an a rchitectural t emplate for parallel loops and
play

An A rchitectural T emplate for Parallel Loops and Sections - PowerPoint PPT Presentation

Oct 09, 2022 •621 likes •821 views

An A rchitectural T emplate for Parallel Loops and Sections Symposium on Software Performance 2018 Markus Frank & Alireza Hakamian The Problem with Palladio and Parallelism Manual Modelling Overhead Performance Inaccurate Model

  1. An A rchitectural T emplate for Parallel Loops and Sections Symposium on Software Performance 2018 Markus Frank & Alireza Hakamian

  2. The Problem with Palladio and Parallelism Manual Modelling Overhead Performance Inaccurate Model Predictions Palladio John Simulator Data Analyst Off by 63 % Me Software BI Performance Performance Prediction System Engineer Results M. Frank and M. Hilbrich. “Performance Prediction for Multicore Environments—An Experiment Report”. In: Proceedings of the Symposium on Software Performance 2016, Kiel, Germany Frank, M., Staude, S. & Hilbrich, M. Is the PCM Ready for ACTORs and Multicore CPUs? - A Use Case-based Evaluation. In Proceedings of the Symposium on Software Performance 2017, Karlsruhe, Germany. M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 2

  3. Idea and Process Common parallelisation patterns as easy to use architectural templates Pattern Pattern A rchitectural Accuracy Identification Modelling T emplate Improvement (Code) (Palladio) Creation M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 3

  4. Pattern Identification Pattern Pattern A rchitectural Accuracy Identification Modelling T emplate Improvement (Code) (Palladio) Creation

  5. Parallel Loop Code example Software Parallelisation Repetitions Resource Demand John Data Analyst M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 5

  6. Pattern Modelling Pattern Pattern A rchitectural Accuracy Identification Modelling T emplate Improvement (Code) (Palladio) Creation

  7. Parallel Loop SEFF Example (Single Core) Software Model << loopAction >> rep = matrixASizeM.VALUE * matrixASizeN.VALUE * matrixBSizeJ.VALUE << InternalAction >> calculationA Resource Demand ResourceDemands 0.00000069 * matrixASizeM.VALUE * << InternalAction >> matrixASizeN.VALUE * calculation Repetitions matrixBSizeJ.VALUE <CPU> ResourceDemands 0.00000069 <CPU> M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 7

  8. Parallel Loop SEFF Example (2 Threads) Software Model << Fork >> ForkedBehaviours << Synchronisation Point >> << InternalAction >> << InternalAction >> Resource Demand calculationA calculationB ResourceDemands ResourceDemands 0.00000069 * matrixASizeM.VALUE * 0.00000069 * matrixASizeM.VALUE * matrixASizeN.VALUE * matrixASizeN.VALUE * Repetitions matrixBSizeJ.VALUE / 2 <CPU> matrixBSizeJ.VALUE / 2 <CPU> Parallelisation M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 8

  9. Parallel Loop SEFF Example (18 Threads) Software Model << Fork >> ForkedBehaviours << Synchronisation Point >> Manual modelling overhead increase << InternalAction >> << InternalAction >> with each thread calculationA calculationB ResourceDemands ResourceDemands 0.00000069 * matrixASizeM.VALUE * 0.00000069 * matrixASizeM.VALUE * matrixASizeN.VALUE * matrixASizeN.VALUE * matrixBSizeJ.VALUE / 2 <CPU> matrixBSizeJ.VALUE / 2 <CPU> M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 9

  10. But, what we want to have… Software Model << loopAction >> @Parallel rep = matrixASizeM.VALUE * matrixASizeN.VALUE * matrixBSizeJ.VALUE Efficiency threadPooleSize = threadNumber Resource Demand << InternalAction >> calculation Repetitions ResourceDemands 0.00000069 <CPU> Parallelisation M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 10

  11. Architectural Templates Pattern Pattern A rchitectural Accuracy Identification Modelling T emplate Improvement (Code) (Palladio) Creation

  12. “Architectural Templates allow software architects to apply reusable patterns to their Palladio models.“ 3-Steps for AT 1 2 Define Stereotype OVT-o Lehrig, Sebastian Michael. Efficiently Conducting Quality-of-Service Analyses by Templating Architectural Knowledge . Vol. 25. KIT Scientific Publishing, 3 Add To Catalog 2018. M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 12

  13. Accuracy Improvement Pattern Pattern A rchitectural Accuracy Identification Modelling T emplate Improvement (Code) (Palladio) Creation

  14. Adding Overhead to Loop AT << loopAction >> Each thread needs additional @Parallel Software Model rep = matrixASizeM.VALUE * matrixASizeN.VALUE * resources to span and matrixBSizeJ.VALUE synchronize threadPooleSize = threadNumber overhead = threadPoolSize * 0.0000079 <CPU> overhead = 0.00078 * threadNumbers Accuracy Additional overhead due to << InternalAction >> communication or waiting calculation condition may occure ResourceDemands 0.00000069 <CPU> M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 14

  15. Adding Overhead to Model << Fork >> ForkedBehaviours << Synchronisation Point >> Software Model << InternalAction >> << InternalAction >> calculationA calculationB ResourceDemands ResourceDemands Accuracy 0.00000069 * matrixASizeM.VALUE * 0.00000069 * matrixASizeM.VALUE * matrixASizeN.VALUE * matrixASizeN.VALUE * matrixBSizeJ.VALUE / 2 <CPU> matrixBSizeJ.VALUE / 2 <CPU> << InternalAction >> overhead ResourceDemands overhead <CPU> M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 15

  16. Limitations & Future Work

  17. Limitations & Future Work Limitations Future Work MPI and ACTORS Equal workload Reference curves for different resource demands Only OpenMP-like Abstraction to architectural level Estimation of Include addition overhead function properties in Simulations M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 17

  18. The Problem with Palladio and Parallelism Manual Modelling Overhead Performance Inaccurate Model Predictions Palladio John Simulator Data Analyst Off by 63 % Me Software BI Performance System Performance Prediction Engineer Results M. Frank - Symposium on Software Performance 2018 - An Architectural Template for Parallel Loops 08.11.18 18

  19. Institut für Softwaretechnologie Reliable Software Systems Thank you! Markus Frank E-Mail markus.frank@informatk.uni-stuttgart.de Telefon +49 (0) 711 685- 88272 iste.uni-stuttgart.de/rss.html www. Universität Stuttgart Reliable Software Systems Universitätsstraße 38, 70569 Stuttgart, Germany

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Controlling Loops in Parallel Mercury Code Paul Bone, Zoltan Somogyi and Peter Schachte National

Controlling Loops in Parallel Mercury Code Paul Bone, Zoltan Somogyi and Peter Schachte National

Controlling Loops in Parallel Mercury Code Controlling Loops in Parallel Mercury Code Paul Bone, Zoltan Somogyi and Peter Schachte National ICT Australia The University of Melbourne Declarative Aspects of Multicore Programming January 28, 2012

517 views • 15 slides
Multithreading 1 A rchitectural State and Context Switches 2 A rchitectural State The A

Multithreading 1 A rchitectural State and Context Switches 2 A rchitectural State The A

Multithreading 1 A rchitectural State and Context Switches 2 A rchitectural State The A rchitectural State of a thread is everything that defines the state of a running program The contents of the register file The current

594 views • 34 slides
Parallel Nested Loops For each tuple s i in S For each tuple t j in T If s i =t j ,

Parallel Nested Loops For each tuple s i in S For each tuple t j in T If s i =t j ,

9/15/2011 Parallel Nested Loops For each tuple s i in S For each tuple t j in T If s i =t j , then add (s i ,t j ) to output Create partitions S 1 , S 2 , T 1 , and T 2 Have processors work on (S 1 ,T 1 ), (S 1 ,T 2 ), (S 2 ,T 1

329 views • 13 slides
Parallel Programming with Thread pools and iterators 1. About me 2. Loops 3. Iterators 4.

Parallel Programming with Thread pools and iterators 1. About me 2. Loops 3. Iterators 4.

Stefan Schindler (@dns2utf8) March 20, 2019 Rust Zrichsee, Schweiz, CH - hosted by Cloud Solutions Amsterdam, NL Parallel Programming with Thread pools and iterators 1. About me 2. Loops 3. Iterators 4. Modes of Execution 5.

612 views • 39 slides
Just tired of endless loops! or parallel : Stata module for parallel computing George G. Vega Yon 1

Just tired of endless loops! or parallel : Stata module for parallel computing George G. Vega Yon 1

Just tired of endless loops! or parallel : Stata module for parallel computing George G. Vega Yon 1 Brian Quistorff 2 1 University of Southern California vegayon@usc.edu 2 Microsoft AI and Research Brian.Quistorff@microsoft.com Stata Conference

857 views • 62 slides
Concurrent Programming with Parallel Extensions to .NET Joe Duffy Architect &amp; Development

Concurrent Programming with Parallel Extensions to .NET Joe Duffy Architect &amp; Development

Concurrent Programming with Parallel Extensions to .NET Joe Duffy Architect &amp; Development Lead Parallel Extensions Talk Outline Overview 5 things about Parallel Extensions 1. Tasks and futures 2. Parallel loops 3. Parallel LINQ

650 views • 36 slides
Tutorial 3 Loops Side Effects 1 CS 136 Spring 2020 Tutorial 3 Loops: for loops &amp;

Tutorial 3 Loops Side Effects 1 CS 136 Spring 2020 Tutorial 3 Loops: for loops &amp;

Tutorial 3 Loops Side Effects 1 CS 136 Spring 2020 Tutorial 3 Loops: for loops &amp; while loops Using a loop to solve a problem is called iteration . while is similar to if statements but while repeatedly loops back and

457 views • 8 slides
Parallel Nested Loops Parallel Partition-Based Create n partitions of S by hashing each

Parallel Nested Loops Parallel Partition-Based Create n partitions of S by hashing each

9/15/2011 Parallel Nested Loops Parallel Partition-Based Create n partitions of S by hashing each S-tuple s, e.g., to For each tuple s i in S bucket number (s mod n) For each tuple t j in T Create n partitions of T in the same way

321 views • 5 slides
Loops! Flow of Control: Loops (Savitch, Chapter 4) TOPICS while Loops do while

Loops! Flow of Control: Loops (Savitch, Chapter 4) TOPICS while Loops do while

Loops! Flow of Control: Loops (Savitch, Chapter 4) TOPICS while Loops do while Loops for Loops break Statement continue Statement CS 160, Fall Semester 2012 2 An Example While Loop

446 views • 11 slides
Shared Memory Programming More about parallel loops LASTPRIVATE clause Sometimes need the

Shared Memory Programming More about parallel loops LASTPRIVATE clause Sometimes need the

Shared Memory Programming More about parallel loops LASTPRIVATE clause Sometimes need the value a private variable would have had on exit from loop (normally undefined). Syntax: Fortran: LASTPRIVATE(list) C/C++: lastprivate(list) Also

146 views • 14 slides
Flexible Hierarchical Execution of Parallel Task Loops Michael Robson, Villanova University

Flexible Hierarchical Execution of Parallel Task Loops Michael Robson, Villanova University

Flexible Hierarchical Execution of Parallel Task Loops Michael Robson, Villanova University Kavitha Chandrasekar, University of Illinois Urbana-Champaign Injection Bandwidth vs CPU speeds Kale (Salishan 2018) 2 Motivation Year Machine

792 views • 30 slides
LOOPS Loops Loops Loops! How can we repeat a piece of code without having to write it out over

LOOPS Loops Loops Loops! How can we repeat a piece of code without having to write it out over

Session 6 LOOPS Loops Loops Loops! How can we repeat a piece of code without having to write it out over and over?! With loops! The Game Loop For Loop vs. While Loop For Loops While Loops Start/End are Known Start/End not Always Clear

301 views • 26 slides
Applying Logistic Regression Model on HPX Parallel Loops Zahra Khatami Lukas Troska Hartmut

Applying Logistic Regression Model on HPX Parallel Loops Zahra Khatami Lukas Troska Hartmut

Motivation HPX HPX Current Challenges Proposed Methods Louisiana State University Experimental Results Conclusion Applying Logistic Regression Model on HPX Parallel Loops Zahra Khatami Lukas Troska Hartmut Kaiser J. Ramanujam Louisiana

926 views • 28 slides
C ONNECTING A SSESSMENT P LANNING WITH D ATA M ANAGEMENT P ROCESSES AND A M IDDLE S TATES T EMPLATE

C ONNECTING A SSESSMENT P LANNING WITH D ATA M ANAGEMENT P ROCESSES AND A M IDDLE S TATES T EMPLATE

C ONNECTING A SSESSMENT P LANNING WITH D ATA M ANAGEMENT P ROCESSES AND A M IDDLE S TATES T EMPLATE Brett Everhart, Ed.D. (beverhar@lhup.edu) Special Assistant to the Provost for Assessment &amp; Accreditation /Director of Teacher Education Lock

955 views • 49 slides
Callisto-RTS: Fine-Grain Parallel Loops Tim Harris (Oracle Labs), Stefan Kaestle (ETH Zurich) ATC

Callisto-RTS: Fine-Grain Parallel Loops Tim Harris (Oracle Labs), Stefan Kaestle (ETH Zurich) ATC

Callisto-RTS: Fine-Grain Parallel Loops Tim Harris (Oracle Labs), Stefan Kaestle (ETH Zurich) ATC 2015 1 / 14 Callisto-RTS OpenMP-like runtime system for (NUMA) shared memory machines Aims to scale better than OpenMP, while requiring

532 views • 15 slides
Building Java Programs Chapter 5 Lecture 11: while Loops, Fencepost Loops, and Sentinel Loops

Building Java Programs Chapter 5 Lecture 11: while Loops, Fencepost Loops, and Sentinel Loops

Building Java Programs Chapter 5 Lecture 11: while Loops, Fencepost Loops, and Sentinel Loops reading: 5.1 5.2 (Slides adapted from Stuart Reges, Hlne Martin, and Marty Stepp) 1 2 Methods using charAt Write a method

795 views • 20 slides
2.11. The Maximum of n Random Variables 3.4. Hypothesis Testing 5.4. Long Repeats of the Same

2.11. The Maximum of n Random Variables 3.4. Hypothesis Testing 5.4. Long Repeats of the Same

2.11. The Maximum of n Random Variables 3.4. Hypothesis Testing 5.4. Long Repeats of the Same Nucleotide Prof. Tesler Math 283 Fall 2018 Prof. Tesler Max of n Variables &amp; Long Repeats Math 283 / Fall 2018 1 / 24 Maximum of two rolls of

483 views • 24 slides
Cardinality of a Set We use three different notations for the number of elements in a finite set:

Cardinality of a Set We use three different notations for the number of elements in a finite set:

Cardinality of a Set We use three different notations for the number of elements in a finite set: Alan H. SteinUniversity of Connecticut Cardinality of a Set We use three different notations for the number of elements in a finite set: n (

1.19k views • 56 slides
Man vs. Machine in Conversational Speech Recognition George Saon IBM Research AI Deep Blue vs.

Man vs. Machine in Conversational Speech Recognition George Saon IBM Research AI Deep Blue vs.

Man vs. Machine in Conversational Speech Recognition George Saon IBM Research AI Deep Blue vs. Garry Kasparov, 1997 Man vs. Machine in Conversational Speech Recognition ASRU 2017, Okinawa AlphaGo vs. Lee Sedol, 2016 Man vs. Machine in

874 views • 44 slides
Lecture 7 Adders and Multipliers 1 11/22/2019 Ripple Carry Adder b3 a3 b2 a2 b1 a1

Lecture 7 Adders and Multipliers 1 11/22/2019 Ripple Carry Adder b3 a3 b2 a2 b1 a1

Lecture 7 Adders and Multipliers 1 11/22/2019 Ripple Carry Adder b3 a3 b2 a2 b1 a1 a0 cin (c0) b0 cout FA3 FA2 FA1 FA0 (c4) M4 M3 M2 M1 c1 c2 c3 s3 s2 s1 s0 Key observations, the value of the A B cin cout 0

279 views • 17 slides
Interactive Proofs Lecture 18 AM 1 Interactive Proofs 2 Interactive Proofs IP[k] 2

Interactive Proofs Lecture 18 AM 1 Interactive Proofs 2 Interactive Proofs IP[k] 2

Interactive Proofs Lecture 18 AM 1 Interactive Proofs 2 Interactive Proofs IP[k] 2 Interactive Proofs IP[k] IP[poly] = PSPACE 2 Interactive Proofs IP[k] IP[poly] = PSPACE IP protocol for TQBF using arithmetization 2 Interactive

1.75k views • 136 slides
Telepathwords Preventing Weak Passwords by Reading Users Minds Authors: Saranga Komanduri,

Telepathwords Preventing Weak Passwords by Reading Users Minds Authors: Saranga Komanduri,

Catalina Cumpanasoiu Erin Gibbons Telepathwords Preventing Weak Passwords by Reading Users Minds Authors: Saranga Komanduri, Richard Shay, and Lorrie Faith Cranor (Carnegie Mellon University) Cormac Herley and Stuart Schechter (Microsoft

504 views • 21 slides
JUST THE MATHS SLIDES NUMBER 2.1 SERIES 1 (Elementary progressions and series) by

JUST THE MATHS SLIDES NUMBER 2.1 SERIES 1 (Elementary progressions and series) by

JUST THE MATHS SLIDES NUMBER 2.1 SERIES 1 (Elementary progressions and series) by A.J.Hobson 2.1.1 Arithmetic progressions 2.1.2 Arithmetic series 2.1.3 Geometric progressions 2.1.4 Geometric series 2.1.5 More general progressions

355 views • 17 slides
Automatic Algorithm Configuration based on Local Search Frank Hutter 1 Holger Hoos 1 utzle 2

Automatic Algorithm Configuration based on Local Search Frank Hutter 1 Holger Hoos 1 utzle 2

Automatic Algorithm Configuration based on Local Search Frank Hutter 1 Holger Hoos 1 utzle 2 Thomas St 1 Department of Computer Science University of British Columbia Canada 2 IRIDIA Universit e Libre de Bruxelles Belgium Real-world

835 views • 59 slides

More recommend