Strategies for Developing Visualizations of Parallel Algorithms for

  • Slides: 12
Download presentation
Strategies for Developing Visualizations of Parallel Algorithms for Dense Matrix Operations on Interconnected SIMD

Strategies for Developing Visualizations of Parallel Algorithms for Dense Matrix Operations on Interconnected SIMD Computers by : Justin C. Miller 1/12

Who am I ? Justin Miller Computer Science Major Mathematics Minor Goals: Receive Ph.

Who am I ? Justin Miller Computer Science Major Mathematics Minor Goals: Receive Ph. D Become College Professor 2/12

Why’d I choose this topic? Strategies for Developing Visualizations of Parallel Algorithms for Dense

Why’d I choose this topic? Strategies for Developing Visualizations of Parallel Algorithms for Dense Matrix Operations on Interconnected SIMD Computers Wanted to combine topics… 1. Computer Science 2. Mathematics 3. Education/Teaching 3/12

Outline of Talk 1. Teach you the General Concepts needed to understand what I’ve

Outline of Talk 1. Teach you the General Concepts needed to understand what I’ve done. 2. Quick overlook of what I’ve done. 4/12

Matrices and Matrix Operations What is a matrix? What is a dense matrix? (as

Matrices and Matrix Operations What is a matrix? What is a dense matrix? (as opposed to sparse) What are matrix operations? eg. (multiplication, transposition) 5/12

Performing matrix operations on Parallel Computers The concept of parallel computing. What does it

Performing matrix operations on Parallel Computers The concept of parallel computing. What does it mean to be SIMD? Subclasses of SIMD computers. eg. (shared memory, interconnected) Types of Interconnected Networks. 6/12

How did I make the visualizations? 2 Step Process to animate the algorithms… 1.

How did I make the visualizations? 2 Step Process to animate the algorithms… 1. Understand it using JCSP library public void run(){ int data = get. Starting. Data() ; while(counter < 10){ To. Left. pass(data) 2. Animate it using Animal ; data = From. Right. get(); square “square 1” (10, 10) 40 color blue calculate(data) ; }//while text “title” “Justin’s Demo”write. Final. Answer(data) (40, 40) centered color black font; Serif size 14 bold }//run 7/12

Why design visualizations for these algorithms? Lots of motion & passing of data Difficult

Why design visualizations for these algorithms? Lots of motion & passing of data Difficult to draw, lots of mistakes Quicker preparation time for lectures Attracts student attention 8/12

Actually in Strategies for developing my Thesis animations for these algorithms. paper. Some general

Actually in Strategies for developing my Thesis animations for these algorithms. paper. Some general techniques mentioned in othe papers. eg. Quiz Questions, Descriptive Text, High Contrast, Step-Thru & Play Several specific techniques for parallel matrix algorithms on Interconnected SIMD computers. eg. show channels, pass data along channels, show register contents Example: abstract delays by making certain events appear exactly in parallel. 9/12

Case Study General results of prior research. Had to see it to believe it.

Case Study General results of prior research. Had to see it to believe it. Mostly for feedback. 10/12

Summation Like to idea of visualizations in the classroom. Future possibility of conducting a

Summation Like to idea of visualizations in the classroom. Future possibility of conducting a more controlled scientific study Most importantly, preparation for graduate research Thank you, Questions? 11/12

Further interest? Here’s some free downloads. Download Java http: //java. sun. com/j 2 se/1.

Further interest? Here’s some free downloads. Download Java http: //java. sun. com/j 2 se/1. 4/download. html Download JCSP library http: //www. cs. ukc. ac. uk/projects/ofa/jcsp/ Download Animal http: //www. animal. ahrgr. de/ 12/12