Summer at LSU 2016 REU Student: John Bell PI: Dr. Hartmut Kaiser - - PowerPoint PPT Presentation

summer at lsu 2016 reu student john bell pi dr hartmut
SMART_READER_LITE
LIVE PREVIEW

Summer at LSU 2016 REU Student: John Bell PI: Dr. Hartmut Kaiser - - PowerPoint PPT Presentation

Summer at LSU 2016 REU Student: John Bell PI: Dr. Hartmut Kaiser Introduction HPX - A general purpose C++ runtime system for parallel and distributed applications of any scale and can run on multiple localities. GIDs Global


slide-1
SLIDE 1

Summer at LSU 2016 REU Student: John Bell PI: Dr. Hartmut Kaiser

slide-2
SLIDE 2

Introduction

ž HPX - A general purpose C++ runtime system

for parallel and distributed applications of any scale and can run on multiple localities.

ž GIDs – Global Identification addresses,

currently HPX has a container for mapping.

ž AGAS - Active Global Address Space, key

concept introduced by HPX supporting and enabling the uniform syntax for local and remote operations

ž Associative containers hold data by pairs in a

key, value relationship.

slide-3
SLIDE 3

Project

ž Topic: Work with various associative

containers and either determine or refine

  • ne or several to optimally work with

HPX.

ž Goal: To select the best performing data

structure for AGAS used to lookup information about global identifiers (GIDs).

ž Task: Testing search trees along with

benchmarking.

slide-4
SLIDE 4

Tasks and Timing

ž Identified several data structures.

Concentrating on the main operations preformed on databases.

ž Criteria:

— Able to handle large amounts of insertions,

searches and deletions.

— Preferable to be Standard Template Library

(STL) compatible.

slide-5
SLIDE 5

Benchmarking

ž Key Features:

— Uniform data sets (1million – 10 million

elements).

— Measured time how long for each of three

tasks.

○ Insertions ○ Searches ○ Deletions

slide-6
SLIDE 6

Search Trees

2-3 Bx Map 2-3-4 (Optimal) Binary Search Unordered Map AA Dancing 123 Top-Down List (a,b) H Tree S-Tree AVL Interval Splay B+ Order Statistic T B (Left-leaning) Red Black Treap B* Scapegoat UB

slide-7
SLIDE 7

Revised Search Trees

ž 2-3 ž AVL ž B ž Binary Search Tree ž Map ž Unordered Map

slide-8
SLIDE 8

Next Steps

ž Revisit the list and maybe implement

more search trees

ž Use the benchmark program to evaluate

  • ther programs

ž Create new data sets for better

evaluation

slide-9
SLIDE 9

Thank you!

ž Dr. Hartmut Kaiser ž Stellar Group ž CCT

slide-10
SLIDE 10

Questions?