The pfunc implementation of nas parallel benchmarks
Download
1 / 8

The PFunc Implementation of NAS Parallel Benchmarks. - PowerPoint PPT Presentation


  • 74 Views
  • Uploaded on

The PFunc Implementation of NAS Parallel Benchmarks. Presenter: Shashi Kumar Nanjaiah Advisor: Dr. Chung E Wang Department of Computer Science California State University, Sacramento. Overview.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' The PFunc Implementation of NAS Parallel Benchmarks.' - susane


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
The pfunc implementation of nas parallel benchmarks

The PFunc Implementation of NAS Parallel Benchmarks.

Presenter: Shashi Kumar Nanjaiah

Advisor: Dr. Chung E Wang

Department of Computer Science

California State University, Sacramento


Overview
Overview.

The goal of this project is to prove the efficacy of task parallelism in PFunc to parallelize industry-standard benchmark computation kernels and applications on shared-memory.

  • Introduce PFunc, a new tool for task parallelism.

    • New features and extensions.

    • Fibonacci example.

  • Introduce NAS parallel benchmarks.

    • Briefly explain the 7 benchmarks.


Background pfunc a new tool for task parallelism
BackgroundPFunc - A new tool for task parallelism.

  • Extends existing task parallel feature-set.

    • Cilk, Threading Building Blocks, Fortran M, etc.

  • Portable.

    • Linux, OS X, AIX and Windows.

  • Customizable.

    • Generic and generic programming techniques.

      • No runtime penalty.

  • C and C++ APIs.

  • Released under Eclipse Public License v1.0.

    • http://coin-or.org/projects/PFunc.xml


Example parallelizing fibonacci numbers
Example: Parallelizing Fibonacci numbers.

typedef struct {int n; int fib_n;} fib_t;

void fibonacci (void arg) {

fib_t* fib_arg = (fib_t*) arg; if (0 == fib_arg-> n || 1 == fib_arg-> n{

fib_arg-> fib_n = fib_arg-> n;

}else{ pfunc_cilk_task_t fib_task; fib_t fib_n_1 = {(fib_arg-> n) - 1, 0}; fib_t fib_n_2 = {(fib_arg-> n) – 2, 0}; pfunc_cilk_task_init (&fib_task); pfunc_cilk_spawn_c (fib_task, /* Handle to the task* / NULL, /* Attribute -- use default */ NULL, /* Group -- use default */ fibonacci, /* Function to execute */ &fib_n_1); /* Argument */ fibonacci (&fib_n_2); pfunc_cilk_wait (fib_task); pfunc_cilk_task_clear (&fib_task); fib_arg-> fib_n = fib_n_1.fib_n + fib_n_2.fib_n;

}}


Fibonacci task creation overhead
Fibonacci: task creation overhead.

Fibonacci number 37 (236 ≈ 69 billion tasks).

  • 2x faster than TBB!

  • Only 2x slower than Cilk.

    • But provides more flexibility!

      • Fibonacci is the worst case behavior.

    • Library-based rather than a custom compiler.


Nas parallel benchmarks
NAS Parallel Benchmarks.

  • Stands for NASA Advanced Supercomputing.

  • Help to evaluate performance of parallel tools and machines.

  • Consist of 5 kernels and 3 pseudo applications.

    • Taken mostly from Computational Fluid Dynamics (CFD).

  • Originally written in Fortran, but C versions are available.

  • http://www.nas.nasa.gov/Resources/Software/npb.html

  • NPB OpenMP-C v2.3.

    • Base code taken from Omni group’s implementation.



Conclusion
Conclusion.

  • Modify data-parallel NPB OpenMP-C version to task parallel version.

  • Compare against original NPB OpenMP-C version.

    • For problem sizes in classes A, B and C.


ad