Objectives

The objective of this homework is to write four OpenMP programs, to debug and test them, and to experimentally determine the gains you get in running it in parallel on up to 8 (or possibly 16) processors on carver. The parallelizations are relatively simple, and the results should be pleasing in terms of performance/speedups.

As always, you should measure the performance of your parallelization as a function of the number of processors and the problem size, and analyze, plot and discuss your observations. For these programs, you may use the department machines for debugging and program development, but use carver for the final tests and data collection. Remember the environment/compiler options and/or commands may be different for each setup. However, it must be possible to switch back and forth simply by selecting an option to make.

1. Mandelbrot

2. Jacobi Stencil 1D

3. Jacobi Stencil 2D

4. Matrix-vector Product

The objective of is to write an OpenMP program for the matrix-vector product. The sequential program for the Matrix-vector Product algorithm is provided below. You do not need to analyze this program for your report.

Provided Code

Compile OpenMP Program

We are using gcc on both lab machine and carver. On the department lab machines, OpenMP is part of gcc, but you need to enable the OpenMP flag to compile, the flag for OpenMP is "fopenmp".

Questions to discuss in your report

The main question we want you to address in the report is how the performance (execution time) changes as a function o the number of processors. More specific details (e.g., input parameters and program size values to be used for the data) will be provided soon.

What/When:

• We should be able to compile your programs by simply typing make with the makefile you provided. It should produce executables jacobi_1D, jacobi_2D, mandelbrot, and mat_vec respectively.
• Please specify your thread number in the environment instead of inside the program. The input usage for each program should be
  • Jacobi_1D: jacobi_1D < size > < iteration >
  • Jacobi_2D: jacobi_2D < size > < iteration >
  • Mandelbrot: mandelbrot < iteration >
  • mat_vec: mat_vec < N > < M >
• The default output should be the same as the output for the sequential code when not in debug mode.
• In the report you need to include your performance for mandelbrot and jacobi_2D. You should test the thread number from 1 to 8 for each program, we also provide the test size as below:
  • Mandelbrot: ./madelbrot 10000
  • Jacobi_2D: ./jacobi_2D 3000 2000
  If you got some problem size gave you better performance, please use your problem size and declare it in your report.
• 11:59 pm, Saturday, Sep 10. OpenMP source files and Makefiles to be submitted. A detailed description is given above. Please read carefully and make sure that your programs match the description.
• 11:59 pm, Sunday, Sep 11. Lab report to be checked in, as well as data and scripts. You only need to report on Jacobi 2D and Mandelbrot.

How to Submit:

• You must submit the file from one of the Linux machines, such as crestone or dmx.
• All assignments are submitted electronically via the ~cs475/bin/checkin program.
• You must submit a single tar file containg all of your source code and your Makefile.
• You may name the file anything you wish.
• You may resubmit the file as many times as you like. Subsequent submissions will not overwrite previous submissions.
• If you have more than one file for your program, then you should provide a makefile that will compile all executables. Any assignment submitted that contains code that will not compile will receive a 0%.
• The checkin program should be used as follows:
  ~cs475/bin/checkin HW1 filename.tar
  

• Enter the directory that contains the file or directory you want to submit.
• Use the command:

   tar czvf HW1.tgz [file1] [file2] [file3] ... 

• Use the command:

   man tar

  for more information.

Good luck and have fun!