Installation Wizards

GPU610/DPS915 | Student List | Group and Project Index | Student Resources | Glossary

Installation Wizards

Team Members

1. Matthew Bell
2. Kevin Ramsamujh

Email All

Image Processing

Much like previous students who have taken this course, we decided to look over image processing (leveraging Christopher Ginac's PGM parser http://www.dreamincode.net/forums/topic/76816-image-processing-tutorial/ ).

For our profiling we scale an image up 10 times to better profile larger data sets, as our source image was only ~800kb in size. We then negate the image, and then reflect it.

The time taken to perform all 3 operations:

real    0m22.057s
user    0m4.584s
sys     0m14.221s

The results of our tests:

Flat profile:
Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total
 time   seconds   seconds    calls   s/call   s/call  name
 42.92      6.70     6.70        3     2.23     2.23  Image::operator=(Image const&)
 29.15     11.25     4.55        2     2.27     2.27  Image::Image(int, int, int)
 15.31     13.64     2.39        1     2.39     2.39  Image::Image(Image const&)
  4.04     14.27     0.63                             writeImage(char*, Image&)
  2.56     14.67     0.40                             Image::negateImage(Image&)
  2.37     15.04     0.37 86832000     0.00     0.00  Image::getPixelVal(int, int)
  2.18     15.38     0.34                             Image::reflectImage(bool, Image&)
  1.47     15.61     0.23                             Image::enlargeImage(int, Image&)
  0.00     15.61     0.00   868320     0.00     0.00  Image::setPixelVal(int, int, int)
  0.00     15.61     0.00        3     0.00     0.00  Image::~Image()
  0.00     15.61     0.00        1     0.00     0.00  _GLOBAL__sub_I__ZN5ImageC2Ev
  0.00     15.61     0.00        1     0.00     0.00  _GLOBAL__sub_I_main
  0.00     15.61     0.00        1     0.00     0.00  Image::getImageInfo(int&, int&, int&)

As we can see, negating the image as well as reflecting and enlarging it are all costly operations, and we believe that we can optimize it to take far less time.

Sudoku Brute Force Solver

For this first assignment I decided to look into a brute force sudoku solver to see if it could benefit from parallel programming. The sudoku solver that was investigated was provided by Bryan Smith (https://github.com/bryanesmith/Sudoku-solver).

As is, this solver is quick to solve a 9x9 sudoku puzzle. The time taken to solve a sudoku puzzle is under a second as seen in our time results:

real    0m0.097s
user    0m0.072s
sys     0m0.016s

However if this we were to scale this solver up to solve nxn sized puzzles we would be able to see a use for optimizing this code.

Results from profiling:

Flat profile:
Each sample counts as 0.01 seconds.
  %   cumulative   self              self     total
 time   seconds   seconds    calls  ms/call  ms/call  name
 60.00      0.03     0.03   468316     0.00     0.00  SudokuPuzzle::verifyValue(int, int)
 20.00      0.04     0.01   468316     0.00     0.00  SudokuPuzzle::printTracerTryingValue(int, int)
 20.00      0.05     0.01        1    10.00    50.00  SudokuPuzzle::solve(int, int)
  0.00      0.05     0.00   445799     0.00     0.00  SudokuPuzzle::setBoardValue(int, int, int)
  0.00      0.05     0.00        2     0.00     0.00  SudokuPuzzle::print()
  0.00      0.05     0.00        1     0.00     0.00  _GLOBAL__sub_I__ZN12SudokuPuzzleC2Ev
  0.00      0.05     0.00        1     0.00     0.00  _GLOBAL__sub_I_main
  0.00      0.05     0.00        1     0.00     0.00  __static_initialization_and_destruction_0(int, int)
  0.00      0.05     0.00        1     0.00     0.00  __static_initialization_and_destruction_0(int, int)
  0.00      0.05     0.00        1     0.00    50.00  SudokuPuzzle::solve()
  0.00      0.05     0.00        1     0.00     0.00  SudokuPuzzle::SudokuPuzzle() 

The profiling of this solver shows that the verifyValue function is taking up most of the execution time at 60% and would benefit from optimizing.