Programming for Hybrid Multi Many-core MPP Systems
  • Home
  • Table of Contents
    • Chapter 1 / Introduction
  • Previous Publications
    • High Performance Computing
    • A Guidebook
  • About the Authors
    • John Levesque
    • Aaron Vose

Chapter 5


Link to the solution:
5.1   Compare and contrast sampling and tracing profiles.​​
​5.2   Why is it important to profile an application with a problem set that is representative of those used in production?
5.3   Why is it useful to obtain a profile which includes loop trip counts (consider -hprofile_generate or similar)?​
5.4   What are hardware performance counters and what kinds of information do they provide?​
5.5   Construct a case study: Select an application and problem set for further examination.
     a. Obtain a sampling profile of the application.
     b. Obtain a tracing profile of the application.
     c. Obtain a profile of the application including a call-tree.
     d. Obtain a profile of the application including loop statistics.
     e. Obtain a profile of the application with hardware counter data.
     f. From the above profiles, identify:
          
i. Degree of load-imbalance.
        ii. Important routines and looping structures.
       iii. Locations of any significant MPI communication.

​​
 

5.1   Compare and contrast sampling and tracing profiles.​
 

5.2   Why is it important to profile an application with a problem set that is representative of those used in production?​
 

5.3   Why is it useful to obtain a profile which includes loop trip counts (consider -hprofile_generate or similar)?​
 

5.4   What are hardware performance counters and what kinds of information do they provide?​
 

5.5   Construct a case study: Select an application and problem set for further examination.
     a. Obtain a sampling profile of the application.
     b. Obtain a tracing profile of the application.
     c. Obtain a profile of the application including a call-tree.
     d. Obtain a profile of the application including loop statistics.
     e. Obtain a profile of the application with hardware counter data.
     f. From the above profiles, identify:
           i. Degree of load-imbalance.
         ii. Important routines and looping structures.
       iii. Locations of any significant MPI communication.
Proudly powered by Weebly
  • Home
  • Table of Contents
    • Chapter 1 / Introduction
  • Previous Publications
    • High Performance Computing
    • A Guidebook
  • About the Authors
    • John Levesque
    • Aaron Vose