DOCSLIB.ORG

Openmp 3.1 API C/C++ Syntax Quick Reference Card

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Openmp 3.1 API C/C++ Syntax Quick Reference Card OpenMP API 3.1 C/C++ Page 1 OpenMP 3.1 API C/C++ Syntax Quick Reference Card OpenMP Application Program Interface(API) is a portable, scalable OpenMP supports multi-platform shared-memory parallel model that gives shared-memory parallel programmers a simple programming in C/C++ and Fortran on all architectures, including ® and flexible interface for developing parallel applications for Unix platforms and Windows NT platforms. platforms ranging from the desktop to the supercomputer. A separate OpenMP reference card for Fortran is also available. C/C++ [n.n.n] refers to sections in the OpenMP API Specification available at www.openmp.org. Directives An OpenMP executable directive applies to the Single [2.5.3] Taskyield [2.7.2] succeeding structured block or an OpenMP Construct. The single construct specifies that the associated The taskyield construct specifies that the current task can A structured-block is a single statement or a compound structured block is executed by only one of the threads be suspended in favor of execution of a different task. statement with a single entry at the top and a single exit in the team (not necessarily the master thread), in the #pragma omp taskyield at the bottom. context of its implicit task. Master [2.8.1] Parallel [2.4] #pragma omp single [clause[ [, ]clause] ...] structured-block The master construct specifies a structured block that The parallel construct forms a team of threads and starts is executed by the master thread of the team. There is parallel execution. clause: no implied barrier either on entry to, or exit from, the private(list) #pragma omp parallel [clause[ [, ]clause] ...] master construct. structured-block firstprivate(list) #pragma omp master clause: copyprivate(list) structured-block if(scalar-expression) nowait Critical [2.8.2] num_threads(integer-expression) Parallel Loop [2.6.1] The critical construct restricts execution of the associated default(shared | none) The parallel loop construct is a shortcut for specifying structured block to a single thread at a time. private(list) a parallel construct containing one or more associated #pragma omp critical[ (name)] firstprivate(list) loops and no other statements. structured-block shared(list) #pragma omp parallel for [clause[ [, ]clause] ...] copyin(list) for-loop Barrier [2.8.3] reduction(operator: list) clause: The barrier construct specifies an explicit barrier at the Any accepted by the parallel or for directives, except point at which the construct appears. Loop [2.5.1] the nowait clause, with identical meanings and #pragma omp barrier The loop construct specifies that the iterations of restrictions. loops will be distributed among and executed by the Taskwait [2.8.4] encountering team of threads. The taskwait construct specifies a wait on the completion #pragma omp for [clause[ [, ]clause] ...] Simple Parallel Loop Example of child tasks of the current task. for-loops The following example demonstrates how to #pragma omp taskwait clause: parallelize a simple loop using the parallel loop private(list) construct. Atomic [2.8.5] firstprivate(list) The atomic construct ensures that a specific storage Most common form location is updated atomically, rather than exposing it to lastprivate(list) void simple(int n, float *a, float *b) of the for loop: the possibility of multiple, simultaneous writing threads. reduction(operator: list) { schedule(kind[, chunk_size]) for(var = lb; int i; #pragma omp atomic [read | write | update | capture] expression-stmt collapse(n) var relational-op b; #pragma omp parallel for ordered var += incr) for (i=1; i<n; i++) /* i is private by default */ #pragma omp atomic capture structured-block nowait b[i] = (a[i] + a[i-1]) / 2.0; } where expression-stmt may be one of the following forms: kind: • static: Iterations are divided into chunks of size if clause is... expression-stmt: chunk_size. Chunks are assigned to threads in the Parallel Sections [2.6.2] read v = x; team in round-robin fashion in order of thread The parallel sections construct is a shortcut for specifying write x = expr; number. a parallel construct containing one sections construct and update or x++; x--; ++x; • dynamic: Each thread executes a chunk of iterations no other statements. is not present --x; x binop = expr; x = x binop expr; then requests another chunk until no chunks remain #pragma omp parallel sections [clause[ [, ]clause] ...] capture v = x++; v = x--; v = ++x; to be distributed. v = --x; v = x binop= expr; • guided: Each thread executes a chunk of iterations { then requests another chunk until no chunks remain [#pragma omp section] and structured-block may be one of the following forms: to be assigned. The chunk sizes start large and shrink structured-block {v = x; x binop= expr;} {x binop= expr; v = x;} to the indicated chunk_size as chunks are scheduled. [#pragma omp section {v = x; x = x binop expr;} {x = x binop expr; v = x;} structured-block] {v = x; x++;} {v = x; ++x;} {++x; v = x;} {x++; v = x;} • auto: The decision regarding scheduling is delegated ... to the compiler and/or runtime system. {v = x; x--;} {v = x; --x;} {--x; v = x;} {x--; v = x;} } • runtime: The schedule and chunk size are taken from Flush [2.8.6] the run-sched-var ICV. clause: Any of the clauses accepted by the parallel or sections The flush construct executes the OpenMP flush operation, which makes a thread’s temporary view of Sections [2.5.2] directives, except the nowait clause, with identical memory consistent with memory, and enforces an order The sections construct contains a set of structured blocks meanings and restrictions. on the memory operations of the variables. that are to be distributed among and executed by the encountering team of threads. Task [2.7.1] #pragma omp flush[ (list)] #pragma omp sections[clause[[ ,] clause] ...] The task construct defines an explicit task. The data Ordered [2.8.7] { environment of the task is created according to the The ordered construct specifies a structured block in a [#pragma omp section] data-sharing attribute clauses on the task construct loop region that will be executed in the order of the loop structured-block and any defaults that apply. iterations. This sequentializes and orders the code within [#pragma omp section #pragma omp task [clause[ [, ]clause] ...] an ordered region while allowing code outside the region structured-block] structured-block to run in parallel. ... clause: #pragma omp ordered } if(scalar-expression) structured-block final(scalar-expression) clause: Threadprivate [2.9.2] untied The threadprivate directive specifies that variables are private(list) default(shared | none) replicated, with each thread having its own copy. firstprivate(list) mergeable lastprivate(list) #pragma omp threadprivate(list) private(list) list: reduction(operator: list) firstprivate(list) nowait A comma-separated list of file-scope, namespace- shared(list) scope, or static block-scope variables that do not have incomplete types. © 2011 OpenMP ARB OMP0711C OpenMP API 3.1 C/C++ Page 2 Runtime Library Routines Execution Environment Routines [3.2] int omp_in_parallel(void); int omp_get_thread_limit(void); Lock Routines [3.3] Execution environment routines affect Returns true if the call to the routine is Returns the value of the thread-limit-var Lock routines support synchronization with and monitor threads, processors, and the enclosed by an active parallel region; ICV, which is the maximum number of OpenMP locks. parallel environment. otherwise, it returns false. OpenMP threads available to the program. void omp_init_lock(omp_lock_t *lock); void omp_set_num_threads( void omp_set_dynamic( void omp_set_max_active_levels( void omp_init_nest_lock( int num_threads); int dynamic_threads); int max_levels); omp_nest_lock_t *lock); Affects the number of threads used for Enables or disables dynamic adjustment of Limits the number of nested active parallel These routines initialize an OpenMP lock. subsequent parallel regions that do not the number of threads available by setting regions, by setting max-active-levels-var specify a num_threads clause. the value of the dyn-var ICV. ICV. void omp_destroy_lock(omp_lock_t *lock); int omp_get_num_threads(void); void omp_destroy_nest_lock( int omp_get_dynamic(void); int omp_get_max_active_levels(void); omp_nest_lock_t *lock); Returns the number of threads in the Returns the value of the dyn-var ICV, Returns the value of max-active-levels-var current team. These routines ensure that the OpenMP determining whether dynamic adjustment ICV, which determines the maximum lock is uninitialized. of the number of threads is enabled or int omp_get_max_threads(void); number of nested active parallel regions. disabled. void omp_set_lock(omp_lock_t *lock); Returns maximum number of threads that int omp_get_level(void); could be used to form a new team using a void omp_set_nested(int nested); void omp_set_nest_lock( Returns the number of nested parallel omp_nest_lock_t *lock); parallel construct without a num_threads Enables or disables nested parallelism, by regions enclosing the task that contains clause. setting the nest-var ICV. These routines provide a means of setting the call. an OpenMP lock. int omp_get_thread_num(void); int omp_get_nested(void); int omp_get_ancestor_thread_num( void omp_unset_lock(omp_lock_t *lock); Returns the ID of the encountering thread Returns the value of the nest-var ICV, which int level); void omp_unset_nest_lock( where ID ranges from zero to the size of determines if nested parallelism is enabled Returns, for a given nested level
Load more

Recommended publications
  • Heterogeneous Task Scheduling for Accelerated Openmp
    Heterogeneous Task Scheduling for Accelerated OpenMP ? ? Thomas R. W. Scogland Barry Rountree† Wu-chun Feng Bronis R. de Supinski† ? Department of Computer Science, Virginia Tech, Blacksburg, VA 24060 USA † Center for Applied Scientific Computing, Lawrence Livermore National Laboratory, Livermore, CA 94551 USA [email protected] [email protected] [email protected] [email protected] Abstract—Heterogeneous systems with CPUs and computa- currently requires a programmer either to program in at tional accelerators such as GPUs, FPGAs or the upcoming least two different parallel programming models, or to use Intel MIC are becoming mainstream. In these systems, peak one of the two that support both GPUs and CPUs. Multiple performance includes the performance of not just the CPUs but also all available accelerators. In spite of this fact, the models however require code replication, and maintaining majority of programming models for heterogeneous computing two completely distinct implementations of a computational focus on only one of these. With the development of Accelerated kernel is a difficult and error-prone proposition. That leaves OpenMP for GPUs, both from PGI and Cray, we have a clear us with using either OpenCL or accelerated OpenMP to path to extend traditional OpenMP applications incrementally complete the task. to use GPUs. The extensions are geared toward switching from CPU parallelism to GPU parallelism. However they OpenCL’s greatest strength lies in its broad hardware do not preserve the former while adding the latter. Thus support. In a way, though, that is also its greatest weak- computational potential is wasted since either the CPU cores ness. To enable one to program this disparate hardware or the GPU cores are left idle.
    [Show full text]
  • Parallel Programming
    Parallel Programming Libraries and implementations Outline • MPI – distributed memory de-facto standard • Using MPI • OpenMP – shared memory de-facto standard • Using OpenMP • CUDA – GPGPU de-facto standard • Using CUDA • Others • Hybrid programming • Xeon Phi Programming • SHMEM • PGAS MPI Library Distributed, message-passing programming Message-passing concepts Explicit Parallelism • In message-passing all the parallelism is explicit • The program includes specific instructions for each communication • What to send or receive • When to send or receive • Synchronisation • It is up to the developer to design the parallel decomposition and implement it • How will you divide up the problem? • When will you need to communicate between processes? Message Passing Interface (MPI) • MPI is a portable library used for writing parallel programs using the message passing model • You can expect MPI to be available on any HPC platform you use • Based on a number of processes running independently in parallel • HPC resource provides a command to launch multiple processes simultaneously (e.g. mpiexec, aprun) • There are a number of different implementations but all should support the MPI 2 standard • As with different compilers, there will be variations between implementations but all the features specified in the standard should work. • Examples: MPICH2, OpenMPI Point-to-point communications • A message sent by one process and received by another • Both processes are actively involved in the communication – not necessarily at the same time • Wide variety of semantics provided: • Blocking vs. non-blocking • Ready vs. synchronous vs. buffered • Tags, communicators, wild-cards • Built-in and custom data-types • Can be used to implement any communication pattern • Collective operations, if applicable, can be more efficient Collective communications • A communication that involves all processes • “all” within a communicator, i.e.
    [Show full text]
  • Openmp API 5.1 Specification
    OpenMP Application Programming Interface Version 5.1 November 2020 Copyright c 1997-2020 OpenMP Architecture Review Board. Permission to copy without fee all or part of this material is granted, provided the OpenMP Architecture Review Board copyright notice and the title of this document appear. Notice is given that copying is by permission of the OpenMP Architecture Review Board. This page intentionally left blank in published version. Contents 1 Overview of the OpenMP API1 1.1 Scope . .1 1.2 Glossary . .2 1.2.1 Threading Concepts . .2 1.2.2 OpenMP Language Terminology . .2 1.2.3 Loop Terminology . .9 1.2.4 Synchronization Terminology . 10 1.2.5 Tasking Terminology . 12 1.2.6 Data Terminology . 14 1.2.7 Implementation Terminology . 18 1.2.8 Tool Terminology . 19 1.3 Execution Model . 22 1.4 Memory Model . 25 1.4.1 Structure of the OpenMP Memory Model . 25 1.4.2 Device Data Environments . 26 1.4.3 Memory Management . 27 1.4.4 The Flush Operation . 27 1.4.5 Flush Synchronization and Happens Before .................. 29 1.4.6 OpenMP Memory Consistency . 30 1.5 Tool Interfaces . 31 1.5.1 OMPT . 32 1.5.2 OMPD . 32 1.6 OpenMP Compliance . 33 1.7 Normative References . 33 1.8 Organization of this Document . 35 i 2 Directives 37 2.1 Directive Format . 38 2.1.1 Fixed Source Form Directives . 43 2.1.2 Free Source Form Directives . 44 2.1.3 Stand-Alone Directives . 45 2.1.4 Array Shaping . 45 2.1.5 Array Sections .
    [Show full text]
  • Openmp Made Easy with INTEL® ADVISOR
    OpenMP made easy with INTEL® ADVISOR Zakhar Matveev, PhD, Intel CVCG, November 2018, SC’18 OpenMP booth Why do we care? Ai Bi Ai Bi Ai Bi Ai Bi Vector + Processing Ci Ci Ci Ci VL Motivation (instead of Agenda) • Starting from 4.x, OpenMP introduces support for both levels of parallelism: • Multi-Core (think of “pragma/directive omp parallel for”) • SIMD (think of “pragma/directive omp simd”) • 2 pillars of OpenMP SIMD programming model • Hardware with Intel ® AVX-512 support gives you theoretically 8x speed- up over SSE baseline (less or even more in practice) • Intel® Advisor is here to assist you in: • Enabling SIMD parallelism with OpenMP (if not yet) • Improving performance of already vectorized OpenMP SIMD code • And will also help to optimize for Memory Sub-system (Advisor Roofline) 3 Don’t use a single Vector lane! Un-vectorized and un-threaded software will under perform 4 Permission to Design for All Lanes Threading and Vectorization needed to fully utilize modern hardware 5 A Vector parallelism in x86 AVX-512VL AVX-512BW B AVX-512DQ AVX-512CD AVX-512F C AVX2 AVX2 AVX AVX AVX SSE SSE SSE SSE Intel® microarchitecture code name … NHM SNB HSW SKL 6 (theoretically) 8x more SIMD FLOP/S compared to your (–O2) optimized baseline x - Significant leap to 512-bit SIMD support for processors - Intel® Compilers and Intel® Math Kernel Library include AVX-512 support x - Strong compatibility with AVX - Added EVEX prefix enables additional x functionality Don’t leave it on the table! 7 Two level parallelism decomposition with OpenMP: image processing example B #pragma omp parallel for for (int y = 0; y < ImageHeight; ++y){ #pragma omp simd C for (int x = 0; x < ImageWidth; ++x){ count[y][x] = mandel(in_vals[y][x]); } } 8 Two level parallelism decomposition with OpenMP: fluid dynamics processing example B #pragma omp parallel for for (int i = 0; i < X_Dim; ++i){ #pragma omp simd C for (int m = 0; x < n_velocities; ++m){ next_i = f(i, velocities(m)); X[i] = next_i; } } 9 Key components of Intel® Advisor What’s new in “2019” release Step 1.
    [Show full text]
  • Introduction to Openmp Paul Edmon ITC Research Computing Associate
    Introduction to OpenMP Paul Edmon ITC Research Computing Associate FAS Research Computing Overview • Threaded Parallelism • OpenMP Basics • OpenMP Programming • Benchmarking FAS Research Computing Threaded Parallelism • Shared Memory • Single Node • Non-uniform Memory Access (NUMA) • One thread per core FAS Research Computing Threaded Languages • PThreads • Python • Perl • OpenCL/CUDA • OpenACC • OpenMP FAS Research Computing OpenMP Basics FAS Research Computing What is OpenMP? • OpenMP (Open Multi-Processing) – Application Program Interface (API) – Governed by OpenMP Architecture Review Board • OpenMP provides a portable, scalable model for developers of shared memory parallel applications • The API supports C/C++ and Fortran on a wide variety of architectures FAS Research Computing Goals of OpenMP • Standardization – Provide a standard among a variety shared memory architectures / platforms – Jointly defined and endorsed by a group of major computer hardware and software vendors • Lean and Mean – Establish a simple and limited set of directives for programming shared memory machines – Significant parallelism can be implemented by just a few directives • Ease of Use – Provide the capability to incrementally parallelize a serial program • Portability – Specified for C/C++ and Fortran – Most majors platforms have been implemented including Unix/Linux and Windows – Implemented for all major compilers FAS Research Computing OpenMP Programming Model Shared Memory Model: OpenMP is designed for multi-processor/core, shared memory machines Thread Based Parallelism: OpenMP programs accomplish parallelism exclusively through the use of threads Explicit Parallelism: OpenMP provides explicit (not automatic) parallelism, offering the programmer full control over parallelization Compiler Directive Based: Parallelism is specified through the use of compiler directives embedded in the C/C++ or Fortran code I/O: OpenMP specifies nothing about parallel I/O.
    [Show full text]
  • Parallel Programming with Openmp
    Parallel Programming with OpenMP OpenMP Parallel Programming Introduction: OpenMP Programming Model Thread-based parallelism utilized on shared-memory platforms Parallelization is either explicit, where programmer has full control over parallelization or through using compiler directives, existing in the source code. Thread is a process of a code is being executed. A thread of execution is the smallest unit of processing. Multiple threads can exist within the same process and share resources such as memory OpenMP Parallel Programming Introduction: OpenMP Programming Model Master thread is a single thread that runs sequentially; parallel execution occurs inside parallel regions and between two parallel regions, only the master thread executes the code. This is called the fork-join model: OpenMP Parallel Programming OpenMP Parallel Computing Hardware Shared memory allows immediate access to all data from all processors without explicit communication. Shared memory: multiple cpus are attached to the BUS all processors share the same primary memory the same memory address on different CPU's refer to the same memory location CPU-to-memory connection becomes a bottleneck: shared memory computers cannot scale very well OpenMP Parallel Programming OpenMP versus MPI OpenMP (Open Multi-Processing): easy to use; loop-level parallelism non-loop-level parallelism is more difficult limited to shared memory computers cannot handle very large problems An alternative is MPI (Message Passing Interface): require low-level programming; more difficult programming
    [Show full text]
  • Intel Threading Building Blocks
    Praise for Intel Threading Building Blocks “The Age of Serial Computing is over. With the advent of multi-core processors, parallel- computing technology that was once relegated to universities and research labs is now emerging as mainstream. Intel Threading Building Blocks updates and greatly expands the ‘work-stealing’ technology pioneered by the MIT Cilk system of 15 years ago, providing a modern industrial-strength C++ library for concurrent programming. “Not only does this book offer an excellent introduction to the library, it furnishes novices and experts alike with a clear and accessible discussion of the complexities of concurrency.” — Charles E. Leiserson, MIT Computer Science and Artificial Intelligence Laboratory “We used to say make it right, then make it fast. We can’t do that anymore. TBB lets us design for correctness and speed up front for Maya. This book shows you how to extract the most benefit from using TBB in your code.” — Martin Watt, Senior Software Engineer, Autodesk “TBB promises to change how parallel programming is done in C++. This book will be extremely useful to any C++ programmer. With this book, James achieves two important goals: • Presents an excellent introduction to parallel programming, illustrating the most com- mon parallel programming patterns and the forces governing their use. • Documents the Threading Building Blocks C++ library—a library that provides generic algorithms for these patterns. “TBB incorporates many of the best ideas that researchers in object-oriented parallel computing developed in the last two decades.” — Marc Snir, Head of the Computer Science Department, University of Illinois at Urbana-Champaign “This book was my first introduction to Intel Threading Building Blocks.
    [Show full text]
  • Unified Parallel C for GPU Clusters: Language Extensions and Compiler Implementation
    Unified Parallel C for GPU Clusters: Language Extensions and Compiler Implementation Li Chen1, Lei Liu1, Shenglin Tang1, Lei Huang2, Zheng Jing1, Shixiong Xu1, Dingfei Zhang1, Baojiang Shou1 1 Key Laboratory of Computer System and Architecture, Institute of Computing Technology, Chinese Academy of Sciences, China {lchen,liulei2007,tangshenglin,jingzheng,xushixiong, zhangdingfei, shoubaojiang}@ict.ac.cn; 2 Department of Computer Science, University of Houston; Houston, TX, USA [email protected] 5 Abstract. Unified Parallel C (UPC), a parallel extension to ANSI C, is designed for high performance computing on large-scale parallel machines. With General-purpose graphics processing units (GPUs) becoming an increasingly important high performance computing platform, we propose new language extensions to UPC to take advantage of GPU clusters. We extend UPC with hierarchical data distribution, revise the execution model of UPC to mix SPMD with fork-join execution model, and modify the semantics of upc_forall to reflect the data-thread affinity on a thread hierarchy. We implement the compiling system, including affinity-aware loop tiling, GPU code generation, and several memory optimizations targeting NVIDIA CUDA. We also put forward unified data management for each UPC thread to optimize data transfer and memory layout for separate memory modules of CPUs and GPUs. The experimental results show that the UPC extension has better programmability than the mixed MPI/CUDA approach. We also demonstrate that the integrated compile-time and runtime optimization is effective to achieve good performance on GPU clusters. 1 Introduction Following closely behind the industry-wide move from uniprocessor to multi-core and many-core systems, HPC computing platforms are undergoing another major change: from homogeneous to heterogeneous platform.
    [Show full text]
  • Lecture 12: Introduction to Openmp (Part 1)
    Lecture 12: Introduction to OpenMP (Part 1) 1 What is OpenMP Open specifications for Multi Processing Long version: Open specifications for MultiProcessing via collaborative work between interested parties from the hardware and software industry, government and academia. • An Application Program Interface (API) that is used to explicitly direct multi-threaded, shared memory parallelism. • API components: – Compiler directives – Runtime library routines – Environment variables • Portability – API is specified for C/C++ and Fortran – Implementations on almost all platforms including Unix/Linux and Windows • Standardization – Jointly defined and endorsed by major computer hardware and software vendors – Possibility to become ANSI standard 2 Brief History of OpenMP • In 1991, Parallel Computing Forum (PCF) group invented a set of directives for specifying loop parallelism in Fortran programs. • X3H5, an ANSI subcommittee developed an ANSI standard based on PCF. • In 1997, the first version of OpenMP for Fortran was defined by OpenMP Architecture Review Board. • Binding for C/C++ was introduced later. • Version 3.1 is available since 2011. 3 4 Thread • A process is an instance of a computer program that is being executed. It contains the program code and its current activity. • A thread of execution is the smallest unit of processing that can be scheduled by an operating system. • Differences between threads and processes: – A thread is contained inside a process. Multiple threads can exist within the same process and share resources such as memory. The threads of a process share the latter’s instructions (code) and its context (values that its variables reference at any given moment). – Different processes do not share these resources.
    [Show full text]
  • Programming Your GPU with Openmp*
    Programming Your GPU with OpenMP* Simon McIntosh-Smith, Matt Martineau, Andrei Poenaru, Patrick Atkinson University of Bristol [email protected] 1 * The name “OpenMP” is the property of the OpenMP Architecture Review Board. 1 Preliminaries: Part 2 • Our plan for the day .. Active learning! – We will mix short lectures with short exercises. – You will use your laptop to connect to a multiprocessor server. • Please follow these simple rules – Do the exercises that we assign and then change things around and experiment. – EmBrace active learning! –Don’t cheat: Do Not look at the solutions before you complete an exercise … even if you get really frustrated. 2 • Collaboration between GW4 Alliance (universities of Bristol, Bath, Cardiff, Exeter), the UK Met Office, Cray, and Arm • ~£3 million, funded by EPSRC • Expected to be the first large-scale, Arm-based production supercomputer • 10,000+ Armv8 cores • Also contains some Intel Xeon Phi (KNL), Intel Xeon (Broadwell), and NVIDIA P100 GPUs • Cray compilers provide the OpenMP implementation for GPU offloading 3 Plan Module Concepts Exercises OpenMP overview • OpenMP recap • None … we’ll use demo • Hardcore jargon of OpenMP mode so we can move fast The device model in • Intro to the Target directive • Vadd program OpenMP with default data movement Understanding • Intro to nvprof • nvprof ./vadd execution Working with the • CPU and GPU execution • Vadd with combined target directive models and the fundamental directive using nvprof to combined directive understand execution Basic memory •
    [Show full text]
  • A “Hands-On” Introduction to Openmp*
    A “Hands-on” Introduction to OpenMP* Tim Mattson Larry Meadows Principal Engineer Principal Engineer Intel Corporation Intel Corporation [email protected] [email protected] 1 * The name “OpenMP” is the property of the OpenMP Architecture Review Board. Preliminaries: part 1 z Disclosures The views expressed in this tutorial are those of the people delivering the tutorial. – We are not speaking for our employers. – We are not speaking for the OpenMP ARB z This is a new tutorial for us: Help us improve … tell us how you would make this tutorial better. 2 Preliminaries: Part 2 z Our plan for the day .. Active learning! We will mix short lectures with short exercises. You will use your laptop for the exercises … that way you’ll have an OpenMP environment to take home so you can keep learning on your own. z Please follow these simple rules Do the exercises we assign and then change things around and experiment. – Embrace active learning! Don’t cheat: Do Not look at the solutions before you complete an exercise … even if you get really frustrated. 3 Our Plan for the day Topic Exercise concepts I. OMP Intro Install sw, Parallel regions hello_world II. Creating threads Pi_spmd_simple Parallel, default data environment, runtime library calls Break III. Synchronization Pi_spmd_final False sharing, critical, atomic IV. Parallel loops Pi_loop For, reduction V. Odds and ends No exercise Single, master, runtime libraries, environment variables, synchronization, etc. lunch VI. Data Environment Pi_mc Data environment details, modular software, threadprivate VII. Worksharing and Linked list, For, schedules, sections schedule matmul Break VIII.
    [Show full text]
  • Openmp Tutorial
    OpenMP Tutorial Seung-Jai Min ([email protected]) School of Electrical and Computer Engineering Purdue University, West Lafayette, IN ECE 563 Programming Parallel Machines 1 Parallel Programming Standards • Thread Libraries - Win32 API / Posix threads • Compiler Directives OUR FOCUS - OpenMP (Shared memory programming) • Message Passing Libraries - MPI (Distributed memory programming) ECE 563 Programming Parallel Machines 2 Shared Memory Parallel Programming in the Multi-Core Era • Desktop and Laptop – 2, 4, 8 cores and … ? • A single node in distributed memory clusters – Steele cluster node: 2 8 (16) cores • Shared memory hardware Accelerators • Cell processors: 1 PPE and 8 SPEs • Nvidia Quadro GPUs: 128 processing units ECE 563 Programming Parallel Machines 3 OpenMP: Some syntax details to get us started • Most of the constructs in OpenMP are compiler directives or pragmas. – For C and C++, the pragmas take the form: #pragma omp construct [clause [clause]…] – For Fortran, the directives take one of the forms: C$OMP construct [clause [clause]…] !$OMP construct [clause [clause]…] *$OMP construct [clause [clause]…] • Include files #include “omp.h” ECE 563 Programming Parallel Machines 4 How is OpenMP typically used? • OpenMP is usually used to parallelize loops: • Find your most time consuming loops. • Split them up between threads. Parallel Program Sequential Program #include “omp.h” void main() void main() { { int i, k, N=1000; int i, k, N=1000; double A[N], B[N], C[N]; double A[N], B[N], C[N]; for (i=0; i<N; i++) { #pragma omp parallel
    [Show full text]