User:Pmillar

From REU@MU
Revision as of 15:27, 30 July 2013 by Pmillar (Talk)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Power Measurement

Performance Measurement

Magma

Research Overview

My name is Patrick B. Millar and I am currently a senior in the college of Engineering. I am majoring in Bio-Computing engineering and will be graduating in December 2013. During the summer of 2013 I will be working on Self-Adaptive Power-Aware Heterogeneous Multicore computing with the assistance of my mentor Dr. Rong.

Weekly Goals

Week 1 -Orientation Activities

-Edit Makefile

-Review Parrallelization Papers

-Complete CPU and GPU Testing Programs

Week 2

-Complete clock speed alterations

-Finish data collection

-Review literature on CPU power optimization

-Learn more about OpenMp functions

Week 3

-Analyze data

-Continue working with CUDA syntax

-Determine steps to improve performance

-Start testing GPU and CPU usage amounts

Week 4

-Review literature on CPU and GPU power consumption

-Review literature on GPU performance

-Begin optimal clock speed testing


Week 5

-Integrate clock speed with CPU usage

-Integrate clock speed with GPU usage


Week 6

-Draw conclusions on the improvements made

-Attempt to make further improvements to TSP problem

Week 7

-Asses completed work

-Determine future work

Week 8

-First draft of paper

-Peer Review paper

Week 9

-Make final paper edits

-Prepare presentation

-Deliver sample presentation

Week 10

-Poster Presentation

-Turn in final paper

-Clean up workspaces

Weekly Log

Week 1

Tuesday

-Campus Tour

-Completed Paperwork

-Lunch with Mentor

Wednesday

-Started reviewing Parallelization Paper

-Downloaded correct files

-Modified Makefile for Compatabillity

-Tested programs for validity

Thursday

-Worked on developing Openmp knowledge

-Learned how to modify CPU and GPU Usage

-Adjusted program run time

-Added command line arguments for GPU / CPU

-Proper research practices lunch

Friday

-Completed demonstration of data collection

-Discovered optimal paths of data

-Familiarized self with data collection software

Week 2

Monday

-Created a bash script to run all possible tests

Tuesday

-Created a bash script for accumulating data

Wednesday

-Worked on modifying program to have all runs perform an equal amount of work

-Created two additional scripts to run all GPU/CPU combinations

Thursday

-Performed testing of bash scripts created on Monday/Tuesday/Wednesday

-Checked validity of results

Friday

-Brewers Game Tailgating

-Collected data on 75 three-five minute tests

-Browsed data to ensure the results were valid

Week 3

Monday

-Graphed results from initial tests

-Graphed results from power modified tests

-Attempted to analyze data differences

Tuesday

-Created and reviewed FSM Data

-Ran new FSM tests

-Began implementing CPU speed changes

-Attended presentation on good research practices

Wednesday

-Reran FSM tests do to data errors

-Read through "A Scalable Heterogeneous Parallelization Framework for Iterative Local Searches" to learn more about the FSM program

-Began CPU speed data collection

Thursday

-Found several errors with CPU scripts

-Corrected script errors

-Adjusted data collection parameters

Friday

-Modified Data Analysis script to work for other team members

-Debugged modified script to help retrieve correct data

-Finalized CPU Speed Data Collection plan

Week 4

Monday

-Began final round of CPU Data collection

-Worked to organize data in readable format

Tuesday

-Finalized FSM testing

-Began scripts for TSP testing

-Analyzed final FSM data

Wednesday

-Generated the CPU Speed TSP data

-Analyzed TSP data and realized the data was flawed

-Began running a complete new set of TSP data

Thursday

-Worked on analyzing new TSP data

Friday

-Learned the GPU commands

-Created GPU testing scripts

Sunday

-Ran tests to collect data on TSP and FSM (With varying GPU speeds)

Week 5

Monday

-Analyzed GPU Data

-Started preparing presentation

Tuesday

-Began work in Matlab

-Prepared all graphs for presentation

Wednesday

-Finalized and proofed presentation

-Finish matlab scripts

Thursday

-Mini Presentations


Friday

-Began work on conjugateGradient program

-Began work on simpleCUFFT program

Week 6

Monday

-Prepared CUFFT and CG scripts

Tuesday

-Collected CG and CUFFT Data

-Analyzed data

Wednesday

-Collected CG and CUFFT Data with varying GPU Speeds

Thursday

-Fourth of July

Friday

-Analyzed GPU Speed data

Week 7

Monday

-Began work on BlackScholes Program -Heavily researched adding OpenMP

Tuesday

-Attempted to implement CUDA timing -Finalized CUDA timing

Wednesday

-Modified program with OpenMP commands -Modified the Makefile to compile OpenMP

Thursday

-Edited Transpose code and created scripts in order to get TP working

-Created Matlab scripts in order to correctly analyze TP data

Friday

-Collected data for the Transpose program

Week 8

Monday

-Analyzed Transpose Data

-Error was found that made Transpose data invalid

-Recollected Transpose Data

Tuesday

-Started working on simpleCUFFT due to error in previous collection

-Attempted to fix bug limiting the GPU run time of the code


Wednesday

-Worked on setting up GFLOPs calculation for simpleCUFFT

-Continued work on GPU runtime


Thursday

-simpleCUFFT run time bug was found to be impossible to fix

-Stopped working on simpleCUFFT and focused on Transpose and TSP

Friday

-Worked on finding a template and organizing poster format

-Decided on the information to present via the poster

Week 9

Monday

-Worked on poster

Tuesday

-Proofread Poster

-Made last minute poster changes

-Analyzed data

Wednesday

-Submitted final poster

Thursday

-Began making corrections to Transpose code

-Began changing TSP code for new tests

Friday

-Edited all scripts to change errors that occurred due to Sandy having 4 graphics cards

Week 10

Monday

-Wrote abstract for talk

-Worked on editing TSP to run with several CPU's at high speeds

Tuesday

-Worked on organizing,commenting, and gathering all of the work done

-Poster presentation


Wednesday


Thursday


Friday