Weekly Report REU 2006
Got familiar with STAPL. Read the ppopp05 and ppl to get things started. Then learnt STL ,learnt how to code in STL. Got to know how learners function, as to
how they maintain a decision tree model and how they choose the best algorithm dynamically depending on the input and the no. of processors .Read about RMIs, pContainers,views and prange and their implementations ,also tried running some test algorithms on them.
Coded matrix multiplication in a parallel way ,but not using pcontainers ,only local arrays and async and sync rmis. The good thing about it was purely asynchronous. Also got to know about the limitations of shipping the data from one processor to another but a wrapper for the std:vector and std:list were already implemented which helped in making my life easier. Then i was assigned a task to make test modules for crashing the RMI model in STAPL, checking their intensity (no of calls and passing data structure as an argument). Made 3 test files which checked its consistency and nestedness. Read about PAPI ,a library which aims to provide the tool designer and application engineer with a consistent interface and methodology for use of the performance counter hardware found in most major microprocessors. Ran some test files and got familiar with it. Attended gabriel's proposal on pContainers .
Picked up Palgorithms , understood the test files and checked their consistency. Also went through the header files and got an idea of how to put in task dependencies and applying workfunctions to views. Dr. Rauchwerger advised me to work on graph algorithms as there was a lot of job to be done there and also i was very much interested in impleneting them. Picked up CLR , got through most of the standard algorithms and thought of a procedure to implement each of them in parallel. Couldn't do much about DFS and all the derived algoriths from DFS like SCC,Topo sort,etc. Met my mentor(Chids) by the end of the week and then felt like i would be assigned a real task finally! Went to the Socail security office at Bryan to apply for my SSN. Had a meeting with Dr Amato's group on Friday ,where i was introduced with the concept of graph generation and stencils. They also mentioned graph coloring in the end..and thought of 4 ways to do it in parallel over the weekend....but..
..on Monday i was informed that coloring is not yet supported in STAPL. Attended the welcome breakfast for USRG,REU guys at Zachry. Met Harsh ,who according to Dr. Amato, is my undergraduate mentor. We had regular meetings at 9o'clock in the morning.We were assigned a task to think about the design of the base stencil class and a module for the random structured graph generation.It seemed to be very trivial in the beginning but as days or rather hours passed by i realised that its quite a CHALLENGE! Thought of numerous ways to implement it ,discussed it with Dr.Amato, Roger,Olga and Chids regularly ,but still could't arrive at a conclusion as to how to design it. Made a report on random graphs , mainly generating random DAGS,random networks, random graphs with hamiltonian cycles,etc and also about my idea on random graph generation with meshes.
Met Kelli Bacon ,she is another undergraduate and was informed that she will be assisiting us in our project.The regular meetings at 9 continued ...worked on implementing the design of the property class and listed all the properties that we are goin to support. Implemented bellmanford algorithm and diameter algorithm. Also thought of a way to make Dijikstra work on negative weight edges.(which i need to get checked!) Had a meeting with Dr.Amato and got an overview of the library we were going to implement finally, i.e. Generic Graph Generator Library . Went through the interface developed by Olga and studied it . Tried to figure out a way to generate graphs according to the properties and their priority + precision, but it was getting complicated and messy because the properties of the graphs we listed were not independent, so started thinking about the non-mutable property methods and their static and dynamic implementations .I added some functions to the graph generator interface, namely the inter and the intra merging functions . I and Kelli sorted out procedures to implement some properties dynamically and documented our ideas.Also Roger wanted to make Dijikstra and other algorithms to be generic w.r.t the weights and gave me some ideas as to how to implement it. Made my first draft of research plan and my 5 week schedule report at a go!
Parasol Home | Research | People | General info | Seminars | Resources
Parasol Laboratory, 425 Harvey R. Bright Bldg, 3112 TAMU, College Station, TX 77843-3112
email@example.com Phone 979.458.0722 Fax 979.458.0718
Department of Computer Science and Engineering | Dwight Look College of Engineering | Texas A&M University
Privacy statement: Computer Science and Engineering Engineering TAMU
Web Accessibility Policy and Law - Web Accessibility and Usability Standards - Contact Webmaster