Scribd Logo
Upload

Welcome to Scribd!

  • 211 views

    HPC Architecture and ECO System PDF

    Uploaded by

    Er Umesh Thoriya
    This document outlines the course details for the subject "HPC Architecture and ECO system" offered as an elective for the Master of Engineering program. The course aims to provide in-depth knowledge of parallel architectures and different shared/distributed memory architectures to support exascale computing. It covers topics like parallel system organization, message passing with MPI and shared memory with OpenMP. Students will learn performance analysis tools and apply them to practical problems involving matrix multiplication, Jacobi iteration, and Monte Carlo simulation using CUDA and OpenACC. The course assessments include a theory exam, practical exam, and projects.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    HPC Architecture and ECO System PDF

    Uploaded by

    Er Umesh Thoriya
    211 views3 pages
    This document outlines the course details for the subject "HPC Architecture and ECO system" offered as an elective for the Master of Engineering program. The course aims to provide in-depth knowledge of parallel architectures and different shared/distributed memory architectures to support exascale computing. It covers topics like parallel system organization, message passing with MPI and shared memory with OpenMP. Students will learn performance analysis tools and apply them to practical problems involving matrix multiplication, Jacobi iteration, and Monte Carlo simulation using CUDA and OpenACC. The course assessments include a theory exam, practical exam, and projects.

    Original Title

    3720220_HPC Architecture and ECO system.pdf

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document outlines the course details for the subject "HPC Architecture and ECO system" offered as an elective for the Master of Engineering program. The course aims to provide in-depth knowledge of parallel architectures and different shared/distributed memory architectures to support exascale computing. It covers topics like parallel system organization, message passing with MPI and shared memory with OpenMP. Students will learn performance analysis tools and apply them to practical problems involving matrix multiplication, Jacobi iteration, and Monte Carlo simulation using CUDA and OpenACC. The course assessments include a theory exam, practical exam, and projects.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    211 views3 pages

    HPC Architecture and ECO System PDF

    Uploaded by

    Er Umesh Thoriya
    This document outlines the course details for the subject "HPC Architecture and ECO system" offered as an elective for the Master of Engineering program. The course aims to provide in-depth knowledge of parallel architectures and different shared/distributed memory architectures to support exascale computing. It covers topics like parallel system organization, message passing with MPI and shared memory with OpenMP. Students will learn performance analysis tools and apply them to practical problems involving matrix multiplication, Jacobi iteration, and Monte Carlo simulation using CUDA and OpenACC. The course assessments include a theory exam, practical exam, and projects.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3

    GUJARAT TECHNOLOGICAL UNIVERSITY

    Master of Engineering
    Subject Code: 3720220
    Semester – II
    Subject Name: HPC Architecture and ECO system

    Type of course: Elective

    Prerequisite: Computer Architecture and Organization, Data Structure

    Rationale: Algorithmic processing performed in High Performance Computing environments impacts the
    lives of billions of people, and planning for exascale computing presents significant power challenges to the
    industry. The objective of the course is to impart in depth knowledge of parallel architectures and different
    shared/distributed memory architectures in support to exascale computing. MPI and OpenMP are discussed
    along with their applications.

    Teaching and Examination Scheme:

    Teaching Scheme Credits Examination Marks Total


    L T P C Theory Marks Practical Marks Marks
    ESE (E) PA (M) ESE (V) PA (I)
    3 0 2 4 70 30 30 20 150

    Content:

    Sr. No. Content Total


    Hrs

    1 Overview of parallel system organization 4

    2 Examples of Scientific Computing; Parallel Languages, Embarrassingly parallel problems; 12


    problem decomposition, graph partitioning and load balancing
    3 Introduction to message passing and MPI programming 7

    4 Introduction to shared memory and OpenMP programming 7

    5 Recent trends in OpenMP and MPI programming, application areas of scientific 9


    computing.
    6 Performance analysis tools; HPCToolkit, OpenSpeedShop, 9
    Debugging and Profiling tools; GDB, PAPI, mpiP, ompP, Valgrind, MPI Program Profiler
    Total 48

    Reference Books:

    1. 1. Parallel Programming for Multicore and Cluster Systems by Thomas Rauber and Gudula Runger.
    2. Scientific Parallel Computing by Scott, Clark, and Bagheri.
    Page 1 of 3

    w.e.f. AY 2018-19
    GUJARAT TECHNOLOGICAL UNIVERSITY
    Master of Engineering
    Subject Code: 3720220
    3. Using OpenMP: Portable Shared Memory Parallel Programming by Chapman, Jost, and van der
    Pas.

    Course Outcomes:
    Sr. No. CO statement Marks % weightage

    CO-1 Develop parallel algorithms for high performance systems. 25%

    CO-2 Perform problem decomposition and load balancing using MPI and 25%
    OpenMP.
    CO-3 Identify hot spots in the codes using performance analysis tools. 25%

    CO-4 Evaluate the performance of the code using the tools 25%

    Distribution of marks weightage for cognitive level

    Bloom’s Taxonomy for Cognitive Domain Marks


    % weightage
    Recall 5
    Comprehension 5
    Application 20
    Analysis 25
    Evaluate 25
    Create 20
    Practical List:

    Use Valgrind, Vtune Amplifier, Nvidia Visual Profiler and Nvidia Nsight to identify hotspots and other
    parameters for detailed analysis of following the practicals.
    1) Calculate standard deviation using Pthread, OpenMP and MPI.
    2) Write parallel code for Matrix Matrix Multiplication using MPI cluster of 4 nodes, OpenMP, PVM
    cluster of 4 nodes, OpenACC and CUDA and compare and plot the performance in terms of
    execution time for Matrix size of 1000 x 1000, 5000 x 5000, 10,000 x 10,000, 20,000 x 20,000.
    3) Write the programs in MPD or in C with the Pthreads library for the following:
    a) Sequential Jacobi iteration program
    b) Parallel Jacobi iteration program
    c) Sequential multigrid program
    d) Parallel multigrid program
    4) Perform Monte Carlo simulation using NVIDIA's CURAND library for random number
    generation.
    Write your own small program to compute the average value of
    az2 + bz + c
    Page 2 of 3

    w.e.f. AY 2018-19
    GUJARAT TECHNOLOGICAL UNIVERSITY
    Master of Engineering
    Subject Code: 3720220
    where z is a standard Normal random variable (i.e. zero mean and unit variance, which is what the
    random number generator produces) and a, b, c are constants which you should store in constant
    memory. It is suggested to use each thread to average over 100 values, and then write this to a
    device array which gets copied back to the host for the averaging over the contributions from each
    of the threads.
    (Note: the average value should be close to a + c.)
    5) Implement 3D Laplace Finite Solver using CUDA and OpenACC.

    Page 3 of 3

    w.e.f. AY 2018-19

    You might also like