Basics What Does Computational Physics Do? Of .

What does Computational Physics do?Basicsof Computational Physics Atomic Physics studies atomsNuclear Physics studies nuclei What is Computational Physics? Basic computer hardware Plasma Physics studies plasmas Software 1: operating systems Solid State Physics studies solids Software 2: Programming languages Software 3: Problem-solving environmentComputational physics does not studycomputers2What is Computational Physics?What is Computational Physics?“Computational physics is a synthesis oftheoretical analysis, numerical algorithms, andcomputer programming.”Computational physics is a tool for solvingcomplex numerical problems in physicsP. L. DeVries Am. J. Phys., vol. 64, 364 (1996)3Example: university physics courses4Why do we need computational physics? few analyticallysolvable problems(unless you are a poet or a philosopher) solutions to verymany problems5In physics we answer how nature works.Quite often we need equations Using equations we create models to describenatureExact (analytic) solutions are very rare unlessa model is a very simple oneWe need computational physics when– we cannot solve problems analytically– we have too much data to processMany, if not the most, problems in contemporaryphysics could never be solved without computers61

Computational physics in contemporary physics Numerical calculations: solutions of well definedmathematical problems to produce numerical solutionsExamples: systems of differential equations, integration,systems of linear equations, Visualization and animation: the human eye the visualprocessing power of the brain very sophisticated tooltraditional presentation: 2D and 3D plotsnew presentations: animation, using colors and texturesComputer simulation: testing models of natureExamples: weather forecast, Data collection and analysis in experimental researchExample: LabViewSymbolic manipulation:Examples: Maple, Mathematica, 7 Classification of computational modelsDeterministic or Stochastic modelsÖ Deterministic models: Results of deterministic modelsdepend on initial conditions.Ö Stochastic models: an element of chance exists.Dynamic or Static modelsÖ A dynamic models changes in time.Ö A static model does not consider time8Computer Simulation (few examples)9 Molecular Dynamic Simulation9 Weather forecast9 Design of complex systems (aircrafts, )9 Financial markets9 Traffic9 War games9 9more Many natural phenomena are nonlinear, and asmall change in a variable might produce alarge effect.But just few nonlinear problems can be solvedanalytically.Interest in systems with many variables ormany degrees of freedomMillennium Simulation - the largest N-body simulationcarried out thus far (more than 1010 particles).A 3-dimensional visualization of the Millennium Simulationshows a journey through the simulated lennium/1110Computational Physicsis a multidisciplinary fieldComputational s122

ComputersComputers in computational physics Hardware– Amazing progress: two times morepowerful processors in 18 months(Moore’s law: the number of transistorsper square inch on integrated circuitsdoubles every 18 months)– Do we have twice more results in physicseach 18 months? Software13Basic computer hardware14Hardware (internal) CPU - central processing unit (speed in GHz),cache memory: cache 1, cache 2 RAM -random-access memory (MB or GB)communication with CPU by bus (MHz) PCI – Peripheral Component Interconnect USB – Universal Serial Bus HDD – Hard Disk Drive (GB) Graphic card Network Interface (Mb/sec) (modems 56 Kb/sec)15Hardware (peripheral) Keyboard (I/O) Mouse (I/O) Printer (I/O) Monitor (Graphic card) Ethernet (network) Scanner, external storage, 16Critical hardware for calculations 17Desktops– CPU– RAMClusters– CPU and RAM– number of CPUs– fast network communication betweennodeshttp://www.top500.org/183

SoftwareSoftware 1: Operating SystemsOperating system – a set of programs to manage communication between hardware(device drivers)communication between a user and a computerrunning applications (software)file systemsecurity19Types of Operating Systems 20Some of Operating Systemsmulti-user: Allows two or more users to run programs atthe same time. Some operating systems permithundreds or even thousands of concurrent users.multiprocessing: Supports running a program on morethan one CPU.multitasking: Allows more than one program to runconcurrently.multithreading: Allows different parts of a singleprogram to run concurrently.real time: Responds to input instantly. General-purposeoperating systems, (Windows, Linux are not real-time).Alive Windows Linux Mac OS UnixDead by now DOS IBM OS/2 VMS IBM OS/40021What OS is better for computationalphysics?22Software 2: Programming LanguagesThe answer depends on a problem Desktops – Windows, Linux, Mac OSClusters – LinuxSupercomputers – Unix, LinuxParameters to consider:– Available hardware, software andcomputer codes– Stability– Analysis of results and ourse.des/cis400/244

Most common in physicsC/C FortranThe basic ideas behindcomputational physics arelanguage independentJava25“The relevance of C to scientific computing is somewhatcontroversial”What language to use? 26“A First Course in Computational Physics and Object OrientedProgramming with C ”, by D. Yevickchoice depends on a problem– numerical simulation– system programming– web programmingavailable libraries and computer codesexperienceC is also one of the most complex programming languages,with many pitfalls for the unwary.“C a reference manual” (5th edition) by S.P. Harbison and G. L.Steele Jr.Spend your intellectual energies on the current problem - noton fancy tools. When the volume and sophistication of yourproblems demand these weapons you will know it. That is thetime to learn a new tool - and learn it by re-doing an alreadysolved problem, not a new one.F.S. Acton “Real Computing made real27Fortran, C/C , and others Fortran – legacy! Very many computercodes and librariesFortran – easy-to-learn and easy-to-useNormally, scientific C programs cannot beeffectively optimized as Fortran programs(C codes run slower - from 10% to 10times)Java and C# poses formal advantages(however, C is rather for industry)Scientific software – may solve problemsfaster2928Fortran 77, C, C , Fortran 901 – excellent2 – good3 – fair4 – 5

Compilers: C and C Languages and Compilers For effective and efficient work you need toa) Select a language that is right for you now andin nearest future.You may need to know/learn more than one language.b) Have a good book (with examples) to learn thelanguage that you selectedc) Have a compiler that fits you demands(and budget) Dev-C Compiler (open source) a full-featured IDE for the C/C The last version was released in 2005 (no activity with the projectsince that)Code::Blocks IDE (open source) for Windows and MacOpen Watcom C/C compiler, (open source)Microsoft Visual C 2008 Express Edition (free)Borland C Compiler. free and proprietary versionsIntel C Compilers for Windows, Linux and Mac OS (License type proprietary) Non-commercial customers can download free Intel C Compiler for Linux (registration required)Microsoft Visual C 2010 (License type - proprietary)C Builder 2010 C (License type - proprietary)The Portland Group C and C compilers for 32-bit x86 and 64-bit x64processor-based Linux and Windows workstations, servers andclusters.3132Compilers: Fortran Intel Visual Fortran Compilers for Windows, Linux and Mac OS(License type - proprietary)The Portland Group Fortran compilers for Linux and Windowsworkstations, servers and clusters. (License type - proprietary)NAGWare f95 Compiler Available on a wide range of Unix andWindows platforms. (License type - proprietary)Absoft Fortran Compiler and associated tools are available for HPC,Linux, Windows and MacOS environments. (License type - proprietary)Lahey/Fujitsu Fortran includes full Fortran 95/90/77 and Fortran for.NET compilers. (License type - proprietary)Silverfrost FTN95 - Fortran for Windows (free for personal andevaluation use)Open 64 – open source for LinuxOpen Watcom Fortran compiler (free): Latest Release (February 2009)G95 (free): latest release (March 2009) (Open source)Software 3: Problem-solving environment lving environment 34Mathematica is a huge system of remarkable capabilitiescloaked in a stupefying variety of commands. But after sixmonths of frequent experimentation, I still find that threequarters of my time goes into trying to discover why I gotan error message instead of the answer I was.Problem-solving environment is goodfor small and medium projectsProgramming with compiled languagesgives more control, power, flexibility fornumerically and logically intensivetasksF.S. Acton “Real Computing made real”35366

Project Management in computational physicsSoftware 4: Applications The art or skill of directing and organizing the workGraphicsSpreadsheetsWord processorsInternet 37387

(device drivers) communication between a user and a computer running applications (software) file system security Operating system – a set of programs to manage 21 Types of Operating Systems multi-user: Allows two or more users to run programs at the same time. Some operating syste