Personal tools

ThreadScope Tour/Profile

From HaskellWiki

< ThreadScope Tour
Revision as of 15:57, 7 December 2011 by EricKow (Talk | contribs)

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

Contents

1 Objectives

Examine a ThreadScope profile on a realistic program

Threadscope on Sudoku2

2 Concepts

  1. HEC - Haskell Execution Context; basically there is one HEC for each core you enable with the RTS option -N

3 Steps

  1. Build sudoku2 and view its eventlog in ThreadScope

    rm sudoku2
    ghc -O2 sudoku2.hs -threaded -rtsopts -eventlog
    ./sudoku2 sudoku17.1000.txt +RTS -N2 -ls
    threadscope sudoku2.eventlog
  2. Examine the activity graph. The dashed horizontal lines provide hints About the amount of activity you might expect. Here for example we are using two cores, which is reflected the line dividing the graph into two pieces.

    Threadscope activity

    In the example above, we can see both cores nearly maxed out until around 1.85s at which point the amount of activity drops down to half.

  3. To see how this activity is distributed between cores, have a look at the HECs just below.

    One of the cores stopped working

    It so happens that the reason the activity dropped in half is that one of the cores stopped working.

  4. Observe the link between activity and garbage collection.

    Activity ceased during garbage collection

4 Questions and exercises

  1. What do you think is causing the uneven division between the 2 cores?
  2. If you have more than 2 course available, what happened when you enable them?
  3. Try profiling the k-means example and studying the profile.