules for Bandwidth, Energy, and Distance. OPTICS: Modulate/receive high bandwidth data stream once per communication event. Broadband switch routes entire multi-wavelength stream. Off-chip BW = On-chip BW for nearly same power. ELECTRONICS: Buffer, receive and re-transmit at every router. Each bus lane routed independently. (P  NLANES) Off-chip BW is pin-limited and power hungry. TX RX RX RX RX RX RX TX RX TX TX TX TX TX TX TX TX International Symposium on Networks-on-Chip 11/11/2018

Silicon Photonic Integration Cornell, 2005 MIT, 2008 IBM, 2007 Sandia, 2008 Ghent, 2007 Columbia, 2008

Related Work Shacham, NOCS ‘07 Vantrease, ISCA ‘08 Batten, HOTI ‘08 International Symposium on Networks-on-Chip 11/11/2018

Hybrid Photonic Network Photonic Transmission Electronic Control Compute International Symposium on Networks-on-Chip 11/11/2018

Hybrid Photonic Network International Symposium on Networks-on-Chip 11/11/2018

Hybrid Photonic Network International Symposium on Networks-on-Chip 11/11/2018

Contributions This work achieves: Accurate simulation Application-based workloads Comparison of electronic and photonic networks International Symposium on Networks-on-Chip 11/11/2018

NanoPhotonic Devices Silicon waveguides Ring resonator (filter) Laser Ring resonator (modulator) Photodetectors Electronic data L. Chen, OE, 2008 International Symposium on Networks-on-Chip 11/11/2018

Switching Building Blocks Broadband 2×2 Switch B. G. Lee, ECOC 2008 Cross State Bar State  Transmission International Symposium on Networks-on-Chip 11/11/2018

Switch Characterization Broadband 1×2 Switch ER IL [A. Biberman et al., LEOS, 2007] Loss Parameter Value Waveguide propagation 0.5 dB/cm Waveguide crossing 0.05 dB Waveguide bend 0.005 dB/90o Passing by Micro-Ring (OFF) 0 dB Coupling into Micro-Ring (ON) 0.5 dB International Symposium on Networks-on-Chip 11/11/2018

Higher Order Switches N E S W International Symposium on Networks-on-Chip 11/11/2018

Simulation Environment Built in OMNeT++ Processing Plane Random, Trace, Execution Electronic Plane Routers XY routing Bubble Flow Control 4 VCs Pipelined – input, arbitration, output ORION – energy Circuit path setup logic Wires Custom lengths Photonic Plane Switches, modulators, detectors, filters, waveguides International Symposium on Networks-on-Chip 11/11/2018

Optical Loss Analysis Modulators Nonlinear effects Laser Switch Switch Total Injected power . PI = pi × Nλ Switch Injected power per wavelength Worst-case Insertion Loss Switch Received power Detectors Detector sensitivity Optical power International Symposium on Networks-on-Chip 11/11/2018

Insertion Loss Analysis International Symposium on Networks-on-Chip 11/11/2018

Experiment setup Networks Parameters Traffic Results International Symposium on Networks-on-Chip 11/11/2018

Experiment setup Networks Mesh Parameters Traffic Results Electronic Conc. Mesh Conc. Torus Electronic Mesh Concentrated Electronic Mesh Concentrated Electronic Torus International Symposium on Networks-on-Chip 11/11/2018

Experiment setup Networks Mesh Parameters Traffic Results Photonic Torus Networks Mesh Conc. Mesh Conc. Torus --------------- Ph. Torus Parameters Traffic Results International Symposium on Networks-on-Chip 11/11/2018

Selective Photonic Torus Experiment setup Parameters Traffic Results Networks Mesh Conc. Mesh Conc. Torus --------------- Ph. Torus Selective Selective Photonic Torus Ph. Torus El. Mesh Bandwidth 256B Msg size International Symposium on Networks-on-Chip 11/11/2018

Concentrated Photonic Torus Experiment setup Networks Mesh Conc. Mesh Conc. Torus --------------- Ph. Torus Selective Conc. Sel. Parameters Traffic Results Concentrated Photonic Torus Core Core Gateway Core Core International Symposium on Networks-on-Chip 11/11/2018

Simulation Parameters Networks Traffic Results Parameter Value Cores 64 Clock Frequency 5 GHz Data rate 10 Gb/s Parameters Network Channel Width Buffer Size (b) Electronic Mesh 128 1024 Conc. Electronic Mesh 2048 Conc. Electronic Torus Photonic Torus 32 512 Selective Photonic Torus 64 Conc. Photonic Torus Selective Conc. Ph. Tor. Energy Parameter Value PSE dynamic energy 375 fJ PSE static (OFF) power 400 µW Modulation dynamic energy 25 fJ/bit Modulation static power 30 µW Detector Energy 50 fJ/b Wire Energy ~50 fJ/bit/mm International Symposium on Networks-on-Chip 11/11/2018

Synthetic Benchmarks Networks Parameters Results Traffic Each transfer occurs 100 times Two versions: small (96B), large (128kB) Networks Parameters Results Traffic Bitreverse Random Neighbor Tornado International Symposium on Networks-on-Chip 11/11/2018

Scientific Applications Profiled by overloading communication functions in Linux Traces broken into phases to preserve order Random mapping Networks Parameters Results Traffic Gyrokinetic Toroidal Code (GTC) Cactus Application Num Phases Num Msgs Avg. Msg. Size (b) Cactus 2 285 25600 GTC 63 129796 MADbench 195 15414 5613 PARATEC 34 126059 43.3 PARAllel Total Energy Code (PARATEC) MADbench International Symposium on Networks-on-Chip 11/11/2018

Results – Synthetic (Small) Conc. Mesh Photonic Torus Conc. Torus Conc. Torus Selective Conc. Selective Electronic Photonic International Symposium on Networks-on-Chip 11/11/2018

Results – Synthetic (Large) Conc. Mesh Photonic Torus Conc. Torus Conc. Torus Selective Conc. Selective Electronic Photonic International Symposium on Networks-on-Chip 11/11/2018

Results - Applications Conc. Mesh Photonic Torus Conc. Torus Conc. Torus Selective Conc. Selective Electronic Photonic International Symposium on Networks-on-Chip 11/11/2018

Conclusions Detailed physically accurate simulations of future networks are informative. Photonics wins on energy consumption Significant difference in performance across different apps. Large messages/distances do well. Synergistic co-design of electronic and photonic planes may be beneficial International Symposium on Networks-on-Chip 11/11/2018