OpenMP - Revision history

Tim-cramer-c1e6@rwth-aachen.de at 12:27, 6 November 2020

2020-11-06T12:27:34Z

Paul-kapinos-e26d@rwth-aachen.de: /* Teaching material */

2020-05-04T11:36:24Z

Teaching material

Paul-kapinos-e26d@rwth-aachen.de: /* Books */

2020-05-04T11:08:33Z

Books

Paul-kapinos-e26d@rwth-aachen.de: /* Videos */

2020-05-04T08:40:12Z

Videos

Paul-kapinos-e26d@rwth-aachen.de: /* Materials */

2020-05-04T08:38:32Z

Materials

Paul-kapinos-e26d@rwth-aachen.de: /* Books */

2020-05-04T08:37:55Z

Books

Paul-kapinos-e26d@rwth-aachen.de: /* Materials */

2020-05-04T08:25:29Z

Materials

Paul-kapinos-e26d@rwth-aachen.de: /* Videos */

2020-05-04T08:00:03Z

Videos

Paul-kapinos-e26d@rwth-aachen.de at 06:57, 4 May 2020

2020-05-04T06:57:56Z

Paul-kapinos-e26d@rwth-aachen.de at 09:09, 22 April 2020

2020-04-22T09:09:13Z

@@ Line 333: / Line 333: @@
 == Teaching material ==
 * [https://doc.itc.rwth-aachen.de/download/attachments/35947076/01_IntroductionToOpenMP.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 1: Introduction]
 * [https://doc.itc.rwth-aachen.de/download/attachments/35947076/02_OpenMPTaskingInDepth.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 2: Tasking in Depth]
@@ Line 343: / Line 347: @@
 * [https://www.springer.com/gp/book/9783540731221 S. Hoffmann, R.Lienhart: “OpenMP - Eine Einführung in die parallele Programmierung mit C/C++”, Springer, 2008. ISBN-13 978-3540731238]
 * [https://www.openmp.org/specifications/ OpenMP Specifications]

@@ Line 332: / Line 332: @@
 Please check the more detailed tutorials in the References.
-== Materials ==
+== Teaching material ==
 * [https://doc.itc.rwth-aachen.de/download/attachments/35947076/01_IntroductionToOpenMP.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 1: Introduction]
 * [https://doc.itc.rwth-aachen.de/download/attachments/35947076/02_OpenMPTaskingInDepth.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 2: Tasking in Depth]

@@ Line 339: / Line 339: @@
 == Books ==
-* [https://mitpress.mit.edu/books/using-openmp R. van der Pas, B. Chapman, G. Jost: “Using OpenMP”, MIT Press. ISBN-10: 0-262-53302-2 / ISBN-13: 978-0-262-53302-7.]
+* [https://mitpress.mit.edu/books/using-openmp R. van der Pas, B. Chapman, G. Jost: “Using OpenMP”, MIT Press. ISBN-10: 0262533022 / ISBN-13: 978-0262533027]
-* R. Chandra, L. Dagum, D. Kohr, D. Maydan, J. McDonald, R.Menon: “Parallel Programming in OpenMP”, Morgan Kaufmann Publishers, 2001.  ISBN-10: 1558606718 / ISBN-13: 978-1558606715
+* [https://mitpress.mit.edu/books/using-openmp-next-step R. van der Pas, E. Stotzer and C. Terboven: “Using OpenMP - The Next Step”, MIT Press. ISBN: 9780-262534789]
-* [https://www.springer.com/gp/book/9783540731221 S. Hoffmann, R.Lienhart: “OpenMP - Eine Einführung in die parallele Programmierung mit C/C++”, Springer, 2008. ISBN-13 978-3-540-73123-8]
+* [https://www.springer.com/gp/book/9783540731221 S. Hoffmann, R.Lienhart: “OpenMP - Eine Einführung in die parallele Programmierung mit C/C++”, Springer, 2008. ISBN-13 978-3540731238]
 * [https://www.openmp.org/specifications/ OpenMP Specifications]

@@ Line 345: / Line 345: @@
 == Videos ==
+* Introduction to OpenMP from PPCES 2014, part 1
 <youtube>6FMn7M5jxrM</youtube>
+* Introduction to OpenMP from PPCES 2014, part 2
 <youtube>Whq28OaPW08</youtube>
+* Getting OpenMP Up To Speed, from PPCES 2014, by Ruud van der Pas
 <youtube>u6N33FhfgZU</youtube>

@@ Line 333: / Line 333: @@
 == Materials ==
- [https://doc.itc.rwth-aachen.de/download/attachments/35947076/01_IntroductionToOpenMP.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 1: Introduction]
+* [https://doc.itc.rwth-aachen.de/download/attachments/35947076/01_IntroductionToOpenMP.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 1: Introduction]
- [https://doc.itc.rwth-aachen.de/download/attachments/35947076/02_OpenMPTaskingInDepth.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 2: Tasking in Depth]
+* [https://doc.itc.rwth-aachen.de/download/attachments/35947076/02_OpenMPTaskingInDepth.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 2: Tasking in Depth]
- [https://doc.itc.rwth-aachen.de/download/attachments/35947076/03_OpenMPNumaSimd.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 3: NUMA & SIMD]
+* [https://doc.itc.rwth-aachen.de/download/attachments/35947076/03_OpenMPNumaSimd.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 3: NUMA & SIMD]
- [https://doc.itc.rwth-aachen.de/download/attachments/35947076/04_OpenMPSummary.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 4: Summary]
+* [https://doc.itc.rwth-aachen.de/download/attachments/35947076/04_OpenMPSummary.pdf Introduction to OpenMP from PPCES (@RWTH Aachen) Part 4: Summary]
 == Books ==

@@ Line 1: / Line 1: @@
 [[Category:HPC-Developer]]
-OpenMP is an open standard API for Shared Memory [[Parallel_Programming|parallelization]] in C, C++ and Fortran which consist of compiler dircetives, runtime routines and environment variables.
+[https://www.openmp.org OpenMP] is an open standard API for Shared Memory [[Parallel_Programming|parallelization]] in C, C++ and Fortran which consist of compiler dircetives, runtime routines and environment variables.
-The first version was published for Fortran in 1997 with the C/C++ standard following a year later. In 2000, OpenMP 2.0 was introduced. For the first two versions the main focus had been on parallelizing highly regular loops, which were considered the main performance limitation for scientific programs. OpenMP 3.0 introduced the concept of tasking in 2008 while OpenMP 4.0 added support for accelerators in 2013. The current version OpenMP 5.0 was released only in November 2018 and is therefore not yet fully supported by all compilers and OSes.
+The first version was published for Fortran in 1997 with the C/C++ standard following a year later. Version 2.0 was available in 2000 for Fortran and C/C++ in 2002, respectively. The version 2.5 from 2005 was the first combined specification for both Fortran and C/C++ . By then the main focus had been on parallelizing highly regular loops, which were considered the main performance hotspot for scientific programs. OpenMP 3.0 introduced the concept of tasking in 2008 while OpenMP 4.0 added support for accelerators in 2013. The current version OpenMP 5.0 was released in November 2018. As the implementation of sophisticated features can be cumbersome, not all achievements must be fully supported by all compilers and OSes.
 Information on how to run an existing OpenMP program can be found in the "[[How to Use OpenMP]]"-Section.
@@ Line 13: / Line 13: @@
 [[File:OpenMPSharedMem.png|400px]]
-A program is launched with one single thread as a master thread. When entering a parallel region, worker threads are spawned and form a team with the master. Upon leaving the parallel region these worker threads are tpically put to sleep (instead of actual joining) by the OpenMP runtime until the program reaches its next parallel region. This concept is known as Fork-Join and it allows incremental [[Parallel_Programming|parallelization]]. As a result all parallel regions must be defined explicitly. Furthermore there may only be exactly one entry point at the top of the region and exactly one exit point at the bottom and Branching out is not allowed. Termination, however, is allowed. The number of work threads to be spawned is controlled by internal control variable, which value can be defined explicitely at the entry point. If this is not specified, OpenMP runtime will use the environment variable <code>OMP_NUM_THREADS</code>, and if not set choose a number of threads according implementation-defined rules (often: 'use all cores').
+A program is launched with one single thread as a master thread. When entering a parallel region, worker threads are spawned and form a team with the master. Upon leaving the parallel region these worker threads are typically put to sleep (instead of actual joining) by the OpenMP runtime until the program reaches its next parallel region. This Fork-Join concept allows incremental [[Parallel_Programming|parallelization]]: starting from serial code, the hot spots are parallelised one by one (typically still allowing serial compilation and execution). As a result all parallel regions must be defined explicitly. Furthermore there may only be exactly one entry point at the top of the region and exactly one exit point at the bottom and Branching out is not allowed. Termination, however, is allowed. The number of work threads to be spawned is controlled by an internal control variable, which value can be defined explicitely at the entry point. If this is not specified, OpenMP runtime will use the environment variable <code>OMP_NUM_THREADS</code>, and if it is not set choose a number of threads according implementation-defined rules (often: 'use all cores').
 [[File:OpenMPForkJoin.png|300px]]
@@ Line 81: / Line 81: @@
-However, not all loops can be parallelized this way. In order to split the workload as shown there must not be any dependencies between the iterations, i.e. no thread may write to a memory location that another thread reads from.  If in doubt, check if a backward exectuion of the loop would have the same result. Pleas not that this is only a simple test that works for most cases, but not all!
+However, not all loops can be parallelized this way. In order to split the workload as shown there must not be any dependencies between the iterations, i.e. no thread may write to a memory location that another thread reads from.  If in doubt, check if a backward exectuion of the loop would have the same result. Pleas note that this simple test is only a necessary condition that works for most cases, but not a sufficient condition!
 === Data Scoping ===
-In OpenMP, variables are divided into shared and private. This can be set when creating a parallel region or ''for''-construct. The default for parallel regions is shared while all variables created within a for-construct are private. Static variables are also shared. The following options are available:
+In OpenMP, variables are divided into shared and private. This can be set when creating a parallel region or ''for''-construct. The default for parallel regions is shared for all already-existing variables while all variables created within a parallel region are private. Static variables are also shared. The following options are available:
 * <code>shared(VariableList)</code>: one shared instance that can be read from and written to by all threads