OpenMP

Table of Contents

1. Abstract
2. Introduction
   1. What is OpenMP?
   2. History
   3. Goals of OpenMP
3. OpenMP Programming Model
4. OpenMP Directives
   1. Fortran Directive Format
   2. C/C++ Directive Format
   3. Directive Scoping
   4. PARALLEL Construct
   5. Work-Sharing Constructs
      1. DO / for Directive
      2. SECTIONS Directive
      3. WORKSHARE Directive
      4. SINGLE Directive
   6. Combined Parallel Work-Sharing Constructs
      1. PARALLEL DO / parallel for Directive
      2. PARALLEL SECTIONS Directive
      3. PARALLEL WORKSHARE Directive
   7. TASK Construct
   8. Synchronization Constructs
      1. MASTER Directive
      2. CRITICAL Directive
      3. BARRIER Directive
      4. TASKWAIT Directive
      5. ATOMIC Directive
      6. FLUSH Directive
      7. ORDERED Directive
   9. THREADPRIVATE Directive
   10. Data Scope Attribute Clauses
       1. PRIVATE Clause
       2. SHARED Clause
       3. DEFAULT Clause
       4. FIRSTPRIVATE Clause
       5. LASTPRIVATE Clause
       6. COPYIN Clause
       7. COPYPRIVATE Clause
       8. REDUCTION Clause
   11. Clauses / Directives Summary
   12. Directive Binding and Nesting Rules
5. Run-Time Library Routines
   1. OMP_SET_NUM_THREADS
   2. OMP_GET_NUM_THREADS
   3. OMP_GET_MAX_THREADS
   4. OMP_GET_THREAD_NUM
   5. OMP_GET_THREAD_LIMIT
   6. OMP_GET_NUM_PROCS
   7. OMP_IN_PARALLEL
   8. OMP_SET_DYNAMIC
   9. OMP_GET_DYNAMIC
   10. OMP_SET_NESTED
   11. OMP_GET_NESTED
   12. OMP_SET_SCHEDULE
   13. OMP_GET_SCHEDULE
   14. OMP_SET_MAX_ACTIVE_LEVELS
   15. OMP_GET_MAX_ACTIVE_LEVELS
   16. OMP_GET_LEVEL
   17. OMP_GET_ANCESTOR_THREAD_NUM
   18. OMP_GET_TEAM_SIZE
   19. OMP_GET_ACTIVE_LEVEL
   20. OMP_INIT_LOCK
   21. OMP_DESTROY_LOCK
   22. OMP_SET_LOCK
   23. OMP_UNSET_LOCK
   24. OMP_TEST_LOCK
   25. OMP_INIT_NEST_LOCK
   26. OMP_DESTROY_NEST_LOCK
   27. OMP_SET_NEST_LOCK
   28. OMP_UNSET_NEST_LOCK
   29. OMP_TEST_NEST_LOCK
   30. OMP_GET_WTIME
   31. OMP_GET_WTICK
6. Environment Variables
7. Memory and Performance Issues
8. LLNL Specific Information and Recommendations
9. References and More Information
10. Exercise

OpenMP is an Application Program Interface (API), jointly defined by a group of major computer hardware and software vendors. OpenMP provides a portable, scalable model for developers of shared memory parallel applications. The API supports C/C++ and Fortran on multiple architectures, including UNIX & Windows NT. This tutorial covers most of the major features of OpenMP, including its various constructs and directives for specifying parallel regions, work sharing, synchronization and data environment. Runtime library functions and environment variables are also covered. This tutorial includes both C and Fortran example codes and a lab exercise.

Level/Prerequisites: Geared to those who are new to parallel programming with OpenMP. Basic understanding of parallel programming in C or Fortran assumed. For those who are unfamiliar with Parallel Programming in general, the material covered in EC3500: Introduction to Parallel Computing would be helpful.

Introduction

What is OpenMP?

• An Application Program Interface (API) that may be used to explicitly direct multi-threaded, shared memory parallelism

• Comprised of three primary API components:
  • Compiler Directives
  • Runtime Library Routines
  • Environment Variables

• Portable:
  • The API is specified for C/C++ and Fortran
  • Most major platforms have been implemented including Unix/Linux platforms and Windows NT

• Standardized:
  • Jointly defined and endorsed by a group of major computer hardware and software vendors
  • Expected to become an ANSI standard later???

• What does OpenMP stand for?
  • Short version: Open Multi-Processing
  • Long version: Open specifications for Multi-Processing via collaborative work between interested parties from the hardware and software industry, government and academia.

OpenMP Is Not:

• Meant for distributed memory parallel systems (by itself)

• Necessarily implemented identically by all vendors

• Guaranteed to make the most efficient use of shared memory

• Required to check for data dependencies, data conflicts, race conditions, or deadlocks

• Required to check for code sequences that cause a program to be classified as non-conforming

• Meant to cover compiler-generated automatic parallelization and directives to the compiler to assist such parallelization

• Designed to guarantee that input or output to the same file is synchronous when executed in parallel. The programmer is responsible for synchronizing input and output.

References:

• OpenMP website: openmp.org
  API specifications, FAQ, presentations, discussions, media releases, calendar, membership application and more...

• Wikipedia: en.wikipedia.org/wiki/OpenMP

History

• In the early 90's, vendors of shared-memory machines supplied similar, directive-based, Fortran programming extensions:
  • The user would augment a serial Fortran program with directives specifying which loops were to be parallelized
  • The compiler would be responsible for automatically parallelizing such loops across the SMP processors

• Implementations were all functionally similar, but were diverging (as usual)

• First attempt at a standard was the draft for ANSI X3H5 in 1994. It was never adopted, largely due to waning interest as distributed memory machines became popular.

• The OpenMP standard specification started in the spring of 1997, taking over where ANSI X3H5 had left off, as newer shared memory machine architectures started to become prevalent.

• Led by the OpenMP Architecture Review Board (ARB). Original ARB members included: (Disclaimer: all partner names derived from the OpenMP web site)
  • Compaq / Digital
  • Hewlett-Packard Company
  • Intel Corporation
  • International Business Machines (IBM)
  • Kuck & Associates, Inc. (KAI)
  • Silicon Graphics, Inc.
  • Sun Microsystems, Inc.
  • U.S. Department of Energy ASCI program

• Other contributors included:

  Endorsing application developers: ADINA R&D, Inc. ANSYS, Inc. Dash Associates Fluent, Inc. ILOG CPLEX Division Livermore Software Technology Corporation (LSTC) MECALOG SARL Oxford Molecular Group PLC The Numerical Algorithms Group Ltd.(NAG)

  Endorsing software vendors: Absoft Corporation Edinburgh Portable Compilers GENIAS Software GmBH Myrias Computer Technologies, Inc. The Portland Group, Inc. (PGI)

• For more news and membership information about the OpenMP ARB, visit: openmp.org/wp/about-openmp.

Release History

Introduction

Goals of OpenMP

• Provide a standard among a variety of shared memory architectures/platforms

• Establish a simple and limited set of directives for programming shared memory machines. Significant parallelism can be implemented by using just 3 or 4 directives.

Ease of Use:

• Provide capability to incrementally parallelize a serial program, unlike message-passing libraries which typically require an all or nothing approach

• Provide the capability to implement both coarse-grain and fine-grain parallelism

Portability:

• Supports Fortran (77, 90, and 95), C, and C++

• Public forum for API and membership

Shared Memory, Thread Based Parallelism:

• OpenMP is based upon the existence of multiple threads in the shared memory programming paradigm. A shared memory process consists of multiple threads.

Explicit Parallelism:

• OpenMP is an explicit (not automatic) programming model, offering the programmer full control over parallelization.

Fork - Join Model:

• OpenMP uses the fork-join model of parallel execution:

  

• All OpenMP programs begin as a single process: the master thread. The master thread executes sequentially until the first parallel region construct is encountered.

• FORK: the master thread then creates a team of parallel threads

• The statements in the program that are enclosed by the parallel region construct are then executed in parallel among the various team threads

• JOIN: When the team threads complete the statements in the parallel region construct, they synchronize and terminate, leaving only the master thread

Compiler Directive Based:

• Most OpenMP parallelism is specified through the use of compiler directives which are imbedded in C/C++ or Fortran source code.

Nested Parallelism Support:

• The API provides for the placement of parallel constructs inside of other parallel constructs.

• Implementations may or may not support this feature.

Dynamic Threads:

• The API provides for dynamically altering the number of threads which may used to execute different parallel regions.

• Implementations may or may not support this feature.

I/O:

• OpenMP specifies nothing about parallel I/O. This is particularly important if multiple threads attempt to write/read from the same file.

• If every thread conducts I/O to a different file, the issues are not as significant.

• It is entirely up to the programmer to insure that I/O is conducted correctly within the context of a multi-threaded program.

Memory Model: FLUSH Often?

• OpenMP provides a "relaxed-consistency" and "temporary" view of thread memory (in their words). In other words, threads can "cache" their data and are not required to maintain exact consistency with real memory all of the time.

• When it is critical that all threads view a shared variable identically, the programmer is responsible for insuring that the variable is FLUSHed by all threads as needed.

• More on this later...

Example OpenMP Code Structure

Fortran - General Code Structure PROGRAM HELLO INTEGER VAR1, VAR2, VAR3 Serial code . . . Beginning of parallel section. Fork a team of threads. Specify variable scoping !$OMP PARALLEL PRIVATE(VAR1, VAR2) SHARED(VAR3) Parallel section executed by all threads . . . All threads join master thread and disband !$OMP END PARALLEL Resume serial code . . . END

C / C++ - General Code Structure #include <omp.h> main () { int var1, var2, var3; Serial code . . . Beginning of parallel section. Fork a team of threads. Specify variable scoping #pragma omp parallel private(var1, var2) shared(var3) { Parallel section executed by all threads . . . All threads join master thread and disband } Resume serial code . . . }

Fortran Directives Format

sentinel	directive-name	[clause ...]
All Fortran OpenMP directives must begin with a sentinel. The accepted sentinels depend upon the type of Fortran source. Possible sentinels are: !$OMP C$OMP *$OMP	A valid OpenMP directive. Must appear after the sentinel and before any clauses.	Optional. Clauses can be in any order, and repeated as necessary unless otherwise restricted.

Example:

!$OMP PARALLEL DEFAULT(SHARED) PRIVATE(BETA,PI)

Fixed Form Source:

• !$OMP C$OMP *$OMP are accepted sentinels and must start in column 1

• All Fortran fixed form rules for line length, white space, continuation and comment columns apply for the entire directive line

• Initial directive lines must have a space/zero in column 6.

• Continuation lines must have a non-space/zero in column 6.

Free Form Source:

• !$OMP is the only accepted sentinel. Can appear in any column, but must be preceded by white space only.

• All Fortran free form rules for line length, white space, continuation and comment columns apply for the entire directive line

• Initial directive lines must have a space after the sentinel.

• Continuation lines must have an ampersand as the last non-blank character in a line. The following line must begin with a sentinel and then the continuation directives.

• Comments can not appear on the same line as a directive

• Only one directive-name may be specified per directive

• Fortran compilers which are OpenMP enabled generally include a command line option which instructs the compiler to activate and interpret all OpenMP directives.

• Several Fortran OpenMP directives come in pairs and have the form shown below. The "end" directive is optional but advised for readability.

  !$OMP directive [ structured block of code ] !$OMP end directive

C / C++ Directives Format

#pragma omp	directive-name	[clause, ...]	newline
Required for all OpenMP C/C++ directives.	A valid OpenMP directive. Must appear after the pragma and before any clauses.	Optional. Clauses can be in any order, and repeated as necessary unless otherwise restricted.	Required. Precedes the structured block which is enclosed by this directive.

Example:

#pragma omp parallel default(shared) private(beta,pi)

General Rules:

• Case sensitive

• Directives follow conventions of the C/C++ standards for compiler directives

• Only one directive-name may be specified per directive

• Each directive applies to at most one succeeding statement, which must be a structured block.

• Long directive lines can be "continued" on succeeding lines by escaping the newline character with a backslash ("\") at the end of a directive line.

Directive Scoping

Static (Lexical) Extent:

• The code textually enclosed between the beginning and the end of a structured block following a directive.

• The static extent of a directives does not span multiple routines or code files

• An OpenMP directive that appears independently from another enclosing directive is said to be an orphaned directive. It exists outside of another directive's static (lexical) extent.

• Will span routines and possibly code files

Dynamic Extent:

• The dynamic extent of a directive includes both its static (lexical) extent and the extents of its orphaned directives.

Example:

PROGRAM TEST ... !$OMP PARALLEL ... !$OMP DO DO I=... ... CALL SUB1 ... ENDDO ... CALL SUB2 ... !$OMP END PARALLEL	SUBROUTINE SUB1 ... !$OMP CRITICAL ... !$OMP END CRITICAL END SUBROUTINE SUB2 ... !$OMP SECTIONS ... !$OMP END SECTIONS ... END
STATIC EXTENT The DO directive occurs within an enclosing parallel region	ORPHANED DIRECTIVES The CRITICAL and SECTIONS directives occur outside an enclosing parallel region
DYNAMIC EXTENT The CRITICAL and SECTIONS directives occur within the dynamic extent of the DO and PARALLEL directives.

Why Is This Important?

• OpenMP specifies a number of scoping rules on how directives may associate (bind) and nest within each other

• Illegal and/or incorrect programs may result if the OpenMP binding and nesting rules are ignored

• See Directive Binding and Nesting Rules for specific details

PARALLEL Region Construct

• A parallel region is a block of code that will be executed by multiple threads. This is the fundamental OpenMP parallel construct.

Format:

Fortran	!$OMP PARALLEL [clause ...] IF (scalar_logical_expression) PRIVATE (list) SHARED (list) DEFAULT (PRIVATE | FIRSTPRIVATE | SHARED | NONE) FIRSTPRIVATE (list) REDUCTION (operator: list) COPYIN (list) NUM_THREADS (scalar-integer-expression) block !$OMP END PARALLEL
C/C++	#pragma omp parallel [clause ...] newline if (scalar_expression) private (list) shared (list) default (shared | none) firstprivate (list) reduction (operator: list) copyin (list) num_threads (integer-expression) structured_block

Notes:

• When a thread reaches a PARALLEL directive, it creates a team of threads and becomes the master of the team. The master is a member of that team and has thread number 0 within that team.

• Starting from the beginning of this parallel region, the code is duplicated and all threads will execute that code.

• There is an implied barrier at the end of a parallel section. Only the master thread continues execution past this point.

• If any thread terminates within a parallel region, all threads in the team will terminate, and the work done up until that point is undefined.

How Many Threads?

• The number of threads in a parallel region is determined by the following factors, in order of precedence:

  1. Evaluation of the IF clause

  2. Setting of the NUM_THREADS clause

  3. Use of the omp_set_num_threads() library function

  4. Setting of the OMP_NUM_THREADS environment variable

  5. Implementation default - usually the number of CPUs on a node, though it could be dynamic (see next bullet).

• Threads are numbered from 0 (master thread) to N-1

Dynamic Threads:

• Use the omp_get_dynamic() library function to determine if dynamic threads are enabled.

• If supported, the two methods available for enabling dynamic threads are:

  1. The omp_set_dynamic() library routine

  2. Setting of the OMP_DYNAMIC environment variable to TRUE

Nested Parallel Regions:

• Use the omp_get_nested() library function to determine if nested parallel regions are enabled.

• The two methods available for enabling nested parallel regions (if supported) are:

  1. The omp_set_nested() library routine

  2. Setting of the OMP_NESTED environment variable to TRUE

• If not supported, a parallel region nested within another parallel region results in the creation of a new team, consisting of one thread, by default.

Clauses:

• IF clause: If present, it must evaluate to .TRUE. (Fortran) or non-zero (C/C++) in order for a team of threads to be created. Otherwise, the region is executed serially by the master thread.

• The remaining clauses are described in detail later, in the Data Scope Attribute Clauses section.

Restrictions:

• A parallel region must be a structured block that does not span multiple routines or code files

• It is illegal to branch into or out of a parallel region

• Only a single IF clause is permitted

• Only a single NUM_THREADS clause is permitted

Example: Parallel Region

• Simple "Hello World" program
  • Every thread executes all code enclosed in the parallel section
  • OpenMP library routines are used to obtain thread identifiers and total number of threads

  Fortran - Parallel Region Example PROGRAM HELLO INTEGER NTHREADS, TID, OMP_GET_NUM_THREADS, + OMP_GET_THREAD_NUM C Fork a team of threads with each thread having a private TID variable !$OMP PARALLEL PRIVATE(TID) C Obtain and print thread id TID = OMP_GET_THREAD_NUM() PRINT *, 'Hello World from thread = ', TID C Only master thread does this IF (TID .EQ. 0) THEN NTHREADS = OMP_GET_NUM_THREADS() PRINT *, 'Number of threads = ', NTHREADS END IF C All threads join master thread and disband !$OMP END PARALLEL END

  C / C++ - Parallel Region Example #include <omp.h> main () { int nthreads, tid; /* Fork a team of threads with each thread having a private tid variable */ #pragma omp parallel private(tid) { /* Obtain and print thread id */ tid = omp_get_thread_num(); printf("Hello World from thread = %d\n", tid); /* Only master thread does this */ if (tid == 0) { nthreads = omp_get_num_threads(); printf("Number of threads = %d\n", nthreads); } } /* All threads join master thread and terminate */ }

Work-Sharing Constructs

• A work-sharing construct divides the execution of the enclosed code region among the members of the team that encounter it.

• Work-sharing constructs do not launch new threads

• There is no implied barrier upon entry to a work-sharing construct, however there is an implied barrier at the end of a work sharing construct.

Types of Work-Sharing Constructs:

NOTE: The Fortran workshare construct is not shown here, but is discussed later.

DO / for - shares iterations of a loop across the team. Represents a type of "data parallelism".	SECTIONS - breaks work into separate, discrete sections. Each section is executed by a thread. Can be used to implement a type of "functional parallelism".	SINGLE - serializes a section of code

Restrictions:

• A work-sharing construct must be enclosed dynamically within a parallel region in order for the directive to execute in parallel.

• Work-sharing constructs must be encountered by all members of a team or none at all

• Successive work-sharing constructs must be encountered in the same order by all members of a team

Work-Sharing Constructs
DO / for Directive

• The DO / for directive specifies that the iterations of the loop immediately following it must be executed in parallel by the team. This assumes a parallel region has already been initiated, otherwise it executes in serial on a single processor.

Format:

Fortran	!$OMP DO [clause ...] SCHEDULE (type [,chunk]) ORDERED PRIVATE (list) FIRSTPRIVATE (list) LASTPRIVATE (list) SHARED (list) REDUCTION (operator | intrinsic : list) COLLAPSE (n) do_loop !$OMP END DO [ NOWAIT ]
C/C++	#pragma omp for [clause ...] newline schedule (type [,chunk]) ordered private (list) firstprivate (list) lastprivate (list) shared (list) reduction (operator: list) collapse (n) nowait for_loop

Clauses:

• SCHEDULE: Describes how iterations of the loop are divided among the threads in the team. The default schedule is implementation dependent.

  STATIC

  Loop iterations are divided into pieces of size chunk and then statically assigned to threads. If chunk is not specified, the iterations are evenly (if possible) divided contiguously among the threads.

  DYNAMIC

  Loop iterations are divided into pieces of size chunk, and dynamically scheduled among the threads; when a thread finishes one chunk, it is dynamically assigned another. The default chunk size is 1.

  GUIDED

  For a chunk size of 1, the size of each chunk is proportional to the number of unassigned iterations divided by the number of threads, decreasing to 1. For a chunk size with value k (greater than 1), the size of each chunk is determined in the same way with the restriction that the chunks do not contain fewer than k iterations (except for the last chunk to be assigned, which may have fewer than k iterations). The default chunk size is 1.

  RUNTIME

  The scheduling decision is deferred until runtime by the environment variable OMP_SCHEDULE. It is illegal to specify a chunk size for this clause.

  AUTO

  The scheduling decision is delegated to the compiler and/or runtime system.

• NO WAIT / nowait: If specified, then threads do not synchronize at the end of the parallel loop.

• ORDERED: Specifies that the iterations of the loop must be executed as they would be in a serial program.

• COLLAPSE: Specifies how many loops in a nested loop should be collapsed into one large iteration space and divided according to the schedule clause. The sequential execution of the iterations in all associated loops determines the order of the iterations in the collapsed iteration space.

• Other clauses are described in detail later, in the Data Scope Attribute Clauses section.

Restrictions:

• The DO loop can not be a DO WHILE loop, or a loop without loop control. Also, the loop iteration variable must be an integer and the loop control parameters must be the same for all threads.

• Program correctness must not depend upon which thread executes a particular iteration.

• It is illegal to branch out of a loop associated with a DO/for directive.

• The chunk size must be specified as a loop invarient integer expression, as there is no synchronization during its evaluation by different threads.

• ORDERED, COLLAPSE and SCHEDULE clauses may appear once each.

• See the OpenMP specification document for additional restrictions.

Example: DO / for Directive

• Simple vector-add program
  • Arrays A, B, C, and variable N will be shared by all threads.
  • Variable I will be private to each thread; each thread will have its own unique copy.
  • The iterations of the loop will be distributed dynamically in CHUNK sized pieces.
  • Threads will not synchronize upon completing their individual pieces of work (NOWAIT).

  Fortran - DO Directive Example PROGRAM VEC_ADD_DO INTEGER N, CHUNKSIZE, CHUNK, I PARAMETER (N=1000) PARAMETER (CHUNKSIZE=100) REAL A(N), B(N), C(N) ! Some initializations DO I = 1, N A(I) = I * 1.0 B(I) = A(I) ENDDO CHUNK = CHUNKSIZE !$OMP PARALLEL SHARED(A,B,C,CHUNK) PRIVATE(I) !$OMP DO SCHEDULE(DYNAMIC,CHUNK) DO I = 1, N C(I) = A(I) + B(I) ENDDO !$OMP END DO NOWAIT !$OMP END PARALLEL END

  C / C++ - for Directive Example #include <omp.h> #define CHUNKSIZE 100 #define N 1000 main () { int i, chunk; float a[N], b[N], c[N]; /* Some initializations */ for (i=0; i < N; i++) a[i] = b[i] = i * 1.0; chunk = CHUNKSIZE; #pragma omp parallel shared(a,b,c,chunk) private(i) { #pragma omp for schedule(dynamic,chunk) nowait for (i=0; i < N; i++) c[i] = a[i] + b[i]; } /* end of parallel section */ }

Work-Sharing Constructs
SECTIONS Directive

• The SECTIONS directive is a non-iterative work-sharing construct. It specifies that the enclosed section(s) of code are to be divided among the threads in the team.

• Independent SECTION directives are nested within a SECTIONS directive. Each SECTION is executed once by a thread in the team. Different sections may be executed by different threads. It is possible that for a thread to execute more than one section if it is quick enough and the implementation permits such.

Format:

Fortran	!$OMP SECTIONS [clause ...] PRIVATE (list) FIRSTPRIVATE (list) LASTPRIVATE (list) REDUCTION (operator | intrinsic : list) !$OMP SECTION block !$OMP SECTION block !$OMP END SECTIONS [ NOWAIT ]
C/C++	#pragma omp sections [clause ...] newline private (list) firstprivate (list) lastprivate (list) reduction (operator: list) nowait { #pragma omp section newline structured_block #pragma omp section newline structured_block }

Clauses:

• There is an implied barrier at the end of a SECTIONS directive, unless the NOWAIT/nowait clause is used.

• Clauses are described in detail later, in the Data Scope Attribute Clauses section.

Questions:

	What happens if the number of threads and the number of SECTIONs are different? More threads than SECTIONs? Less threads than SECTIONs?
	Which thread executes which SECTION?

Restrictions:

• It is illegal to branch into or out of section blocks.

• SECTION directives must occur within the lexical extent of an enclosing SECTIONS directive

Example: SECTIONS Directive

• Simple program demonstrating that different blocks of work will be done by different threads.

  Fortran - SECTIONS Directive Example PROGRAM VEC_ADD_SECTIONS INTEGER N, I PARAMETER (N=1000) REAL A(N), B(N), C(N), D(N) ! Some initializations DO I = 1, N A(I) = I * 1.5 B(I) = I + 22.35 ENDDO !$OMP PARALLEL SHARED(A,B,C,D), PRIVATE(I) !$OMP SECTIONS !$OMP SECTION DO I = 1, N C(I) = A(I) + B(I) ENDDO !$OMP SECTION DO I = 1, N D(I) = A(I) * B(I) ENDDO !$OMP END SECTIONS NOWAIT !$OMP END PARALLEL END

  C / C++ - sections Directive Example #include <omp.h> #define N 1000 main () { int i; float a[N], b[N], c[N], d[N]; /* Some initializations */ for (i=0; i < N; i++) a[i] = i * 1.5; b[i] = i + 22.35; #pragma omp parallel shared(a,b,c,d) private(i) { #pragma omp sections nowait { #pragma omp section for (i=0; i < N; i++) c[i] = a[i] + b[i]; #pragma omp section for (i=0; i < N; i++) d[i] = a[i] * b[i]; } /* end of sections */ } /* end of parallel section */ }

Work-Sharing Constructs
WORKSHARE Directive

• Fortran only

• The WORKSHARE directive divides the execution of the enclosed structured block into separate units of work, each of which is executed only once.

• The structured block must consist of only the following:
  • array assignments
  • scalar assignments
  • FORALL statements
  • FORALL constructs
  • WHERE statements
  • WHERE constructs
  • atomic constructs
  • critical constructs
  • parallel constructs

• See the OpenMP API documentation for additional information.

Format:

Fortran	!$OMP WORKSHARE structured block !$OMP END WORKSHARE [ NOWAIT ]

Restrictions:

• The construct must not contain any user defined function calls unless the function is ELEMENTAL.

Example: WORKSHARE Directive

• Simple array and scalar assigments shared by the team of threads. A unit of work would include:
  • Any scalar assignment
  • For array assignment statements, the assignment of each element is a unit of work

  Fortran - WORKSHARE Directive Example PROGRAM WORKSHARE INTEGER N, I, J PARAMETER (N=100) REAL AA(N,N), BB(N,N), CC(N,N), DD(N,N), FIRST, LAST ! Some initializations DO I = 1, N DO J = 1, N AA(J,I) = I * 1.0 BB(J,I) = J + 1.0 ENDDO ENDDO !$OMP PARALLEL SHARED(AA,BB,CC,DD,FIRST,LAST) !$OMP WORKSHARE CC = AA * BB DD = AA + BB FIRST = CC(1,1) + DD(1,1) LAST = CC(N,N) + DD(N,N) !$OMP END WORKSHARE NOWAIT !$OMP END PARALLEL END

Work-Sharing Constructs
SINGLE Directive

• The SINGLE directive specifies that the enclosed code is to be executed by only one thread in the team.

• May be useful when dealing with sections of code that are not thread safe (such as I/O)

Format:

Fortran	!$OMP SINGLE [clause ...] PRIVATE (list) FIRSTPRIVATE (list) block !$OMP END SINGLE [ NOWAIT ]
C/C++	#pragma omp single [clause ...] newline private (list) firstprivate (list) nowait structured_block

Clauses:

• Threads in the team that do not execute the SINGLE directive, wait at the end of the enclosed code block, unless a NOWAIT/nowait clause is specified.

• Clauses are described in detail later, in the Data Scope Attribute Clauses section.

Restrictions:

• It is illegal to branch into or out of a SINGLE block.

Combined Parallel Work-Sharing Constructs

• OpenMP provides three directives that are merely conveniences:
  • PARALLEL DO / parallel for
  • PARALLEL SECTIONS
  • PARALLEL WORKSHARE (fortran only)

• For the most part, these directives behave identically to an individual PARALLEL directive being immediately followed by a separate work-sharing directive.

• Most of the rules, clauses and restrictions that apply to both directives are in effect. See the OpenMP API for details.

• An example using the PARALLEL DO / parallel for combined directive is shown below.

  Fortran - PARALLEL DO Directive Example PROGRAM VECTOR_ADD INTEGER N, I, CHUNKSIZE, CHUNK PARAMETER (N=1000) PARAMETER (CHUNKSIZE=100) REAL A(N), B(N), C(N) ! Some initializations DO I = 1, N A(I) = I * 1.0 B(I) = A(I) ENDDO CHUNK = CHUNKSIZE !$OMP PARALLEL DO !$OMP& SHARED(A,B,C,CHUNK) PRIVATE(I) !$OMP& SCHEDULE(STATIC,CHUNK) DO I = 1, N C(I) = A(I) + B(I) ENDDO !$OMP END PARALLEL DO END

  C / C++ - parallel for Directive Example #include <omp.h> #define N 1000 #define CHUNKSIZE 100 main () { int i, chunk; float a[N], b[N], c[N]; /* Some initializations */ for (i=0; i < N; i++) a[i] = b[i] = i * 1.0; chunk = CHUNKSIZE; #pragma omp parallel for \ shared(a,b,c,chunk) private(i) \ schedule(static,chunk) for (i=0; i < n; i++) c[i] = a[i] + b[i]; }

TASK Construct

• New construct with OpenMP 3.0

• The TASK construct defines an explicit task, which may be executed by the encountering thread, or deferred for execution by any other thread in the team.

• The data environment of the task is determined by the data sharing attribute clauses.

• Task execution is subject to task scheduling - see the OpenMP 3.0 specification document for details.

Format:

Fortran	!$OMP TASK [clause ...] IF (scalar expression) UNTIED DEFAULT (PRIVATE | FIRSTPRIVATE | SHARED | NONE) PRIVATE (list) FIRSTPRIVATE (list) SHARED (list) block !$OMP END TASK
C/C++	#pragma omp task [clause ...] newline if (scalar expression) untied default (shared | none) private (list) firstprivate (list) shared (list) structured_block

Clauses and Restrictions:

• Please consult the OpenMP 3.0 specifications document for details.

Synchronization Constructs

• Consider a simple example where two threads on two different processors are both trying to increment a variable x at the same time (assume x is initially 0):

  THREAD 1: increment(x) { x = x + 1; } THREAD 1: 10 LOAD A, (x address) 20 ADD A, 1 30 STORE A, (x address)

  THREAD 2: increment(x) { x = x + 1; } THREAD 2: 10 LOAD A, (x address) 20 ADD A, 1 30 STORE A, (x address)

• One possible execution sequence:
  1. Thread 1 loads the value of x into register A.
  2. Thread 2 loads the value of x into register A.
  3. Thread 1 adds 1 to register A
  4. Thread 2 adds 1 to register A
  5. Thread 1 stores register A at location x
  6. Thread 2 stores register A at location x

  The resultant value of x will be 1, not 2 as it should be.

• To avoid a situation like this, the incrementation of x must be synchronized between the two threads to insure that the correct result is produced.

• OpenMP provides a variety of Synchronization Constructs that control how the execution of each thread proceeds relative to other team threads.

Synchronization Constructs
MASTER Directive

• The MASTER directive specifies a region that is to be executed only by the master thread of the team. All other threads on the team skip this section of code

• There is no implied barrier associated with this directive

Format:

Fortran	!$OMP MASTER block !$OMP END MASTER
C/C++	#pragma omp master newline structured_block

Restrictions:

• It is illegal to branch into or out of MASTER block.

Synchronization Constructs
CRITICAL Directive

• The CRITICAL directive specifies a region of code that must be executed by only one thread at a time.

Format:

Fortran	!$OMP CRITICAL [ name ] block !$OMP END CRITICAL
C/C++	#pragma omp critical [ name ] newline structured_block

Notes:

• If a thread is currently executing inside a CRITICAL region and another thread reaches that CRITICAL region and attempts to execute it, it will block until the first thread exits that CRITICAL region.

• The optional name enables multiple different CRITICAL regions to exist:
  • Names act as global identifiers. Different CRITICAL regions with the same name are treated as the same region.
  • All CRITICAL sections which are unnamed, are treated as the same section.

Restrictions:

• It is illegal to branch into or out of a CRITICAL block.

Example: CRITICAL Construct

• All threads in the team will attempt to execute in parallel, however, because of the CRITICAL construct surrounding the increment of x, only one thread will be able to read/increment/write x at any time

  Fortran - CRITICAL Directive Example PROGRAM CRITICAL INTEGER X X = 0 !$OMP PARALLEL SHARED(X) !$OMP CRITICAL X = X + 1 !$OMP END CRITICAL !$OMP END PARALLEL END

  C / C++ - critical Directive Example #include <omp.h> main() { int x; x = 0; #pragma omp parallel shared(x) { #pragma omp critical x = x + 1; } /* end of parallel section */ }

Synchronization Constructs
BARRIER Directive

• The BARRIER directive synchronizes all threads in the team.

• When a BARRIER directive is reached, a thread will wait at that point until all other threads have reached that barrier. All threads then resume executing in parallel the code that follows the barrier.

Format:

Fortran	!$OMP BARRIER
C/C++	#pragma omp barrier newline

Restrictions:

• All threads in a team (or none) must execute the BARRIER region.

• The sequence of work-sharing regions and barrier regions encountered must be the same for every thread in a team.

Synchronization Constructs
TASKWAIT Directive

• The TASKWAIT construct specifies a wait on the completion of child tasks generated since the beginning of the current task.

Format:

Fortran	!$OMP TASKWAIT
C/C++	#pragma omp taskwait newline

Restrictions:

• Because the taskwait construct does not have a C language statement as part of its syntax, there are some restrictions on its placement within a program. The taskwait directive may be placed only at a point where a base language statement is allowed. The taskwait directive may not be used in place of the statement following an if, while, do, switch, or label. See the OpenMP 3.0 specifications document for details.

Synchronization Constructs
ATOMIC Directive

• The ATOMIC directive specifies that a specific memory location must be updated atomically, rather than letting multiple threads attempt to write to it. In essence, this directive provides a mini-CRITICAL section.

Format:

Fortran	!$OMP ATOMIC statement_expression
C/C++	#pragma omp atomic newline statement_expression

Restrictions:

• The directive applies only to a single, immediately following statement

• An atomic statement must follow a specific syntax. See the most recent OpenMP specs for this.

Synchronization Constructs
FLUSH Directive

• The FLUSH directive identifies a synchronization point at which the implementation must provide a consistent view of memory. Thread-visible variables are written back to memory at this point.

• There is a fair amount of discussion on this directive within OpenMP circles that you may wish to consult for more information.

• To quote from the openmp.org FAQ:

  Q17: Is the !$omp flush directive necessary on a cache coherent system?

  A17: Yes the flush directive is necessary. Look in the OpenMP specifications for examples of it's uses. The directive is necessary to instruct the compiler that the variable must be written to/read from the memory system, i.e. that the variable can not be kept in a local CPU register over the flush "statement" in your code.

  Cache coherency makes certain that if one CPU executes a read or write instruction from/to memory, then all other CPUs in the system will get the same value from that memory address when they access it. All caches will show a coherent value. However, in the OpenMP standard there must be a way to instruct the compiler to actually insert the read/write machine instruction and not postpone it. Keeping a variable in a register in a loop is very common when producing efficient machine language code for a loop.

• Also see the most recent OpenMP specs for details.

Format:

Fortran	!$OMP FLUSH (list)
C/C++	#pragma omp flush (list) newline

Notes:

• The optional list contains a list of named variables that will be flushed in order to avoid flushing all variables. For pointers in the list, note that the pointer itself is flushed, not the object it points to.

• Implementations must ensure any prior modifications to thread-visible variables are visible to all threads after this point; ie. compilers must restore values from registers to memory, hardware might need to flush write buffers, etc

• The FLUSH directive is implied for the directives shown in the table below. The directive is not implied if a NOWAIT clause is present.

  Fortran	C / C++
  BARRIER END PARALLEL CRITICAL and END CRITICAL END DO END SECTIONS END SINGLE ORDERED and END ORDERED	barrier parallel - upon entry and exit critical - upon entry and exit ordered - upon entry and exit for - upon exit sections - upon exit single - upon exit

Synchronization Constructs
ORDERED Directive

• The ORDERED directive specifies that iterations of the enclosed loop will be executed in the same order as if they were executed on a serial processor.

• Threads will need to wait before executing their chunk of iterations if previous iterations haven't completed yet.

• Used within a DO / for loop with an ORDERED clause

• The ORDERED directive provides a way to "fine tune" where ordering is to be applied within a loop. Otherwise, it is not required.

Format:

Fortran	!$OMP DO ORDERED [clauses...] (loop region) !$OMP ORDERED (block) !$OMP END ORDERED (end of loop region) !$OMP END DO
C/C++	#pragma omp for ordered [clauses...] (loop region) #pragma omp ordered newline structured_block (endo of loop region)

Restrictions:

• An ORDERED directive can only appear in the dynamic extent of the following directives:
  • DO or PARALLEL DO (Fortran)
  • for or parallel for (C/C++)

• Only one thread is allowed in an ordered section at any time

• It is illegal to branch into or out of an ORDERED block.

• An iteration of a loop must not execute the same ORDERED directive more than once, and it must not execute more than one ORDERED directive.

• A loop which contains an ORDERED directive, must be a loop with an ORDERED clause.

THREADPRIVATE Directive

• The THREADPRIVATE directive is used to make global file scope variables (C/C++) or common blocks (Fortran) local and persistent to a thread through the execution of multiple parallel regions.

Format:

Fortran	!$OMP THREADPRIVATE (/cb/, ...) cb is the name of a common block
C/C++	#pragma omp threadprivate (list)

Notes:

• The directive must appear after the declaration of listed variables/common blocks. Each thread then gets its own copy of the variable/common block, so data written by one thread is not visible to other threads. For example:

  Fortran - THREADPRIVATE Directive Example PROGRAM THREADPRIV INTEGER A, B, I, TID, OMP_GET_THREAD_NUM REAL*4 X COMMON /C1/ A !$OMP THREADPRIVATE(/C1/, X) C Explicitly turn off dynamic threads CALL OMP_SET_DYNAMIC(.FALSE.) PRINT *, '1st Parallel Region:' !$OMP PARALLEL PRIVATE(B, TID) TID = OMP_GET_THREAD_NUM() A = TID B = TID X = 1.1 * TID + 1.0 PRINT *, 'Thread',TID,': A,B,X=',A,B,X !$OMP END PARALLEL PRINT *, '************************************' PRINT *, 'Master thread doing serial work here' PRINT *, '************************************' PRINT *, '2nd Parallel Region: ' !$OMP PARALLEL PRIVATE(TID) TID = OMP_GET_THREAD_NUM() PRINT *, 'Thread',TID,': A,B,X=',A,B,X !$OMP END PARALLEL END Output: 1st Parallel Region: Thread 0 : A,B,X= 0 0 1.000000000 Thread 1 : A,B,X= 1 1 2.099999905 Thread 3 : A,B,X= 3 3 4.300000191 Thread 2 : A,B,X= 2 2 3.200000048 ************************************ Master thread doing serial work here ************************************ 2nd Parallel Region: Thread 0 : A,B,X= 0 0 1.000000000 Thread 2 : A,B,X= 2 0 3.200000048 Thread 3 : A,B,X= 3 0 4.300000191 Thread 1 : A,B,X= 1 0 2.099999905

  C/C++ - threadprivate Directive Example #include <omp.h> int a, b, i, tid; float x; #pragma omp threadprivate(a, x) main () { /* Explicitly turn off dynamic threads */ omp_set_dynamic(0); printf("1st Parallel Region:\n"); #pragma omp parallel private(b,tid) { tid = omp_get_thread_num(); a = tid; b = tid; x = 1.1 * tid +1.0; printf("Thread %d: a,b,x= %d %d %f\n",tid,a,b,x); } /* end of parallel section */ printf("************************************\n"); printf("Master thread doing serial work here\n"); printf("************************************\n"); printf("2nd Parallel Region:\n"); #pragma omp parallel private(tid) { tid = omp_get_thread_num(); printf("Thread %d: a,b,x= %d %d %f\n",tid,a,b,x); } /* end of parallel section */ } Output: 1st Parallel Region: Thread 0: a,b,x= 0 0 1.000000 Thread 2: a,b,x= 2 2 3.200000 Thread 3: a,b,x= 3 3 4.300000 Thread 1: a,b,x= 1 1 2.100000 ************************************ Master thread doing serial work here ************************************ 2nd Parallel Region: Thread 0: a,b,x= 0 0 1.000000 Thread 3: a,b,x= 3 0 4.300000 Thread 1: a,b,x= 1 0 2.100000 Thread 2: a,b,x= 2 0 3.200000

• On first entry to a parallel region, data in THREADPRIVATE variables and common blocks should be assumed undefined, unless a COPYIN clause is specified in the PARALLEL directive

• THREADPRIVATE variables differ from PRIVATE variables (discussed later) because they are able to persist between different parallel sections of a code.

Restrictions:

• Data in THREADPRIVATE objects is guaranteed to persist only if the dynamic threads mechanism is "turned off" and the number of threads in different parallel regions remains constant. The default setting of dynamic threads is undefined.

• The THREADPRIVATE directive must appear after every declaration of a thread private variable/common block.

• Fortran: only named common blocks can be made THREADPRIVATE.

Data Scope Attribute Clauses

• Also called Data-sharing Attribute Clauses

• An important consideration for OpenMP programming is the understanding and use of data scoping

• Because OpenMP is based upon the shared memory programming model, most variables are shared by default

• Global variables include:
  • Fortran: COMMON blocks, SAVE variables, MODULE variables
  • C: File scope variables, static

• Private variables include:
  • Loop index variables
  • Stack variables in subroutines called from parallel regions
  • Fortran: Automatic variables within a statement block

• The OpenMP Data Scope Attribute Clauses are used to explicitly define how variables should be scoped. They include:
  • PRIVATE
  • FIRSTPRIVATE
  • LASTPRIVATE
  • SHARED
  • DEFAULT
  • REDUCTION
  • COPYIN

• Data Scope Attribute Clauses are used in conjunction with several directives (PARALLEL, DO/for, and SECTIONS) to control the scoping of enclosed variables.

• These constructs provide the ability to control the data environment during execution of parallel constructs.

  • They define how and which data variables in the serial section of the program are transferred to the parallel sections of the program (and back)

  • They define which variables will be visible to all threads in the parallel sections and which variables will be privately allocated to all threads.

• Data Scope Attribute Clauses are effective only within their lexical/static extent.

• Important: Please consult the latest OpenMP specs for important details and discussion on this topic.

• A Clauses / Directives Summary Table is provided for convenience.

PRIVATE Clause

• The PRIVATE clause declares variables in its list to be private to each thread.

Format:

Fortran	PRIVATE (list)
C/C++	private (list)

Notes:

• PRIVATE variables behave as follows:

  • A new object of the same type is declared once for each thread in the team

  • All references to the original object are replaced with references to the new object

  • Variables declared PRIVATE should be assumed to be uninitialized for each thread

• Comparison between PRIVATE and THREADPRIVATE:

  	PRIVATE	THREADPRIVATE
  Data Item	C/C++: variable Fortran: variable or common block	C/C++: variable Fortran: common block
  Where Declared	At start of region or work-sharing group	In declarations of each routine using block or global file scope
  Persistent?	No	Yes
  Extent	Lexical only - unless passed as an argument to subroutine	Dynamic
  Initialized	Use FIRSTPRIVATE	Use COPYIN

Questions:

For the C/C++ and Fortran THREADPRIVATE example codes, what output would you expect for alpha[3] and beta[3]? Why?

SHARED Clause

• The SHARED clause declares variables in its list to be shared among all threads in the team.

Format:

Fortran	SHARED (list)
C/C++	shared (list)

Notes:

• A shared variable exists in only one memory location and all threads can read or write to that address

• It is the programmer's responsibility to ensure that multiple threads properly access SHARED variables (such as via CRITICAL sections)

DEFAULT Clause

• The DEFAULT clause allows the user to specify a default scope for all variables in the lexical extent of any parallel region.

Format:

Fortran	DEFAULT (PRIVATE | FIRSTPRIVATE | SHARED | NONE)
C/C++	default (shared | none)

Notes:

• Specific variables can be exempted from the default using the PRIVATE, SHARED, FIRSTPRIVATE, LASTPRIVATE, and REDUCTION clauses

• The C/C++ OpenMP specification does not include private or firstprivate as a possible default. However, actual implementations may provide this option.

• Using NONE as a default requires that the programmer explicitly scope all variables.

Restrictions:

• Only one DEFAULT clause can be specified on a PARALLEL directive

FIRSTPRIVATE Clause

• The FIRSTPRIVATE clause combines the behavior of the PRIVATE clause with automatic initialization of the variables in its list.

Format:

Fortran	FIRSTPRIVATE (list)
C/C++	firstprivate (list)

Notes:

• Listed variables are initialized according to the value of their original objects prior to entry into the parallel or work-sharing construct.

LASTPRIVATE Clause

• The LASTPRIVATE clause combines the behavior of the PRIVATE clause with a copy from the last loop iteration or section to the original variable object.

Format:

Fortran	LASTPRIVATE (list)
C/C++	lastprivate (list)

Notes:

• The value copied back into the original variable object is obtained from the last (sequentially) iteration or section of the enclosing construct.

  For example, the team member which executes the final iteration for a DO section, or the team member which does the last SECTION of a SECTIONS context performs the copy with its own values

COPYIN Clause

• The COPYIN clause provides a means for assigning the same value to THREADPRIVATE variables for all threads in the team.

Format:

Fortran	COPYIN (list)
C/C++	copyin (list)

Notes:

• List contains the names of variables to copy. In Fortran, the list can contain both the names of common blocks and named variables.

• The master thread variable is used as the copy source. The team threads are initialized with its value upon entry into the parallel construct.

COPYPRIVATE Clause

• The COPYPRIVATE clause can be used to broadcast values acquired by a single thread directly to all instances of the private variables in the other threads.

• Associated with the SINGLE directive

• See the most recent OpenMP specs document for additional discussion and examples.

Format:

Fortran	COPYPRIVATE (list)
C/C++	copyprivate (list)

REDUCTION Clause

• The REDUCTION clause performs a reduction on the variables that appear in its list.

• A private copy for each list variable is created for each thread. At the end of the reduction, the reduction variable is applied to all private copies of the shared variable, and the final result is written to the global shared variable.

Format:

Fortran	REDUCTION (operator|intrinsic: list)
C/C++	reduction (operator: list)

Example: REDUCTION - Vector Dot Product:

• Iterations of the parallel loop will be distributed in equal sized blocks to each thread in the team (SCHEDULE STATIC)

• At the end of the parallel loop construct, all threads will add their values of "result" to update the master thread's global copy.

  Fortran - REDUCTION Clause Example PROGRAM DOT_PRODUCT INTEGER N, CHUNKSIZE, CHUNK, I PARAMETER (N=100) PARAMETER (CHUNKSIZE=10) REAL A(N), B(N), RESULT ! Some initializations DO I = 1, N A(I) = I * 1.0 B(I) = I * 2.0 ENDDO RESULT= 0.0 CHUNK = CHUNKSIZE !$OMP PARALLEL DO !$OMP& DEFAULT(SHARED) PRIVATE(I) !$OMP& SCHEDULE(STATIC,CHUNK) !$OMP& REDUCTION(+:RESULT) DO I = 1, N RESULT = RESULT + (A(I) * B(I)) ENDDO !$OMP END PARALLEL DO NOWAIT PRINT *, 'Final Result= ', RESULT END

  C / C++ - reduction Clause Example #include <omp.h> main () { int i, n, chunk; float a[100], b[100], result; /* Some initializations */ n = 100; chunk = 10; result = 0.0; for (i=0; i < n; i++) { a[i] = i * 1.0; b[i] = i * 2.0; } #pragma omp parallel for \ default(shared) private(i) \ schedule(static,chunk) \ reduction(+:result) for (i=0; i < n; i++) result = result + (a[i] * b[i]); printf("Final result= %f\n",result); }

Restrictions:

• Variables in the list must be named scalar variables. They can not be array or structure type variables. They must also be declared SHARED in the enclosing context.

• Reduction operations may not be associative for real numbers.

• The REDUCTION clause is intended to be used on a region or work-sharing construct in which the reduction variable is used only in statements which have one of following forms:

  Fortran	C / C++
  x = x operator expr x = expr operator x (except subtraction) x = intrinsic(x, expr) x = intrinsic(expr, x)	x = x op expr x = expr op x (except subtraction) x binop = expr x++ ++x x-- --x
  x is a scalar variable in the list expr is a scalar expression that does not reference x intrinsic is one of MAX, MIN, IAND, IOR, IEOR operator is one of +, *, -, .AND., .OR., .EQV., .NEQV.	x is a scalar variable in the list expr is a scalar expression that does not reference x op is not overloaded, and is one of +, *, -, /, &, ^, |, &&, || binop is not overloaded, and is one of +, *, -, /, &, ^, |

Clauses / Directives Summary

• The table below summarizes which clauses are accepted by which OpenMP directives.

  Clause	Directive
  	PARALLEL	DO/for	SECTIONS	SINGLE	PARALLEL DO/for	PARALLEL SECTIONS
  IF
  PRIVATE
  SHARED
  DEFAULT
  FIRSTPRIVATE
  LASTPRIVATE
  REDUCTION
  COPYIN
  COPYPRIVATE
  SCHEDULE
  ORDERED
  NOWAIT

• The following OpenMP directives do not accept clauses:
  • MASTER
  • CRITICAL
  • BARRIER
  • ATOMIC
  • FLUSH
  • ORDERED
  • THREADPRIVATE

• Implementations may (and do) differ from the standard in which clauses are supported by each directive.

Directive Binding and Nesting Rules

This section is provided mainly as a quick reference on rules which govern OpenMP directives and binding. Users should consult their implementation documentation and the OpenMP standard for other rules and restrictions.

• Unless indicated otherwise, rules apply to both Fortran and C/C++ OpenMP implementations.

• Note: the Fortran API also defines a number of Data Environment rules. Those have not been reproduced here.

Directive Binding:

• The DO/for, SECTIONS, SINGLE, MASTER and BARRIER directives bind to the dynamically enclosing PARALLEL, if one exists. If no parallel region is currently being executed, the directives have no effect.

• The ORDERED directive binds to the dynamically enclosing DO/for.

• The ATOMIC directive enforces exclusive access with respect to ATOMIC directives in all threads, not just the current team.

• The CRITICAL directive enforces exclusive access with respect to CRITICAL directives in all threads, not just the current team.

• A directive can never bind to any directive outside the closest enclosing PARALLEL.

Directive Nesting:

• A PARALLEL directive dynamically inside another PARALLEL directive logically establishes a new team, which is composed of only the current thread unless nested parallelism is enabled.

• DO/for, SECTIONS, and SINGLE directives that bind to the same PARALLEL are not allowed to be nested inside of each other.

• DO/for, SECTIONS, and SINGLE directives are not permitted in the dynamic extent of CRITICAL, ORDERED and MASTER regions.

• CRITICAL directives with the same name are not permitted to be nested inside of each other.

• BARRIER directives are not permitted in the dynamic extent of DO/for, ORDERED, SECTIONS, SINGLE, MASTER and CRITICAL regions.

• MASTER directives are not permitted in the dynamic extent of DO/for, SECTIONS and SINGLE directives.

• ORDERED directives are not permitted in the dynamic extent of CRITICAL regions.

• Any directive that is permitted when executed dynamically inside a PARALLEL region is also legal when executed outside a parallel region. When executed dynamically outside a user-specified parallel region, the directive is executed with respect to a team composed of only the master thread.

Overview:

• The OpenMP standard defines an API for library calls that perform a variety of functions:

  • Query the number of threads/processors, set number of threads to use

  • General purpose locking routines (semaphores)

  • Portable wall clock timing routines

  • Set execution environment functions: nested parallelism, dynamic adjustment of threads.

• For C/C++, it may be necessary to specify the include file "omp.h".

• For the Lock routines/functions:

  • The lock variable must be accessed only through the locking routines

  • For Fortran, the lock variable should be of type integer and of a kind large enough to hold an address.

  • For C/C++, the lock variable must have type omp_lock_t or type omp_nest_lock_t, depending on the function being used.

• Implementation notes:

  • Your implementation may or may not support nested parallelism and/or dynamic threads. If nested parallelism is supported, it is often only nominal, in that a nested parallel region may only have one thread.

  • Consult your implementation's documentation for details - or experiment and find out for yourself if you can't find it in the documentation.

OMP_SET_NUM_THREADS

• Sets the number of threads that will be used in the next parallel region. Must be a postive integer.

Format:

Fortran	SUBROUTINE OMP_SET_NUM_THREADS(scalar_integer_expression)
C/C++	#include <omp.h> void omp_set_num_threads(int num_threads)

Notes & Restrictions:

• The dynamic threads mechanism modifies the effect of this routine.
  • Enabled: specifies the maximum number of threads that can be used for any parallel region by the dynamic threads mechanism.
  • Disabled: specifies exact number of threads to use until next call to this routine.

• This routine can only be called from the serial portions of the code

• This call has precedence over the OMP_NUM_THREADS environment variable

OMP_GET_NUM_THREADS

• Returns the number of threads that are currently in the team executing the parallel region from which it is called.

Format:

Fortran	INTEGER FUNCTION OMP_GET_NUM_THREADS()
C/C++	#include <omp.h> int omp_get_num_threads(void)

Notes & Restrictions:

• If this call is made from a serial portion of the program, or a nested parallel region that is serialized, it will return 1.

• The default number of threads is implementation dependent.

OMP_GET_MAX_THREADS

• Returns the maximum value that can be returned by a call to the OMP_GET_NUM_THREADS function.

  Fortran	INTEGER FUNCTION OMP_GET_MAX_THREADS()
  C/C++	#include <omp.h> int omp_get_max_threads(void)

Notes & Restrictions:

• Generally reflects the number of threads as set by the OMP_NUM_THREADS environment variable or the OMP_SET_NUM_THREADS() library routine.

• May be called from both serial and parallel regions of code.

OMP_GET_THREAD_NUM

• Returns the thread number of the thread, within the team, making this call. This number will be between 0 and OMP_GET_NUM_THREADS-1. The master thread of the team is thread 0

Format:

Fortran	INTEGER FUNCTION OMP_GET_THREAD_NUM()
C/C++	#include <omp.h> int omp_get_thread_num(void)

Notes & Restrictions:

• If called from a nested parallel region, or a serial region, this function will return 0.

Examples:

• Example 1 is the correct way to determine the number of threads in a parallel region.
• Example 2 is incorrect - the TID variable must be PRIVATE
• Example 3 is incorrect - the OMP_GET_THREAD_NUM call is outside the parallel region

  Fortran - determining the number of threads in a parallel region Example 1: Correct PROGRAM HELLO INTEGER TID, OMP_GET_THREAD_NUM !$OMP PARALLEL PRIVATE(TID) TID = OMP_GET_THREAD_NUM() PRINT *, 'Hello World from thread = ', TID ... !$OMP END PARALLEL END Example 2: Incorrect PROGRAM HELLO INTEGER TID, OMP_GET_THREAD_NUM !$OMP PARALLEL TID = OMP_GET_THREAD_NUM() PRINT *, 'Hello World from thread = ', TID ... !$OMP END PARALLEL END Example 3: Incorrect PROGRAM HELLO INTEGER TID, OMP_GET_THREAD_NUM TID = OMP_GET_THREAD_NUM() PRINT *, 'Hello World from thread = ', TID !$OMP PARALLEL ... !$OMP END PARALLEL END

OMP_GET_THREAD_LIMIT

• New with OpenMP 3.0. Returns the maximum number of OpenMP threads available to a program.

Format:

Fortran	INTEGER FUNCTION OMP_GET_THREAD_LIMIT
C/C++	#include <omp.h> int omp_get_thread_limit (void)

Notes:

• Also see the OMP_THREAD_LIMIT environment variable.

OMP_GET_NUM_PROCS

• Returns the number of processors that are available to the program.

Format:

Fortran	INTEGER FUNCTION OMP_GET_NUM_PROCS()
C/C++	#include <omp.h> int omp_get_num_procs(void)

OMP_IN_PARALLEL

• May be called to determine if the section of code which is executing is parallel or not.

Format:

Fortran	LOGICAL FUNCTION OMP_IN_PARALLEL()
C/C++	#include <omp.h> int omp_in_parallel(void)

Notes & Restrictions:

• For Fortran, this function returns .TRUE. if it is called from the dynamic extent of a region executing in parallel, and .FALSE. otherwise. For C/C++, it will return a non-zero integer if parallel, and zero otherwise.

OMP_SET_DYNAMIC

• Enables or disables dynamic adjustment (by the run time system) of the number of threads available for execution of parallel regions.

Format:

Fortran	SUBROUTINE OMP_SET_DYNAMIC(scalar_logical_expression)
C/C++	#include <omp.h> void omp_set_dynamic(int dynamic_threads)

Notes & Restrictions:

• For Fortran, if called with .TRUE. then the number of threads available for subsequent parallel regions can be adjusted automatically by the run-time environment. If called with .FALSE., dynamic adjustment is disabled.

• For C/C++, if dynamic_threads evaluates to non-zero, then the mechanism is enabled, otherwise it is disabled.

• The OMP_SET_DYNAMIC subroutine has precedence over the OMP_DYNAMIC environment variable.

• The default setting is implementation dependent.

• Must be called from a serial section of the program.

OMP_GET_DYNAMIC

• Used to determine if dynamic thread adjustment is enabled or not.

Format:

Fortran	LOGICAL FUNCTION OMP_GET_DYNAMIC()
C/C++	#include <omp.h> int omp_get_dynamic(void)

Notes & Restrictions:

• For Fortran, this function returns .TRUE. if dynamic thread adjustment is enabled, and .FALSE. otherwise.

• For C/C++, non-zero will be returned if dynamic thread adjustment is enabled, and zero otherwise.

OMP_SET_NESTED

• Used to enable or disable nested parallelism.

Format:

Fortran	SUBROUTINE OMP_SET_NESTED(scalar_logical_expression)
C/C++	#include <omp.h> void omp_set_nested(int nested)

Notes & Restrictions:

• For Fortran, calling this function with .FALSE. will disable nested parallelism, and calling with .TRUE. will enable it.

• For C/C++, if nested evaluates to non-zero, nested parallelism is enabled; otherwise it is disabled.

• The default is for nested parallelism to be disabled.

• This call has precedence over the OMP_NESTED environment variable

OMP_GET_NESTED

• Used to determine if nested parallelism is enabled or not.

Format:

Fortran	LOGICAL FUNCTION OMP_GET_NESTED
C/C++	#include <omp.h> int omp_get_nested (void)

Notes & Restrictions:

• For Fortran, this function returns .TRUE. if nested parallelism is enabled, and .FALSE. otherwise.

• For C/C++, non-zero will be returned if nested parallelism is enabled, and zero otherwise.

OMP_SET_SCHEDULE
OMP_SET_SCHEDULE
OMP_GET_SCHEDULE
OMP_SET_MAX_ACTIVE_LEVELS
OMP_GET_MAX_ACTIVE_LEVELS
OMP_GET_LEVEL
OMP_GET_ANCESTOR_THREAD_NUM
OMP_GET_TEAM_SIZE
OMP_GET_ACTIVE_LEVEL
OMP_INIT_NEST_LOCK
OMP_DESTROY_NEST_LOCK
OMP_SET_NEST_LOCK
OMP_UNSET_NEST_LOCK
OMP_TEST_NEST_LOCK

• These routines are new with OpenMP 3.0. Please consult the most recent specs for details.

OMP_INIT_LOCK

• This subroutine initializes a lock associated with the lock variable.

Format:

Fortran	SUBROUTINE OMP_INIT_LOCK(var) SUBROUTINE OMP_INIT_NEST_LOCK(var)
C/C++	#include <omp.h> void omp_init_lock(omp_lock_t *lock) void omp_init_nest_lock(omp_nest_lock_t *lock)

Notes & Restrictions:

• The initial state is unlocked

• For Fortran, var must be an integer large enough to hold an address, such as INTEGER*8 on 64-bit systems.

OMP_DESTROY_LOCK

• This subroutine disassociates the given lock variable from any locks.

Format:

Fortran	SUBROUTINE OMP_DESTROY_LOCK(var) SUBROUTINE OMP_DESTROY_NEST_LOCK(var)
C/C++	#include <omp.h> void omp_destroy_lock(omp_lock_t *lock) void omp_destroy_nest__lock(omp_nest_lock_t *lock)

Notes & Restrictions:

• It is illegal to call this routine with a lock variable that is not initialized.

• For Fortran, var must be an integer large enough to hold an address, such as INTEGER*8 on 64-bit systems.

OMP_SET_LOCK

• This subroutine forces the executing thread to wait until the specified lock is available. A thread is granted ownership of a lock when it becomes available.

Format:

Fortran	SUBROUTINE OMP_SET_LOCK(var) SUBROUTINE OMP_SET_NEST_LOCK(var)
C/C++	#include <omp.h> void omp_set_lock(omp_lock_t *lock) void omp_set_nest__lock(omp_nest_lock_t *lock)

Notes & Restrictions:

• It is illegal to call this routine with a lock variable that is not initialized.

• For Fortran, var must be an integer large enough to hold an address, such as INTEGER*8 on 64-bit systems.

OMP_UNSET_LOCK

• This subroutine releases the lock from the executing subroutine.

Format:

Fortran	SUBROUTINE OMP_UNSET_LOCK(var) SUBROUTINE OMP_UNSET_NEST_LOCK(var)
C/C++	#include <omp.h> void omp_unset_lock(omp_lock_t *lock) void omp_unset_nest__lock(omp_nest_lock_t *lock)

Notes & Restrictions:

• It is illegal to call this routine with a lock variable that is not initialized.

• For Fortran, var must be an integer large enough to hold an address, such as INTEGER*8 on 64-bit systems.

OMP_TEST_LOCK

• This subroutine attempts to set a lock, but does not block if the lock is unavailable.

Format:

Fortran	SUBROUTINE OMP_TEST_LOCK(var) SUBROUTINE OMP_TEST_NEST_LOCK(var)
C/C++	#include <omp.h> int omp_test_lock(omp_lock_t *lock) int omp_test_nest__lock(omp_nest_lock_t *lock)

Notes & Restrictions:

• For Fortran, .TRUE. is returned if the lock was set successfully, otherwise .FALSE. is returned.

• For Fortran, var must be an integer large enough to hold an address, such as INTEGER*8 on 64-bit systems.

• For C/C++, non-zero is returned if the lock was set successfully, otherwise zero is returned.

• It is illegal to call this routine with a lock variable that is not initialized.

OMP_GET_WTIME

• Provides a portable wall clock timing routine

• Returns a double-precision floating point value equal to the number of elapsed seconds since some point in the past. Usually used in "pairs" with the value of the first call subtracted from the value of the second call to obtain the elapsed time for a block of code.

• Designed to be "per thread" times, and therefore may not be globally consistent across all threads in a team - depends upon what a thread is doing compared to other threads.

Format:

Fortran	DOUBLE PRECISION FUNCTION OMP_GET_WTIME()
C/C++	#include <omp.h> double omp_get_wtime(void)

Notes & Restrictions:

• Requires OpenMP version 2.0 support

OMP_GET_WTICK

• Provides a portable wall clock timing routine

• Returns a double-precision floating point value equal to the number of seconds between successive clock ticks.

Format:

Fortran	DOUBLE PRECISION FUNCTION OMP_GET_WTICK()
C/C++	#include <omp.h> double omp_get_wtick(void)

Notes & Restrictions:

• Requires OpenMP version 2.0 support

• OpenMP provides the following environment variables for controlling the execution of parallel code.

• All environment variable names are uppercase. The values assigned to them are not case sensitive.

OMP_SCHEDULE

Applies only to DO, PARALLEL DO (Fortran) and for, parallel for (C/C++) directives which have their schedule clause set to RUNTIME. The value of this variable determines how iterations of the loop are scheduled on processors. For example:

setenv OMP_SCHEDULE "guided, 4"
setenv OMP_SCHEDULE "dynamic"

OMP_NUM_THREADS

Sets the maximum number of threads to use during execution. For example:

setenv OMP_NUM_THREADS 8

OMP_DYNAMIC

Enables or disables dynamic adjustment of the number of threads available for execution of parallel regions. Valid values are TRUE or FALSE. For example:

setenv OMP_DYNAMIC TRUE

OMP_NESTED

Enables or disables nested parallelism. Valid values are TRUE or FALSE. For example:

setenv OMP_NESTED TRUE

Implementation notes:

• Your implementation may or may not support nested parallelism and/or dynamic threads. If nested parallelism is supported, it is often only nominal, in that a nested parallel region may only have one thread.

• Consult your implementation's documentation for details - or experiment and find out for yourself if you can't find it in the documentation.

OMP_STACKSIZE

New with OpenMP 3.0. Controls the size of the stack for created (non-Master) threads. Examples:

setenv OMP_STACKSIZE 2000500B
setenv OMP_STACKSIZE "3000 k "
setenv OMP_STACKSIZE 10M
setenv OMP_STACKSIZE " 10 M "
setenv OMP_STACKSIZE "20 m "
setenv OMP_STACKSIZE " 1G"
setenv OMP_STACKSIZE 20000

OMP_WAIT_POLICY

New with OpenMP 3.0. Provides a hint to an OpenMP implementation about the desired behavior of waiting threads. A compliant OpenMP implementation may or may not abide by the setting of the environment variable. Valid values are ACTIVE and PASSIVE. ACTIVE specifies that waiting threads should mostly be active, i.e., consume processor cycles, while waiting. PASSIVE specifies that waiting threads should mostly be passive, i.e., not consume processor cycles, while waiting. The details of the ACTIVE and PASSIVE behaviors are implementation defined. Examples:

setenv OMP_WAIT_POLICY ACTIVE
setenv OMP_WAIT_POLICY active
setenv OMP_WAIT_POLICY PASSIVE
setenv OMP_WAIT_POLICY passive

OMP_MAX_ACTIVE_LEVELS

New with OpenMP 3.0. Controls the maximum number of nested active parallel regions. The value of this environment variable must be a non-negative integer. The behavior of the program is implementation defined if the requested value of OMP_MAX_ACTIVE_LEVELS is greater than the maximum number of nested active parallel levels an implementation can support, or if the value is not a non-negative integer. Example:

setenv OMP_MAX_ACTIVE_LEVELS 2

OMP_THREAD_LIMIT

New with OpenMP 3.0. Sets the number of OpenMP threads to use for the whole OpenMP program. The value of this environment variable must be a positive integer. The behavior of the program is implementation defined if the requested value of OMP_THREAD_LIMIT is greater than the number of threads an implementation can support, or if the value is not a positive integer. Example:

setenv OMP_THREAD_LIMIT 8

Thread Stack Size:

• The OpenMP standard does not specify how much stack space a thread should have. Consequently, implementations will differ in the default thread stack size.

• Default thread stack size can be easy to exhaust. It can also be non-portable between compilers For example, the table below shows some approximate thread stack size limits at LC using the default version compilers (Jan '09).

  Compiler	Approx. Stack Limit	Approx. Array Size (doubles)
  IBM POWER5 AIX: xlc, xlf	8 MB	707 x 707
  AMD Opteron Linux icc, ifort	8 MB	707 x 707
  AMD Opteron Linux pgcc, pgf90	16 MB	1024 x 1024
  AMD Opteron Linux gcc, gfortran	4 MB	512 x 512
  AMD Opteron Linux pathcc pathf90	67 MB >1.4GB	2048 x 2048 >9500 x 9500

• Threads that exceed their stack allocation may or may not seg fault. An application may continue to run while data is being corrupted.

• Statically linked codes may be subject to further stack restrictions.

• A user's login shell may also restrict stack size

• Examples for increasing the thread stack size to 12 MB at LC:

  Platform	Shell Commands csh	Shell Commands ksh
  IBM POWER5 AIX	limit stacksize 12288 setenv XLSMPOPTS "stack=12000000"	ulimit -s 12288 export XLSMPOPTS="stack=12000000"
  AMD Opteron Linux	limit stacksize 12288 setenv KMP_STACKSIZE 12000000	ulimit -s 12288 export KMP_STACKSIZE=12000000

LC OpenMP Implementations:

• OpenMP is fully supported all of LC's IBM and Opteron systems.

• LC maintains different versions of compilers. For the most recent information, please see: www.llnl.gov/asci/platforms/bluepac/CompsAvails.html

• All versions should be at least OpenMP 2.0 compliant. None are OpenMP 3.0 compliant as of Jan, 2009.

Compiling:

• Use the following compiler flags to "turn on" OpenMP compilations:

  Compiler	Flag
  IBM	-qsmp=omp
  Intel	-openmp
  PathScale	-mp
  PGI	-mp
  GNU	-fopenmp

Documentation:

• IBM compiler documentation:
  www-4.ibm.com/software/ad/fortran and www-4.ibm.com/software/ad/caix
  Locally: see the /usr/local/doc/xlf* and /usr/local/doc/xlc* files; man pages.

• Intel compiler documentation:
  www.intel.com/software/products/compilers/
  Locally: see the /usr/local/doc/ia32_ref and /usr/local/doc/intel_compilers files; man pages

• PathScale:
  www.pathscale.com
  Locally: man pages

• PGI:
  www.pgroup.com
  Locally: man pages

• GNU:
  gnu.org
  Locally: man pages

This completes the tutorial.

Please complete the online evaluation form - unless you are doing the exercise, in which case please complete it at the end of the exercise.

Where would you like to go now?

• Exercise
• Agenda
• Back to the top

• Author: Blaise Barney, Livermore Computing.

• The OpenMP web site, which includes the C/C++ and Fortran Application Program Interface documents.
  www.openmp.org