DOCSLIB.ORG

Lecture 7: Parallelism

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Lecture 7: Parallelism Lecture 7: Parallelism Fall 2018 Jason Tang Slides based upon Operating System Concept slides, http://codex.cs.yale.edu/avi/os-book/OS9/slide-dir/index.html Copyright Silberschatz, Galvin, and Gagne, 2013 "1 Topics • Overview of Threading! • Multithreading Models! • Pthreads Library! • Implicit Threading "2 Why Threading? • Modern programs, including kernels, are usually multithreaded! • Example: Update display, while! • Handling key strokes, while! • Sending network messages! • Process creation is heavy-weight, thread creation is light-weight! • Can simplify code and increase computer usage "3 Single and Multithreaded Processes "4 Multithreading Benefits • Responsiveness - continue processing while waiting for other I/O, especially in GUIs! • Resource Sharing - because threads share the process’s resources, they may be more e$cient than having di%erent processes communicate via messages! • Economy - thread switching usually faster than process context switching! • Scalability - can run in parallel across multiple processors "5 Concurrency vs. Parallelism • Concurrency - Multiple processes making progress! • If only one CPU, no true parallelism; OS rapidly switches between processes to give illusion of parallelism! • Parallelism - Multiple processes simultaneously making progress! • Modern computers have multiple cores, with true parallelism! • Challenge is to write programs that run in parallel, which is a very hard problem "6 Concurrency vs. Parallelism • Concurrent execution on single core system:! • Concurrent and parallelism on a multicore system: "7 Types of Parallelism • Data Parallelism - distributes subsets of same data across multiple cores; same operation run on each core! • Examples: ray tracer, search engine, BitCoin miner! • Task Parallelism - distributes threads across cores, with each thread performing unique operations! • Examples: web server, database server! • In modern software, programs tend to have both data and task parallelism "8 Amdahl’s Law • Programs have a mix of serial (non-parallel) and parallel components! • Let S = percent of program that must be serial, and N = number of cores! • Example: if process foo is 75% parallel and 25% serial, then by going from 1 to 2 cores, the expected speedup is 1.6x! • As N approaches infinity, speedup approaches S-1 "9 User Threads and Kernel Threads • User Threads - threading handled entirely by user-level threading library; kernel is unaware of any threading! • Examples: Java “green” threads, Stackless Python, Tcl! • Kernel Threads - kernel supported threading! • Supported by nearly all modern OSes (Windows, macOS, Linux, Unix, …) "10 Many-to-One Multithreading Model • Many user-level threads mapped to single kernel thread! • If one thread blocks, entire process blocks! • Only model possible if no kernel support! • Threading library handles scheduling! • Faster to schedule (no context switching)! • Unable to take advantage of multiple processors "11 One-to-One Multithreading Model • Each thread created by user maps to a kernel thread! • OS schedules each thread independently! • Each thread could run on separate processor! • OS may limit number of threads per process! • Used by Windows, Linux, etc. "12 Many-to-Many (Hybrid) Multithreading Model • Many user-level threads map to many kernel threads! • Threading library handles scheduling! • OS creates kernel threads based upon number of processors available! • Much more complex to implement! • Available in Windows 7, older versions of FreeBSD, and in experimental OSes "13 Thread Libraries • Thread Library - provides programmer an API for creating and managing threads! • Threads could be entirely user-space construct (green threads)! • Or library invokes system calls to request OS to create threads! • POSIX Threads (Pthreads) - standard API for threading, used by Linux, macOS, and other Unix-like systems "14 About Pthreads • Pthreads library specifies thread creation, management, synchronizations (next lecture)! • Add #include <pthread.h> to programs! • On Linux, compile and link with -pthread flag! gcc -o foo foo.c -pthread • See man 7 pthreads for an overview! • On Ubuntu, make sure you have installed the manpages-dev package (sudo apt-get install manpages-dev) "15 Thread Creation int pthread_create(pthread_t *thread, const pthread_attr_t *attr, void *(*start_routine) (void *), void *arg) • Use to create a thread! • First parameter is address to store new thread’s unique ID! • Second parameter gives thread settings, or set to NULL for default settings! • Third parameter gives starting address for thread! • Fourth parameter is for arbitrary data to be passed into thread! • Return value is 0 on success, or negative on error "16 Starting Routine void *thread_func(void *) { … } • Function where newly created thread will start running! • Fourth parameter of pthread_create() passed as this function’s parameter! • When this function returns, thread will terminate! • OS stores function’s return value until later retrieved! • In Pthreads, the new thread will be begin in the ready state (assuming thread creation was successful)! • In Windows, two-step process (create thread, and then start running it) "17 Joining Threads int pthread_join(pthread_t thread, void **retval); • Just as with processes, Linux stores a thread’s return status, until something later reaps the thread! • In Linux, threads and processes are implemented uniformly as tasks! • This function blocks until target thread terminates! • First parameter is target thread to join! • Second parameter gives location to store return status, or NULL to ignore return status "18 Example Threading Code #1 #include <pthread.h> #include <stdio.h> static void *example1(void *args) { printf("In thread\n"); return NULL; } int main(void) { pthread_t tid; pthread_create(&tid, NULL, example1, NULL); printf("In main thread\n"); pthread_join(tid, NULL); printf("After join\n"); return 0; } "19 Example Threading Code #2 #include <pthread.h> #include <stdio.h> static void *example2(void *args) { int my_ID = *(int *)(args); printf("I am thread %d\n", my_ID); return NULL; } int main(void) { int i, data[4]; pthread_t tids[4]; for (i = 0; i < 4; i++) { data[i] = i; pthread_create(tids + i, NULL, example2, data + i); } for (i = 0; i < 4; i++) { pthread_join(tids[i], NULL); } printf("After join\n"); return 0; } "20 Example Threading Code #3 #include <pthread.h> #include <stdio.h> static void *example2(void *args) { int my_ID = *(int *)(args); printf("I am thread %d\n", my_ID); return NULL; } int main(void) { int i, data[4]; pthread_t tids[4]; for (i = 0; i < 4; i++) { pthread_create(tids + i, NULL, example2, &i); } for (i = 0; i < 4; i++) { pthread_join(tids[i], NULL); } printf("After join\n"); return 0; } "21 Implicit Threading • Creation and thread management automatically done by compilers and run- time libraries, instead of programmers! • Examples:! • Thread pools! • OpenMP! • Grand Central Dispatch "22 Thread Pools • Pre-create many threads, in a pool awaiting work! • Usually faster to service request when thread already exists, rather than creating a new thread each time! • Number of threads can be customized based upon application workload, amount of RAM, number of CPUs! • Starting with Windows Vista, a process can push work into a thread pool! • Within Linux kernel code, lower-priority tasks run within workqueues "23 OpenMP • Compiler directives that will #include <stdio.h> automatically parallelize #include <math.h> loops! int main(void) { int i; • Creates threads when double vals[1000000], sum = 0; compiler detects parallel #pragma omp parallel for regions! for (i = 0; i < 1000000; i++) { vals[i] = sqrt(i) + logf(i + 1); } • Available for C, C++, Fortran, running on modern /* this loop cannot be optimized */ operating OSes! for (i = 0; i < 1000000; i++) { sum += vals[i]; } • Used in scientific computing printf("Sum is %f\n", sum); return 0; } "24 Grand Central Dispatch • Designed by Apple for macOS and iOS, but has been ported to other OSes! • Programmer designates parts of their program that can be run in parallel! • GCD automatically dispatches that code into thread(s)! • GCD abstracts thread creation and joining! • OS determines how many threads to create based upon system load "25.
Load more

Recommended publications
  • Events, Co-Routines, Continuations and Threads OS (And Application)Execution Models System Building
    Events, Co-routines, Continuations and Threads OS (and application)Execution Models System Building General purpose systems need to deal with • Many activities – potentially overlapping – may be interdependent • Activities that depend on external phenomena – may requiring waiting for completion (e.g. disk read) – reacting to external triggers (e.g. interrupts) Need a systematic approach to system structuring © Kevin Elphinstone 2 Construction Approaches Events Coroutines Threads Continuations © Kevin Elphinstone 3 Events External entities generate (post) events. • keyboard presses, mouse clicks, system calls Event loop waits for events and calls an appropriate event handler. • common paradigm for GUIs Event handler is a function that runs until completion and returns to the event loop. © Kevin Elphinstone 4 Event Model The event model only requires a single stack Memory • All event handlers must return to the event loop CPU Event – No blocking Loop – No yielding PC Event SP Handler 1 REGS Event No preemption of handlers Handler 2 • Handlers generally short lived Event Handler 3 Data Stack © Kevin Elphinstone 5 What is ‘a’? int a; /* global */ int func() { a = 1; if (a == 1) { a = 2; } No concurrency issues within a return a; handler } © Kevin Elphinstone 6 Event-based kernel on CPU with protection Kernel-only Memory User Memory CPU Event Loop Scheduling? User PC Event Code SP Handler 1 REGS Event Handler 2 User Event Data Handler 3 Huh? How to support Data Stack multiple Stack processes? © Kevin Elphinstone 7 Event-based kernel on CPU with protection Kernel-only Memory User Memory CPU PC Trap SP Dispatcher User REGS Event Code Handler 1 User-level state in PCB Event PCB Handler 2 A User Kernel starts on fresh Timer Event Data stack on each trap (Scheduler) PCB B No interrupts, no blocking Data Current in kernel mode Thead PCB C Stack Stack © Kevin Elphinstone 8 Co-routines Originally described in: • Melvin E.
    [Show full text]
  • Designing an Ultra Low-Overhead Multithreading Runtime for Nim
    Designing an ultra low-overhead multithreading runtime for Nim Mamy Ratsimbazafy Weave [email protected] https://github.com/mratsim/weave Hello! I am Mamy Ratsimbazafy During the day blockchain/Ethereum 2 developer (in Nim) During the night, deep learning and numerical computing developer (in Nim) and data scientist (in Python) You can contact me at [email protected] Github: mratsim Twitter: m_ratsim 2 Where did this talk came from? ◇ 3 years ago: started writing a tensor library in Nim. ◇ 2 threading APIs at the time: OpenMP and simple threadpool ◇ 1 year ago: complete refactoring of the internals 3 Agenda ◇ Understanding the design space ◇ Hardware and software multithreading: definitions and use-cases ◇ Parallel APIs ◇ Sources of overhead and runtime design ◇ Minimum viable runtime plan in a weekend 4 Understanding the 1 design space Concurrency vs parallelism, latency vs throughput Cooperative vs preemptive, IO vs CPU 5 Parallelism is not 6 concurrency Kernel threading 7 models 1:1 Threading 1 application thread -> 1 hardware thread N:1 Threading N application threads -> 1 hardware thread M:N Threading M application threads -> N hardware threads The same distinctions can be done at a multithreaded language or multithreading runtime level. 8 The problem How to schedule M tasks on N hardware threads? Latency vs 9 Throughput - Do we want to do all the work in a minimal amount of time? - Numerical computing - Machine learning - ... - Do we want to be fair? - Clients-server - Video decoding - ... Cooperative vs 10 Preemptive Cooperative multithreading:
    [Show full text]
  • The Essentials of Stackless Python Tuesday, 10 July 2007 10:00 (30 Minutes)
    EuroPython 2007 Contribution ID: 62 Type: not specified The Essentials of Stackless Python Tuesday, 10 July 2007 10:00 (30 minutes) This is a re-worked, actualized and improved version of my talk at PyCon 2007. Repeating the abstract: As a surprise for people who think they know Stackless, we present the new Stackless implementation For PyPy, which has led to a significant amount of new insight about parallel programming and its possible implementations. We will isolate the known Stackless as a special case of a general concept. This is a Stackless, not a PyPy talk. But the insights presented here would not exist without PyPy’s existance. Summary Stackless has been around for a long time now. After several versions with different goals in mind, the basic concepts of channels and tasklets turned out to be useful abstractions, and since many versions, Stackless is only ported from version to version, without fundamental changes to the principles. As some spin-off, Armin Rigo invented Greenlets at a Stackless sprint. They are some kind of coroutines and a bit of special semantics. The major benefit is that Greenlets can runon unmodified CPython. In parallel to that, the PyPy project is in its fourth year now, and one of its goals was Stackless integration as an option. And of course, Stackless has been integrated into PyPy in a very nice and elegant way, much nicer than expected. During the design of the Stackless extension to PyPy, it turned out, that tasklets, greenlets and coroutines are not that different in principle, and it was possible to base all known parallel paradigms on one simple coroutine layout, which is as minimalistic as possible.
    [Show full text]
  • Method, 224 Vs. Typing Module Types
    Index A migration metadata, 266 running migrations, 268, 269 Abstract base classes, 223 setting up a new project, 264 @abc.abstractmethod decorator, 224 using constants in migrations, 267 __subclasshook__(...) method, 224 __anext__() method, 330 vs. typing module types, 477 apd.aggregation package, 397, 516, 524 virtual subclasses, 223 clean functions (see clean functions) Actions, 560 database, 254 analysis process, 571, 572 get_data_by_deployment(...) config file, 573, 575 function, 413, 416 DataProcessor class, 561 get_data(...) function, 415 extending, 581 plot_sensor(...) function, 458 IFTTT service, 566 plotting data, 429 ingesting data, 567, 571 plotting functions, 421 logs, 567 query functions, 417 trigger, 563, 564 apd.sensors package, 106 Adafruit, 42, 486 APDSensorsError, 500 Adapter pattern, 229 DataCollectionError, 500 __aenter__() method, 331, 365 directory structure, 108 __aexit__(...) method, 331, 365 extending, 149 Aggregation process, 533 IntermittentSensorFailureError, 500 aiohttp library, 336 making releases, 141 __aiter__() method, 330 sensors script, 32, 36, 130, 147, 148, Alembic, 264 153, 155, 156, 535 ambiguous changes, 268 UserFacingCLIError, 502 creating a new revision, 265 apd.sunnyboy_solar package, 148, current version, 269 155, 173 downgrading, 268, 270 API design, 190 irreversible, migrations, 270 authentication, 190 listing migrations, 269 versioning, 240, 241, 243 merging, migrations, 269 589 © Matthew Wilkes 2020 M. Wilkes, Advanced Python Development, https://doi.org/10.1007/978-1-4842-5793-7 INDEX AssertionError,
    [Show full text]
  • Python Concurrency
    Python Concurrency Threading, parallel and GIL adventures Chris McCafferty, SunGard Global Services Overview • The free lunch is over – Herb Sutter • Concurrency – traditionally challenging • Threading • The Global Interpreter Lock (GIL) • Multiprocessing • Parallel Processing • Wrap-up – the Pythonic Way Reminder - The Free Lunch Is Over How do we get our free lunch back? • Herb Sutter’s paper at: • http://www.gotw.ca/publications/concurrency-ddj.htm • Clock speed increase is stalled but number of cores is increasing • Parallel paths of execution will reduce time to perform computationally intensive tasks • But multi-threaded development has typically been difficult and fraught with danger Threading • Use the threading module, not thread • Offers usual helpers for making concurrency a bit less risky: Threads, Locks, Semaphores… • Use logging, not print() • Don’t start a thread in module import (bad) • Careful importing from daemon threads Traditional management view of Threads Baby pile of snakes, Justin Guyer Managing Locks with ‘with’ • With keyword is your friend • (compare with the ‘with file’ idiom) import threading rlock = threading.RLock() with rlock: print "code that can only be executed while we acquire rlock" #lock is released at end of code block, regardless of exceptions Atomic Operations in Python • Some operations can be pre-empted by another thread • This can lead to bad data or deadlocks • Some languages offer constructs to help • CPython has a set of atomic operations due to the operation of something called the GIL and the way the underlying C code is implemented • This is a fortuitous implementation detail – ideally use RLocks to future-proof your code CPython Atomic Operations • reading or replacing a single instance attribute • reading or replacing a single global variable • fetching an item from a list • modifying a list in place (e.g.
    [Show full text]
  • Thread-Level Parallelism II
    Great Ideas in UC Berkeley UC Berkeley Teaching Professor Computer Architecture Professor Dan Garcia (a.k.a. Machine Structures) Bora Nikolić Thread-Level Parallelism II Garcia, Nikolić cs61c.org Languages Supporting Parallel Programming ActorScript Concurrent Pascal JoCaml Orc Ada Concurrent ML Join Oz Afnix Concurrent Haskell Java Pict Alef Curry Joule Reia Alice CUDA Joyce SALSA APL E LabVIEW Scala Axum Eiffel Limbo SISAL Chapel Erlang Linda SR Cilk Fortan 90 MultiLisp Stackless Python Clean Go Modula-3 SuperPascal Clojure Io Occam VHDL Concurrent C Janus occam-π XC Which one to pick? Garcia, Nikolić Thread-Level Parallelism II (3) Why So Many Parallel Programming Languages? § Why “intrinsics”? ú TO Intel: fix your #()&$! compiler, thanks… § It’s happening ... but ú SIMD features are continually added to compilers (Intel, gcc) ú Intense area of research ú Research progress: 20+ years to translate C into good (fast!) assembly How long to translate C into good (fast!) parallel code? • General problem is very hard to solve • Present state: specialized solutions for specific cases • Your opportunity to become famous! Garcia, Nikolić Thread-Level Parallelism II (4) Parallel Programming Languages § Number of choices is indication of ú No universal solution Needs are very problem specific ú E.g., Scientific computing/machine learning (matrix multiply) Webserver: handle many unrelated requests simultaneously Input / output: it’s all happening simultaneously! § Specialized languages for different tasks ú Some are easier to use (for some problems)
    [Show full text]
  • Uwsgi Documentation Release 1.9
    uWSGI Documentation Release 1.9 uWSGI February 08, 2016 Contents 1 Included components (updated to latest stable release)3 2 Quickstarts 5 3 Table of Contents 11 4 Tutorials 137 5 Articles 139 6 uWSGI Subsystems 141 7 Scaling with uWSGI 197 8 Securing uWSGI 217 9 Keeping an eye on your apps 223 10 Async and loop engines 231 11 Web Server support 237 12 Language support 251 13 Release Notes 317 14 Contact 359 15 Donate 361 16 Indices and tables 363 Python Module Index 365 i ii uWSGI Documentation, Release 1.9 The uWSGI project aims at developing a full stack for building (and hosting) clustered/distributed network applica- tions. Mainly targeted at the web and its standards, it has been successfully used in a lot of different contexts. Thanks to its pluggable architecture it can be extended without limits to support more platforms and languages. Cur- rently, you can write plugins in C, C++ and Objective-C. The “WSGI” part in the name is a tribute to the namesake Python standard, as it has been the first developed plugin for the project. Versatility, performance, low-resource usage and reliability are the strengths of the project (and the only rules fol- lowed). Contents 1 uWSGI Documentation, Release 1.9 2 Contents CHAPTER 1 Included components (updated to latest stable release) The Core (implements configuration, processes management, sockets creation, monitoring, logging, shared memory areas, ipc, cluster membership and the uWSGI Subscription Server) Request plugins (implement application server interfaces for various languages and platforms: WSGI, PSGI, Rack, Lua WSAPI, CGI, PHP, Go ...) Gateways (implement load balancers, proxies and routers) The Emperor (implements massive instances management and monitoring) Loop engines (implement concurrency, components can be run in preforking, threaded, asynchronous/evented and green thread/coroutine modes.
    [Show full text]
  • An Ideal Match?
    24 November 2020 An ideal match? Investigating how well-suited Concurrent ML is to implementing Belief Propagation for Stereo Matching James Cooper [email protected] OutlineI 1 Stereo Matching Generic Stereo Matching Belief Propagation 2 Concurrent ML Overview Investigation of Alternatives Comparative Benchmarks 3 Concurrent ML and Belief Propagation 4 Conclusion Recapitulation Prognostication 5 References Outline 1 Stereo Matching Generic Stereo Matching Belief Propagation 2 Concurrent ML Overview Investigation of Alternatives Comparative Benchmarks 3 Concurrent ML and Belief Propagation 4 Conclusion Recapitulation Prognostication 5 References Outline 1 Stereo Matching Generic Stereo Matching Belief Propagation 2 Concurrent ML Overview Investigation of Alternatives Comparative Benchmarks 3 Concurrent ML and Belief Propagation 4 Conclusion Recapitulation Prognostication 5 References Stereo Matching Generally SM is finding correspondences between stereo images Images are of the same scene Captured simultaneously Correspondences (`disparity') are used to estimate depth SM is an ill-posed problem { can only make best guess Impossible to perform `perfectly' in general case Stereo Matching ExampleI (a) Left camera's image (b) Right camera's image Figure 1: The popular 'Tsukuba' example stereo matching images, so called because they were created by researchers at the University of Tsukuba, Japan. They are probably the most widely-used benchmark images in stereo matching. Stereo Matching ExampleII (a) Ground truth disparity map (b) Disparity map generated using a simple Belief Propagation Stereo Matching implementation Figure 2: The ground truth disparity map for the Tsukuba images, and an example of a possible real disparity map produced by using Belief Propagation Stereo Matching. The ground truth represents what would be expected if stereo matching could be carried out `perfectly'.
    [Show full text]
  • Tcl and Java Performance
    Tcl and Java Performance http://ptolemy.eecs.berkeley.edu/~cxh/java/tclblend/scriptperf/scriptperf.html Tcl and Java Performance by H. John Reekie, University of California at Berkeley Christopher Hylands, University of California at Berkeley Edward A. Lee, University of California at Berkeley Abstract Combining scripting languages such as Tcl with lower−level programming languages such as Java offers new opportunities for flexible and rapid software development. In this paper, we benchmark various combinations of Tcl and Java against the two languages alone. We also provide some comparisons with JavaScript. Performance can vary by well over two orders of magnitude. We also uncovered some interesting threading issues that affect performance on the Solaris platform. "There are lies, damn lies and statistics" This paper is a work in progress, we used the information here to give our group some generalizations on the performance tradeoffs between various scripting languages. Updating the timing results to include JDK1.2 with a Just In Time (JIT) compiler would be useful. Introduction There is a growing trend towards integration of multiple languages through scripting. In a famously controversial white paper (Ousterhout 97), John Ousterhout, now of Scriptics Corporation, argues that scripting −− the use of a high−level, untyped, interpreted language to "glue" together components written in a lower−level language −− provides greater reuse benefits that other reuse technologies. Although traditionally a language such as C or C++ has been the lower−level language, more recent efforts have focused on using Java. Recently, Sun Microsystems laboratories announced two products aimed at fulfilling this goal with the Tcl and Java programming languages.
    [Show full text]
  • Goless Documentation Release 0.6.0
    goless Documentation Release 0.6.0 Rob Galanakis July 11, 2014 Contents 1 Intro 3 2 Goroutines 5 3 Channels 7 4 The select function 9 5 Exception Handling 11 6 Examples 13 7 Benchmarks 15 8 Backends 17 9 Compatibility Details 19 9.1 PyPy................................................... 19 9.2 Python 2 (CPython)........................................... 19 9.3 Python 3 (CPython)........................................... 19 9.4 Stackless Python............................................. 20 10 goless and the GIL 21 11 References 23 12 Contributing 25 13 Miscellany 27 14 Indices and tables 29 i ii goless Documentation, Release 0.6.0 • Intro • Goroutines • Channels • The select function • Exception Handling • Examples • Benchmarks • Backends • Compatibility Details • goless and the GIL • References • Contributing • Miscellany • Indices and tables Contents 1 goless Documentation, Release 0.6.0 2 Contents CHAPTER 1 Intro The goless library provides Go programming language semantics built on top of gevent, PyPy, or Stackless Python. For an example of what goless can do, here is the Go program at https://gobyexample.com/select reimplemented with goless: c1= goless.chan() c2= goless.chan() def func1(): time.sleep(1) c1.send(’one’) goless.go(func1) def func2(): time.sleep(2) c2.send(’two’) goless.go(func2) for i in range(2): case, val= goless.select([goless.rcase(c1), goless.rcase(c2)]) print(val) It is surely a testament to Go’s style that it isn’t much less Python code than Go code, but I quite like this. Don’t you? 3 goless Documentation, Release 0.6.0 4 Chapter 1. Intro CHAPTER 2 Goroutines The goless.go() function mimics Go’s goroutines by, unsurprisingly, running the routine in a tasklet/greenlet.
    [Show full text]
  • Continuations and Stackless Python
    Continuations and Stackless Python Or "How to change a Paradigm of an existing Program" Christian Tismer Virtual Photonics GmbH mailto:[email protected] Abstract 2 Continuations In this paper, an implementation of "Stackless Python" (a Python which does not keep state on the C 2.1 What is a Continuation? stack) is presented. Surprisingly, the necessary changes affect just a small number of C modules, and Many attempts to explain continuations can be found a major rewrite of the C library can be avoided. The in the literature[5-10], more or less hard to key idea in this approach is a paradigm change for the understand. The following is due to Jeremy Hylton, Python code interpreter that is not easy to understand and I like it the best. Imagine a very simple series of in the first place. Recursive interpreter calls are statements: turned into tail recursion, which allows deferring evaluation by pushing frames to the frame stack, x = 2; y = x + 1; z = x * 2 without the C stack involved. In this case, the continuation of x=2 is y=x+1; By decoupling the frame stack from the C stack, we z=x*2. You might think of the second and third now have the ability to keep references to frames and assignments as a function (forgetting about variable to do non-local jumps. This turns the frame stack into scope for the moment). That function is the a tree, and every leaf of the tree can now be a jump continuation. In essence, every single line of code has target.
    [Show full text]
  • Thread Scheduling in Multi-Core Operating Systems Redha Gouicem
    Thread Scheduling in Multi-core Operating Systems Redha Gouicem To cite this version: Redha Gouicem. Thread Scheduling in Multi-core Operating Systems. Computer Science [cs]. Sor- bonne Université, 2020. English. tel-02977242 HAL Id: tel-02977242 https://hal.archives-ouvertes.fr/tel-02977242 Submitted on 24 Oct 2020 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Ph.D thesis in Computer Science Thread Scheduling in Multi-core Operating Systems How to Understand, Improve and Fix your Scheduler Redha GOUICEM Sorbonne Université Laboratoire d’Informatique de Paris 6 Inria Whisper Team PH.D.DEFENSE: 23 October 2020, Paris, France JURYMEMBERS: Mr. Pascal Felber, Full Professor, Université de Neuchâtel Reviewer Mr. Vivien Quéma, Full Professor, Grenoble INP (ENSIMAG) Reviewer Mr. Rachid Guerraoui, Full Professor, École Polytechnique Fédérale de Lausanne Examiner Ms. Karine Heydemann, Associate Professor, Sorbonne Université Examiner Mr. Etienne Rivière, Full Professor, University of Louvain Examiner Mr. Gilles Muller, Senior Research Scientist, Inria Advisor Mr. Julien Sopena, Associate Professor, Sorbonne Université Advisor ABSTRACT In this thesis, we address the problem of schedulers for multi-core architectures from several perspectives: design (simplicity and correct- ness), performance improvement and the development of application- specific schedulers.
    [Show full text]