Livermore Seminar

67 %
33 %
Information about Livermore Seminar

Published on November 21, 2007

Author: Garrick


Scientific Application Performance on Candidate PetaScale Applications:  Scientific Application Performance on Candidate PetaScale Applications Leonid Oliker Future Technologies Group Lawrence Berkeley National Laboratory Winner Best Paper, IPDPS 07, Long Beach, CA, March 26-30, 2007 Overview:  Overview Stagnating application performance is well-know problem in scientific computing By end of decade numerous mission critical applications expected to have 100X computational demands of current levels Many HEC platforms are poorly balanced for demands of leading applications Memory-CPU gap, deep memory hierarchies, poor network-processor integration, low-degree network topology Traditional superscalar trends slowing down Mined most benefits of ILP and pipelining, Clock frequency limited by power concerns Traditional Sources of Performance Improvement are Flat-Lining:  Traditional Sources of Performance Improvement are Flat-Lining New Constraints 15 years of exponential clock rate growth has ended But Moore’s Law continues! How do we use all of those transistors to keep performance increasing at historical rates? Industry Response: #cores per chip doubles every 18 months instead of clock frequency! Figure courtesy of Kunle Olukotun, Lance Hammond, Herb Sutter, and Burton Smith Hardware: What are the problems?:  Current Hardware/Lithography Constraints Power limits leading edge chip designs Intel Tejas Pentium 4 cancelled due to power issues Yield on leading edge processes dropping dramatically IBM quotes yields of 10 – 20% on 8-processor Cell Design/validation leading edge chip is becoming unmanageable Verification teams > design teams on leading edge processors Solution: Small Is Beautiful Expect modestly pipelined (5- to 9-stage) CPUs, FPUs, vector, SIMD PEs Small cores not much slower than large cores Parallel is energy efficient path to performance: Lower threshold and supply voltages lowers energy per op Redundant processors can improve chip yield Cisco Metro 188 CPUs + 4 spares; Sun Niagara sells 6 or 8 CPUs Small, regular processing elements easier to verify Hardware: What are the problems? How Small is “Small”:  How Small is “Small” Power5 (Server) 389mm^2 120W@1900MHz Intel Core2 sc (laptop) 130mm^2 15W@1000MHz ARM Cortex A8 (automobiles) 5mm^2 0.8W@800MHz Tensilica DP (cell phones / printers) 0.8mm^2 0.09W@600MHz Tensilica Xtensa (Cisco router) 0.32mm^2 for 3! 0.05W@600MHz Each core operates at 1/3 to 1/10th efficiency of largest chip, but you can pack 100x more cores onto a chip and consume 1/20 the power The Future of HPC System Concurrency:  The Future of HPC System Concurrency Must ride exponential wave of increasing concurrency for foreseeable future Manycore Concurrency Levels:  Manycore Concurrency Levels Application Evaluation:  Application Evaluation Microbenchmarks, algorithmic kernels, performance modeling and prediction, are important components of understanding and improving architectural efficiency However full-scale application performance is the final arbiter of system utility and necessary as baseline to support all complementary approaches Our evaluation work emphasizes full applications, with real input data, at the appropriate scale Requires coordination of computer scientists and application experts from highly diverse backgrounds Our initial efforts have focused on comparing performance between high-end vector and scalar platforms Currently evaluating ultra-scale systems (soon to be petascale) NERSC 0.1PF/s (XT4) in FY07, 1PF/s in FY10 ORNL 1PF/s in FY08/09 (BAKER) SNL 1PF/s in FY 08 (XT4) ANL 0.5PF/s in FY08/09 (BG/P) LANL 1PF/s in FY08 (RoadRunner) Hypothetical 1PT/s sustained (2011) will contain >1M processors! Important understand algorithmic aspects allow/prevent petascale computation Benefits of Evaluation:  Benefits of Evaluation Full scale application evaluation lead to more efficient use of the community resources For both current installation and future designs Head-to-head comparisons on full applications: Help identify the suitability of a particular architecture for a given application class Give application scientists information about how well various numerical methods perform across systems Reveal performance-limiting system bottlenecks that can aid designers of the next generation systems. Science Driven Architecture In-depth studies reveal limitation of compilers, operating systems, and hardware, since all of these components must work together at scale to achieve high performance. Application Overview:  Application Overview Examining set of applications with a variety of numerical methods and communication patterns with the potential to run at petascale Architectural Comparison:  Architectural Comparison These architectures represent main stream petaflop candidate systems Power5 aggregates 4 DDR233 for high 6.8 GB/s Stream bandwidth Jaguar XT3 dual-core w/ SeaStar routing (Catamount) Jacquard single-core Opteron with non-custom communication integration BG/L, power eff, 5 networks, dual-core w/o L1 cohere, 512MB/node, 2 SIMD FPU Phoenix, X1E custom HPC platform, X1 upgrade (double MSP w/o BW increase) Yes Cost is a critical metric - however we are unable to provide such data Proprietary, pricing varies based on customer and time frame Poorly balanced systems cannot solve important problems/resolutions regardless of cost! Astrophysics: CACTUS:  Astrophysics: CACTUS Numerical solution of Einstein’s equations from theory of general relativity Among most complex in physics: set of coupled nonlinear hyperbolic & elliptic systems with thousands of terms CACTUS evolves these equations to simulate high gravitational fluxes, such as collision of two black holes Evolves PDE’s on regular grid using finite differences Visualization of grazing collision of two black holes Developed at Max Planck Institute, vectorized/ported by John Shalf & Tom Goodale Cactus Performance: Weak Scaling:  Cactus Performance: Weak Scaling Bassi shows highest raw and sustained performance Remains to be seen if scalability will continue to thousands of processors Jacquard/Jaguar show more modest scaling - Phoenix X1 shows lowest % of peak (Cactus crashes on X1E) Small fraction unvectorizable code (boundary condition) leads to large penalty Uses only 1 of 4 SSP within an MSP for an unvectorizable code portion BG/L lowest raw performance - expected simple (power efficient) dual-issue in-order PPC440 Achieves near perfect scalability to highest concurrency ever attained Virtual node mode (32K procs) showed no performance degradation (smaller problem) Topology mapping attempts did not improve performance Overall shows potential of Cactus to run at Petascale Fluid Dynamics: ELB3D:  Fluid Dynamics: ELB3D LBM concept: develop simplified kinetic model with incorporated physics, to reproduce correct macroscopic averaged properties Unlike explicit LBM methods, entropic approach not prone to nonlinear instabilities ELBM develop to simulate Navier-Stokes turbulence Non-linear equation is solved at each grid point and time step to satisfy constraints, followed by streaming of data The equation is solved via a Newton-Raphson iteration (heavy use of log function) Spatial grid overlayed with phase-space velocity lattice Block distributed over processor grid Developed by George Vahala’s group College of William & Mary, ported Jonathan Carter Evolution of vorticity into turbulent structures ELB3D Performance: Strong Scaling:  ELB3D Performance: Strong Scaling Specialized log functions (ASSV, ACML) used on all platforms w/ significant improvement Shows high % of peak and good scaling across all platforms - good load balance Results show that high comp cost of entropic algorithm can be designed in efficient manner X1E attains the highest raw performance: Innermost gridpoint loop taken inside nonlinear equation to allow for vectorization Bassi shows highest fraction of peak: high memory BW, large caches, advanced prefetch BG/L shows lowest efficiency, but value is based on peak with double hummer Except for highly-tuned libraries, peak will almost always be 50% of potential Overall results show promising performance of ELB methods at PetaScale Slide16:  LBMHD Cell Performance Multi-core scientific kernel optimization work by Samuel Williams (UCB/LBNL) Cell achieves impressive performance for MHD problem (explicit LB method) Ability to explicitly control local memory (more programming complexity) Results shown only for collision phase (>>85% of overall time) Multi-blade (MPI) Cell results to come Magnetic Fusion: GTC:  Magnetic Fusion: GTC Gyrokinetic Toroidal Code: transport of thermal energy (plasma microturbulence) Goal magnetic fusion is burning plasma power plant producing cleaner energy GTC solves 3D gyroaveraged gyrokinetic system w/ particle-in-cell approach (PIC) PIC scales N instead of N2 – particles interact w/ electromagnetic field on grid Allows solving equation of particle motion with ODEs (instead of nonlinear PDEs) Vectorization inhibited since multiple particles may attempt to concurrently update same grid point Whole volume and cross section of electrostatic potential field, showing elongated turbulence eddies Developed at PPPL, vectorized/optimized by Stephane Ethier GTC Performance: Weak Scaling:  GTC Performance: Weak Scaling Generally expect low % of peak due to scatter gather Extensive X1E optimization (such as reversing array dimensions) paid off Although dropping at higher concurrency Three commodity superscalers (Power5, Opteron) achieve similar raw performance However Bassi, only attains 1/2 %peak of Opteron High Opteron performance partially due to low main memory latency Bassi, Jaguar, and BG/L show excellent scalability BG/L scales to 32K processors: virtual node mode only 5% slower than coprocessor Optimizations: vector MASSV functions (60%) and mapping onto 3D Torus (30%) Results show GTC prime candidate for Petascale systems Perfect load balancing up to 32K processors and low communication High Energy Physics: BeamBeam3D:  High Energy Physics: BeamBeam3D BB3D models collisions of counter-rotating charge particle beams Used to study beam-beam collisions at world’s highest energy accelerators Particle-in-cell method, where particles are deposited on 3D grid to calculate charge density distribution At collision points electric/magnetic fields calculated using Vlasov-Poisson via FFT Particle are advances using computed fields and accelerator forces High communication requirements: Global gather charge density Broadcast electric/magnetic fields Global FFT transpose Developed by Ji Qiang, ported/optimized by Hongzhang Shan GTC and BB3D both PIC w/ charged particle interactions, but: GTC requires relatively small particle movements Allows local solve of Poisson equation BB3D has longer range particle forces Requires high communication, global FFTs Limited number of subdomains available (2048 our case) Pure domain decomposition not possible BeamBeam3D Performance: Strong Scaling:  BeamBeam3D Performance: Strong Scaling Phoenix attains fastest per-processor performance, however scalability is poor at high P At P=256 > 50% communication, at high P vector length decreases (fixed size problem) At P=512 Bassi surpasses Phoenix - outperforms Opterons by 1.8x, BG/L by 4.5X Jacquard and Jaguar show similar behavior, even though vastly different interconnects Notice all platforms achieve < 5% of peak for high concurrencies! Indirect addressing, global all-to-all comm, extensive (non-flop) data movement BG/L achieves just 1% of peak at high concurrencies Higher scalability experiments not possible due to limited number of domains To reach Petascale for BB3D requires extensive code reengineering: Additional decomposition schemes, reduce communication bottleneck Materials Science: PARATEC:  Materials Science: PARATEC PARATEC performs first-principles quantum mechanical total energy calculation using pseudopotentials & plane wave basis set Density Functional Theory to calc structure & electronic properties of new materials using QM principles DFT calc are one of the largest consumers of supercomputer cycles in the world 33% 3D FFT, 33% BLAS3, 33% Hand coded F90 Part of calculation in real space other in Fourier space Uses specialized 3D FFT to transform wavefunction QDOTs have important photo-luminscent properties Conduction band minimum electron state for CdSe quantum dot Developed by Andrew Canning with Louie and Cohen’s groups (UCB, LBNL) PARATEC Performance: Strong Scaling:  PARATEC Performance: Strong Scaling PARATEC attains high % peak due to 1D FFTs and BLAS3 Bassi achieves highest performance and good scaling up to 512 processors 1024 Power5 performance from LLNL Purple system Jaguar outperforms Jacquard due partially to high interconnect bandwidth BG/L lowest overall raw performance - but uses smaller system due to mem constraints Performance drops between 512 and 1024, due to moving off topology half-plane Phoenix shows lowest % peak Low scalar/vector ratio, vector length drops for custom F90 routines and BLAS3/FFT PARATEC scaling decreases and limited to ~2K processors , due to 1 level decomposition For petascale must introduce second level of decomposition, over electronic band indices Will increase scaling and reduce memory requirements for systems like BG/L Purple data courtesy of Tom Spelce and Bronis de Supinski Gas Dynamics: HyperCLaw:  Gas Dynamics: HyperCLaw Adaptive Mesh Refinement (AMR): powerful technique to solve otherwise intractable problems - typically applied to physical systems solving PDEs AMR dynamically refines underlying grid in regions of scientific interest Naively increasing grid resolution prohibitive in computation and memory Price paid for additional power is complexity: significant software infrastructure Regridding, interpolation between coarse/fine grids, dynamic load balancing Generally written in flexible/modular fashion Causing performance reduction even for unrefined grid calculations Complex communication looks more like many-to-many than stencil exchange Berger-Colella Hyperbolic Conservation laws C++/Fortran hybrid programming model Godunov - Advances the solution at a given refinement level TimeStep - Prepares grids at next level of Godunov solver Regrid - Replace existing grid hierarchy to maintain numerical accuracy Knapsack - Load balancing algorithm Developed by CCSE LBNL, ported/optimized by Michael Lijewski and Michael Welcome X1E Optimization:  X1E Optimization Two X1E optimizations undertaken since our original study in CF06 Knapsack and Regridding originally prevents X1E scalability Knapsack optimized by swapping pointers instead of lists Results in almost cost free X1E Knapsack algorithm Regridding originally required O(N^2) box list intersection calculation Updated hashing scheme reduced overhead to O(NlogN) Significantly reduced X1E regridding overhead HyperCLaw Performance: Weak Scaling:  HyperCLaw Performance: Weak Scaling Due to code complexity, all platforms achieve < 5% of peak Increasing adaptivity level introduces grid management overheads, reducing performance Bassi shows highest raw performance, BG/L the lowest Jacquard and Jaguar shoe similar raw performance X1E shows lowest fraction of peak ~1%, before optimization this was much lower Hierarchical data structures management is scalar work Higher concurrency reduces vector length AMR designed for small grid patches, causes short vector length Adding more complexity (elliptic, chemistry) would certainly reduce vector performance May be possible to design AMR parallel vector code, but only from the ground up Despite low efficiency, systems show good scalability - potential for petascale systems Performance Summary:  Performance Summary Evaluating HPC on realistic problems is complex task, requires diverse group domain scientists Work presents one of most extensive performance analyses to date On average the Power5 (Bassi) achieves highest raw and sustained performance High mem BW, tight interconnect integration, latency hiding via advanced prefetching Phoenix X1E achieved high perf for GTC and ELB3D, but showed poor results for several codes Opteron systems show similar performance However, Jaguar’s XT3 interconnect outperformed Jacquard for GTC & PARATEC BG/L - lowest performance, but achieved unprecedented concurrency on several codes Potential to run at petascale: GTC, Cactus - very encouraging, linear scaling to 32K processors ELB3D, Hyperclaw - promising scaling at lower concurrencies BB3D and Paratec - require reengineering to incorporate additional levels of parallelism System Balance :  System Balance Halving the memory bandwidth has virtually no effect on performance for a broad variety of applications Memory Bandwidth (maintaining system balance):  Memory Bandwidth (maintaining system balance) Neither memory bandwidth nor FLOPs dominate runtime The “other” category dominated by memory latency stalls Points to inadequacies in current CPU core design (inability to tolerate latency) Future and Related Work:  Future and Related Work Future Evaluation Work: Explore more complex methods and irregular data structures Continue evaluating leading HPC systems as they come online Perform in-depth application characterizations Related research activities: Developing an application-derived I/O benchmark based on CMB Investigating automatic tuning for stencil computations Investigating scientific kernel performance on multi-core systems Investigating potential of dynamically reconfigurable interconnects to achieve fat-tree performance at fraction of switch components Papers available at CMB Science:  CMB Science The CMB is a unique probe of the very early Universe Tiny fluctuations in its temperature (1 in 100K) and polarization (1 in 100M) encode the fundamental parameters of cosmology, including the geometry, composition (mass-energy content), and ionization history of the Universe Combined with complementary supernova measurements tracing the dynamical history of the Universe, we have an entirely new “concordance” cosmology: 70% dark energy + 25% dark matter + 5% ordinary matter Dark Matter + Ordinary Matter Dark Energy MADbench Synchronous performance:  MADbench Synchronous performance MADbench Asynchronous performance:  MADbench Asynchronous performance

#cores presentations

Add a comment

Related presentations

Related pages

Livermore Seminar Events | Eventbrite

Looking for the best seminars that Livermore has to offer? Find your next great experience on Eventbrite.
Read more

Seminars - Lawrence Livermore National Laboratory

Lawrence Livermore National Laboratory 7000 East Avenue • Livermore, CA 94550 Operated by Lawrence Livermore National ...
Read more

seminar series | Lawrence Livermore National Laboratory

Lawrence Livermore National Laboratory 7000 East Avenue • Livermore, CA 94550 Operated by Lawrence Livermore National Security, LLC, for the
Read more

Aikido of Livermore Shinrei Dojo

Aikido of Livermore Shinrei Dojo, Livermore, CA. 753 likes · 472 were here. Aikido of Livermore/Shinrei Dojo
Read more

Lab participating in small business seminar | Lawrence ...

Lawrence Livermore National Laboratory 7000 East Avenue • Livermore, CA 94550 Operated by Lawrence Livermore National Security, LLC, for the
Read more

Livermore Events | Eventbrite

Looking for the best that Livermore has to offer? Find your next great experience on Eventbrite.
Read more

Livermore Valley Winegrowers Association | Regional Events

Inner Artiste Sip-N-Paint. April 20, 2016 (6:30 - 9:30pm) Concannon Vineyard 4590 Tesla Road, Livermore, CA 94550 - directions
Read more

Balance-Based Torso-Weighting Augmenting Sensory ...

Livermore, CA 94551 Seminar Price: $150 Register to Attend a BBTW™ Seminar Saturday, February 19, 2011 8 a.m. Name: _____ Clinical Focus ...
Read more

Livermore Super Seminar - part 3

Livermore Super Seminar - part 3 Details Alex 29 January 2008 Print ...
Read more

Livermore Computing Seminar Announcement

Title: A Holistic Approach towards Automated Performance Analysis and Tuning: Date: May 14, 2009 10:00-11:00am: Location: B451 White Room (R1025) Speaker:
Read more