COSC 6374 Parallel Computation Recap for the Quiz Edgar Gabriel Fall 2011 Edgar Gabriel Organizational Details to the Quiz Date Nov 30 2011 1pm 2 30pm Topics discussed on the next pages You can have 4 pages of handwritten cheat sheets you can write on both sides of each sheet any pages containing non handwritten items will be removed during the quiz and student will get 0 for that quiz you have to write it yourself not copy xerox it from somebody else Pages that have been xeroxed from somebody else will be treated as non handwritten COSC 6374 Parallel Computation Edgar Gabriel 1 Overview of relevant lectures Parallel Computer Architectures Introduction to MPI I IV Parallel Measures except for debugging Dense Matrix Operations Performance Modeling Parallel I O I III MPI Derived Datatypes POSIX threads OpenMP Homework 1 3 COSC 6374 Parallel Computation Edgar Gabriel Lectures not included in the Quiz PDEs although concept of ghost cells is relevant Performance Analysis Scientific Data Libraries UPC Data Parallel Approaches Performance Oriented Software Development COSC 6374 Parallel Computation Edgar Gabriel 2 Introduction to MPI I IV Basics of MPI entire lecture especially rank of a process size of communicator message envelope and data description for point to point operations Asynchronous Communication entire lecture especially message matching rules non overtaking rule Collective Communication entire lecture especially Bcast Reduce Allreduce Gather Scatter and Allgather Process Grouping entire lecture especially concept of process groups and communicators Comm split Comm create Process topologies i e Cart create and friends COSC 6374 Parallel Computation Edgar Gabriel Introduction to MPI I IV Derived Data Types Type struct Type contiguous Type vector Type indexed Type subarray do not worry about upper lower bound and resizing a data type In the quiz if you have to deal with an MPI function I will provide you the C prototype of the function No need to put them on the cheat sheet COSC 6374 Parallel Computation Edgar Gabriel 3 Dense Matrix Algorithms entire lecture except matrix matrix multiply operation Parallel I O unix file access disk striping RAID levels parallel file systems algorithmic vs list I O interfaces discontiguous I O collective I O and its derivatives hints MPI I O Performance Modeling LogP LogGP Parallel Measures Speedup Scaleup Parallel Efficiency COSC 6374 Parallel Computation Edgar Gabriel OpenMP parallel regions for loops parallel sections private shared firstprivate lastprivate clauses critical sequential single clauses reduction operations library functions get set num threads get thread num etc POSIX threads thread creation and termination lock unlock condition variables COSC 6374 Parallel Computation Edgar Gabriel 4
View Full Document
Unlocking...