FACULTY OF COMPUTING, ENGINEERING and SCIENCE Final mark awarded:_____ Assessment Cover Sheet and Feedback Form 2016/17 Module Code: CS3S601 Module Title: Concurrent and Parallel Programming Module Lecturer: Marius Miknis Assessment Title and Tasks: Image Processing in TBB Assessment No. 2 of 2 No. of pages submitted in total including this page: Word Count of submission: Date Set: 9/02/17 Submission Date: 17/03/17 Return Date: 14/04/17 Part A: Record of Submission (to be completed by Student) Extenuating Circumstances If there are any exceptional circumstances that may have affected your ability to undertake or submit this assignment, make sure you contact the Advice Zone on your campus prior to your submission deadline. Fit to sit policy: The University operates a fit to sit policy whereby you, in submitting or presenting yourself for an assessment, are declaring that you are fit to sit the assessment. You cannot subsequently claim that your performance in this assessment was affected by extenuating factors. Plagiarism and Unfair Practice Declaration: By submitting this assessment, you declare that it is your own work and that the sources of information and material you have used (including the internet) have been fully identified and properly acknowledged as required1. Additionally, the work presented has not been submitted for any other assessment. You also understand that the Faculty reserves the right to investigate allegations of plagiarism or unfair practice which, if proven, could result in a fail in this assessment and may affect your progress. Details of Submission: Note that all work handed in after the submission date and within 5 working days will be capped at 40%2. No marks will be awarded if the assessment is submitted after the late submission date unless extenuating circumstances are applied for and accepted (Advice Zone to be consulted). Work should be submitted as detailed in your student handbook. You are responsible for checking the method of submission. You are required to acknowledge that you have read the above statements by writing your student number (s) in the box: Student Number(s): 1 University Academic Integrity Regulations 2 Information on exclusions to this rule is available from Campus Advice ShopsIT IS YOUR RESPONSIBILITY TO KEEP A RECORD OF ALL WORK SUBMITTED Part B: Marking and Assessment (to be completed by Module Lecturer) This assignment will be marked out of 100% This assignment contributes to 35% of the total module marks. This assignment is bonded Assessment Task: You are required to apply the techniques discussed in lectures and tutorials to implement several image processing operations using Intel's Threading Building Blocks (TBB) library. Use RGB_Processing.zip provided project as a starting point. The tasks are broken down as follows: 1) Part1: Using the stencil pattern, implement a TBB kernel to blur an image using a Gaussian kernel. Note: You can use the example load function given in lectures and tutorials to load images (greyscale): a. The code should be written both in a sequential and parallel approach. Code should be made parallel using the TBB parallel_for structure. b. The speed differences between the sequential and parallel approaches should be investigated and documented, this should also include comparisons to investigate whether different step sizes in the parallel_for make a difference to performance. c. Blurred image should be saved (grey_blurred.png). 2) Part 2: Update the current image loading code from splitting the RGB values in a sequential approach to a parallel method: a. The code needs to load two colour images and compute an image with the values of the absolute differences between the two input images �(�, �) = |�1(�, �) − �2(�, �)| b. With an image of absolute differences computed a binary threshold operation should be applied to the image that makes any pixels that are not black into white. c. The resulting thresholded image of absolute differences should be saved (RGB_processed.png). 3) You will need to write a short section (500 words) in your report of how the image blurring code in Part 1 could be implemented using OpenCL. (A full OpenCL implementation is not required but code snippets could be included to aid your discussion) 4) You will also be required to explain your implementation in a short 5-minute code demo which will take place in the tutorial sessions after the assignment has been submitted. As part of the code demo you will be required to discuss the techniques used to implement your TBB solution and the results you obtained. The code demo is mandatory. The above sections will also be marked according to how well you demonstrate your understanding of them in the code demo. If you fail to do code demo this will result if failed coursework submission. You may use the tutorial or lecture demo code as a starting point for your implementation. If you wish, you can also implement your solution in the OS of your choice (OS X or Windows) if you don't want to use Linux. If you choose not to implement the assignment in Linux please be aware that it is your responsibility to demonstrate the code in the code demo sessions (on your own laptop for example) and it is your responsibility to make sure that the TBB and OpenCL support is present. Also be aware of version differences as the examples discussed in lectures may be based on a different version of OpenCL (we're using version 1.1) than on the platform you choose for the assignment.Deliverables 1) A zip file containing the source code and executable of your implementation. This is to be submitted to Blackboard no later than the submission date shown on the assignment front sheet. Please name your zip file CS3S601_CW2_EnrolmentNumber_FirstName_LastName.zip (replace the placeholders with your enrolment number, first and last name respectively) 2) A word processed report documenting your test plan, results and discussion. This is to be included in your zip file. 3) A copy of this document is to be included in your zip file, with your Student Enrolment Number filled in on the front sheet and the optional Reflection sheet (see Part C below) filled in accordingly. 4) A 5-minute code demo discussing your implementation, the results obtained and the problems you faced in implementing the assignment. Learning Outcomes to be assessed (as specified in the validated module descriptor https://icis.southwales.ac.uk/): LO1. To critically analyse the advantages and problems inherent in concurrent and parallel programming. LO2. To be able to implement and evaluate the performance of an application using concurrent and parallel programming techniques.Part C: Reflections on Assessment (to be completed by student – optional) Use of previous feedback: In this assessment, I have taken/took note of the following points in feedback on previous work: Please indicate which of the following you feel/felt applies/applied to your submitted work  A reasonable attempt. I could have developed some of the sections further.  A good attempt, displaying my understanding and learning, with analysis in some parts.  A very good attempt. The work demonstrates my clear understanding of the learning supported by relevant literature and scholarly work with good analysis and evaluation.  An excellent attempt, with clear application of literature and scholarly work, demonstrating significant analysis and evaluation. What I found most difficult about this assessment: The areas where I would value/would have valued feedback:Marking grid: Fail Narrow Fail 3rd Class / Pass Lower 2nd Class / Pass Upper 2nd Class / Merit 1st Class / Distinction Gaussian Kernel Blur 10% Gaussian blur not implemented Blur computed incorrectly or not using Gaussian kernel Gaussian blur implemented but with minor issues or artefacts Good implementation Gaussian blur Gaussian blur implemented well with a small kernel Gaussian blur implemented excellently with a large kernel Image subtraction 10% Image subtraction not performed Incorrect absolute difference formula was implemented Image subtraction performed incorrectly Image subtraction was performed with minor issues Image subtraction was performed well Image subtraction was performed very well Image subtraction was performed excellently Correct absolute difference formula was implemented Binary threshold 5% Binary threshold not performed Poor binary threshold Binary threshold was performed with minor issues Binary threshold was performed well Binary threshold was performed very well Binary threshold was performed excellently with a well-adjusted threshold value Final processed colour image 10% Processed colour image was not generated Processed colour image was generated incorrectly Processed colour image was generated with significant artefacts Processed colour image was generated with some artefacts Processed colour image was generated without artefacts Processed colour image was generated with clear object areas TBB and parallelisation 25% Code was not parallelised No lambda functions used Incorrect use of code parallelisation Incorrect use of lambda functions Some parts of the code were parallelised Some lambdas implemented with minor issues Most parts of the code were parallelised well Code correctly implements each lambda All parts of the code were parallelised well Code correctly implements each lambda and correctly separates lambda parameters from captured variables All possible parts of the code were parallelised very well Excellent use of lambda functions with correctly separated lambda parameters from captured variables Report 20% Very poor Report Possible OpenCL implementation not mentioned Poor Report Possible OpenCL implementation mentioned poorly Report is satisfactory with basic comparison of sequential and parallel code performance Possible OpenCL implementation discussed Report is good with a satisfactory comparison of sequential and parallel code performance Possible OpenCL implementation discussed in detail Very good report that provides a good comparison of sequential and parallel code performance Possible OpenCL implementation discussed in good detail but no examples provided Very detailed report that showed a great comparison of sequential and parallel code performance Possible OpenCL implementation discussed in high detail and examples provided Demonstration 20% No Demonstration Does not understand the code Has a simple knowledge of the code Has a reasonable knowledge of the code Has a good understanding of the code Has an excellent understanding of the code