High Performance Computing

Transcription

1 High Performance Computing ADVANCED SCIENTIFIC COMPUTING Dr. Ing. Morris Riedel Adjunct Associated Professor School of Engineering and Natural Sciences, University of Iceland Research Group Leader, Juelich Supercomputing Centre, Germany SHORT LECTURE 14 Molecular Systems & Libraries November 9 th, 2017 Room Ág-303

2 Review of Lecture 13 Systems Biology & Bioinformatics Protein Folding Function of a protein is closely related to its 3D shape BLAST & Role of Databases query target (1) Compares gene sequences to DBs (2) Calculates match significance Monte Carlo Methods Pointwise real data inputs Pointwise sampling from probability distribution Inputs model (w/o monte carlo) model (with monte carlo) Output Output [1] Wikipedia on Monte Carlo method [2] S. Mohanty et al, JSC Simlab Biology modified from [3] Sequence Searching 2/ 27

3 HPC-A[dvanced] Scientific Computing Second Part Consists of techniques for programming large-scale HPC Systems Approach: Get a broad understanding what HPC is and what can be done Goal: Train general HPC techniques and bits of domain-specific applications Domain-specific Science & Engineering A Domain-specific Molecular Systems High Performance Computing (a field of constant changes) HPC A Course Domain-specific Application Libraries Domain-specific Science & Engineering N 3/ 27

4 Outline of the Course 1. High Performance Computing 2. Parallelization Fundamentals 3. Parallel Programming with MPI 4. Advanced MPI Techniques 5. Parallel Algorithms & Data Structures 6. Parallel Programming with OpenMP 7. Hybrid Programming & Patterns 8. Debugging & Profiling Techniques 9. Performance Optimization & Tools 10. Scalable HPC Infrastructures & GPUs 11. Scientific Visualization & Steering 12. Terrestrial Systems & Climate 13. Systems Biology & Bioinformatics 14. Molecular Systems & Libraries 15. Computational Fluid Dynamics 16. Finite Elements Method 17. Machine Learning & Data Mining 18. Epilogue + additional practical lectures for our hands-on exercises in context 4/ 27

5 Outline Molecular Systems Terminology & Motivation Ab Initio Calculations Molecular Docking Molecular Dynamics Selected Methods & Libraries NAMD CPMD MP2C AMBER Parallel Interoperability Application Short Lecture only shows subsets of libraries & applications in domains Selected promises from previous lecture(s): Lecture 2: Lectures will provide details on applied parallelization methods within parallel applications Lecture 4: Lectures will provide parallel application examples that take advantage of the NetCDF Lecture 5: Lecture 14 will give indepth details on parallel molecular systems algorithms, tools, and methods Lecture 10: Lecture 14 will provide insights into application packages that are able to leverage also GPGPUs Lecture 11: Lectures will provide scientific visualizations for different HPC application domain fields 5/ 27

6 Molecular Systems 6/ 27

7 Molecular Systems Terminology & Motivation Molecular systems refers to techniques to model the behaviour (and dynamics) of molecules Used in many domains (e.g. computational chemistry, biology, materials science, drug design, etc.) Domain-specific usage E.g. used to understand small chemical systems or large biological molecules (e.g. proteins, cf. Lecture 13) E.g. used to study large material assemblies (with many thousands to millions of atoms) Different Approaches Common is to enable atomistic level description of the molecular system A wide variety of methods and libraries evolved over time Parallelization techniques are used to make studying molecular systems possible modified from [4] Wikipedia on Molecular Modelling (molecular model of a protein) (molecules are a group of two or more atoms held together by chemical bonds) (atoms are the smallest unit that defines the chemical elements: chemical substances consisting of a single type of atom) [6] Wikipedia on Atoms [5] Wikipedia on Molecule (consists of electron cloud & nucleus of protons and neutrons) 7/ 27

8 Molecular Systems Terminology Different Approaches Molecular Mechanics (aka coarse granular ) Systems treat atoms as the smallest individual unit Potential energy of systems is calculated using force fields Quantum Chemistry / Mechanics (aka more fine granular ) Experiments of chemical systems with focus of quantum mechanics in physical models Quantum mechanics provides a mathematical description (i.e. equations) of much of the behavior and interactions of energy and matter Explicitly models electrons of each atom (force field) [7] Wikipedia on Molecular Mechanics [8] Wikipedia on Quantum Chemistry (Schroedinger Equation) (describe the behaviour of a particle in a field of force) [9] Wikipedia on Quantum Mechanics 8/ 27

9 Molecular Systems Terminology Ab Initio Calculations Ab initio calculations rely on basic and established laws of nature w/o additional assumptions Ab initio calculations in computational chemistry refer to methods based on quantum chemistry Term often used in science & engineering Also known as from first principles calculations or from the beginning E.g. Terrestrial Systems Terrestrial systems: ab initio calculation of liquid water includes the properties of the constituent hydrogen and oxygen atoms Calculation use these properties with the established laws of electrostatics and quantum mechanics (1 water molecule groups of them water) E.g. Bioinformatics / Biophysics modified from [10] Wikipedia on Ab initio Bioinformatics: term used to define methods for making predictions about biological features using a computational model Biophysics: term used to define methods for the prediction of protein structures in protein folding (cf. Lecture 13) as computational model 9/ 27

10 Molecular Systems Terminology Molecular Docking Molecular docking is a method to determine whether one molecule can bind to another Predicting strength of the association or binding affinity between two molecules with score function Approach E.g. Geometric matching shape complementarity methods describe the protein (cf. Lecture 13) and ligand as a set of features that make them dockable Computational complexity Relatively small (simulation of docking becomes more larger) E.g. nicely parallizable (i.e. HTC) Selected applications fields Used to predict the binding orientation of small molecule drug candidates to their protein targets The binding orientation in turn predict the affinity and activity of the small molecule modified from [11] Wikipedia on Molecular Docking (docking of a small molecule ligand to a protein receptor to produce a complex) (once docked a simulation of time is useful to apply) 10 / 27

11 Molecular Systems Terminology Molecular Dynamics Molecular dynamics (MD) refers to the simulation of physical movement of atoms and molecules Atoms (and whole molecules) interact for a period of time to simulate the motion of the atoms Approach (simplified) Trajectories of atoms/molecules are determined by numerically solving Newton s equation of motion for a system of interacting particles Forces between particles and potential energy are defined by molecular mechanics force fields Computational complexity Complex molecular systems consist of a vast number of particles Impossible to find solutions analytically Use numerical methods (cf. Lecture 12) Selected application fields modified from [12] Wikipedia on Molecular Dynamics Chemical physics, materials science, biomolecule modelling (electron sub-atomic particle) (atoms/molecules microscopic particle) (Iterative simulation method calculating forces and solving the equations of motion based on the accelerations obtained from the new forces) 11 / 27

12 [Video] Molecular Dynamics Summary [13] YouTube Video Introduction to MD 12 / 27

13 Selected Methods & Libraries 13 / 27

14 Selected Methods & Libraries NAMD Nanoscale Molecular Dynamics (NAMD) is a parallel code designed for HPC simulations of large biomolecular systems to analyse trajectories and study molecular systems Selected Facts Uses the molecular graphics program VMD for simulation setup Free software package and open source Parallelization Based on Charm++ (abstraction of MPI, cf. Lecture 3) Scalable to hundreds of cores (beyond 200,000) Example atoms simulation on a 1ns simulation timescale modified from [14] NAMD Webpage 14 / 27

15 Selected Methods & Libraries CPMD Selected Facts Plane wave / pseudopotential implementation of density functional theory (DFT) designed for ab initio molecular dynamics Free software package (for non-profit organization) Parallelization Shared memory with OpenMP (cf. Lecture 6) Distributed memory with MPI (cf. Lecture 3) Example Car Parinello Molecular Dynamics (CPMD) is a parallel program used for ab initio electronic structure and molecular dynamics simulations Parallel simulation of 32 water molecules modified from [15] CPMD Webpage (task group optimization) (w/o task group optimization) 15 / 27

16 Molecular Systems Parallel Algorithms - Revisited Scientific case: understanding physical movements of atoms and molecules in the context of n-body simulations Molecular dynamics algorithms for interacting particles Determine trajectories of atoms and molecules Numerically solving the Newton s equations of motion Forces between particles and potential energy is parallel computed according to molecular mechanics force field methods E.g. MP2C code: particle-based hydrodynamics (fluid simulations) Flow field in a gas diffusion membrane 16 / 27

17 Selected Methods & Libraries MP2C MP2C stands for Massively Parallel Multi-Particle Collision Dynamics MP2C is a molecular dynamics code that focusses on mesoscopic particles Selected Facts Implements a hybrid representation of solvated particles in a fluid Parallelization Based on MPI (cf. Lecture 3) Scalable parallel/io (via SionLib, cf. Lecture 4) Example Flow field in gas diffusion membrane Particle-based hydrodynamics application (highly scalable fluid simulation) [16] MP2C Web page (Flow field in a gas diffusion membrane) Lecture 15 will give further details on computational fluid dynamics (CFD) techniques & codes 17 / 27

18 GPGPUs Selected Applications Revisited Human brain research Example: Registration of high resolution brain images [17] NVidia Application Lab Juelich AMBER / CHARMM applications Molecular dynamics package to simulate molecular dynamics on biomolecules Emerging support for GPGPU processing 18 / 27

19 Selected Methods & Libraries AMBER AMBER stands for Assisted Model Building with Energy Refinement AMBER is a set of molecular mechanical force fields for simulating biomolecules & MD software Selected Facts AMBER is a MD suite of several programs Parallelization Full 3D domain decomposition (cf. Lecture 2) Parallel NetCDF (cf. Lecture 4) Offers CUDA GPU use (cf. Lecture 10) Example (Integrase is one of the three essential enzymes required for replication of the HIC-1 virus) Simulation of HIV-1 integrase enzyime Understanding challenging diseases like retroviruses (e.g. HIV) [18] AMBER 19 / 27

20 Parallel Network Common Data Form (NETCDF) - Revisited NETCDF is a portable and self-describing file format used for array-oriented data (e.g. vectors) Simple vector example: Create a NetCDF data set with one one-dimensional variable A local vector of integers should be allocated and initialized with the task number Each task should write his vector to the NetCDF data set as a part of the global vector (cf. Lecture 4) [19] I/O workshop [20] E. Hartnett 20 / 27

21 Using High Throughput Computing & Grids Revisited Use of Interoperability of two different computing paradigms E.g. use of HTC in EGI and HPC in PRACE infrastructures (Web-based portal make complex infrastructure access easy for biological domain scientists) HTC HPC [21] M. Riedel et al., Research Advances by using interoperable e-science Infrastructures, / 27

22 [Video] Towards Simulating Jökulsárlón [22] YouTube Video A molecule's eye view of ice melting 22 / 27

23 Lecture Bibliography 23 / 27

24 Lecture Bibliography (1) [1] Wikipedia on Monte Carlo, Online: [2] JSC Simulation Lab Biology, Online: [3] J. Archuleta et al., A Maintainable Software Architecture for Fast and Modular Bioinformatics Sequence Search, Sequence Searching, 2007 [4] Wikipedia on Molecular Modelling, Online: [5] Wikipedia on Molecule, Online: [6] Wikipedia on Atom, Online: [7] Wikipedia on Molecular Mechanics, Online: [8] Wikipedia on Quantum Chemistry, Online: [9] Wikipedia on Quantum Mechanics, Online: [10] Wikipedia on Ab initio, Online: [11] Wikipedia on Molecular Docking, Online: 24 / 27

25 Lecture Bibliography (2) [12] Wikipedia on Molecular Dynamics, Outline: [13] YouTube Video, An Introduction to Molecular Dynamics, Online: [14] NAMD Web page, Online: [15] CPMD Web page, Online: [16] MP2C Webpage, Online: [17] Juelich NVidia Application Online: [18] D.A. Case, V. Babin, J.T. Berryman, R.M. Betz, Q. Cai, D.S. Cerutti, T.E. Cheatham, III, T.A. Darden, R.E. Duke, H. Gohlke, A.W. Goetz, S. Gusarov, N. Homeyer, P. Janowski, J. Kaus, I. Kolossváry, A. Kovalenko, T.S. Lee, S. LeGrand, T. Luchko, R. Luo, B. Madej, K.M. Merz, F. Paesani, D.R. Roe, A. Roitberg, C. Sagui, R. Salomon-Ferrer, G. Seabra, C.L. Simmerling, W. Smith, J. Swails, R.C. Walker, J. Wang, R.M. Wolf, X. Wu and P.A. Kollman (2014), AMBER 14, University of California, San Francisco. [19] Wolfgang Frings, Florian Janetzko, Michael Stephan, Portable Parallel I/O - Parallel netcdf, Parallel I/O Workshop, Juelich Supercomputing Centre, 2014, Online: blob=publicationfile [20] E. Hartnett, : NetCDF and HDF5 - HDF5 Workshop / 27

26 Lecture Bibliography (3) [21] M. Riedel et al., Research Advances by using Interoperable e-science Infrastructures, Journal of Cluster Computing, 12(4): , [22] YouTube Video, A molecule's eye view of ice melting, Online: 26 / 27

27 27 / 27