Uploaded image for project: 'Moodle'
  1. Moodle
  2. MDL-41674

grouped_parser_processor::grouped_parent_exists is called too much

XMLWordPrintable

    • Icon: Improvement Improvement
    • Resolution: Fixed
    • Icon: Minor Minor
    • 2.9
    • 2.6, 2.7, 2.8, 2.9
    • Backup
    • MOODLE_26_STABLE, MOODLE_27_STABLE, MOODLE_28_STABLE, MOODLE_29_STABLE
    • MOODLE_29_STABLE
    • Hide

      1. Backup a course with users and questions
      2. Restore the course and confirm the users and question information is restored correctly.

      If you have XHProf installed and can restore the course prior to the patch, do that and save the result as a reference run.
      Then restore the course again and compare the results to ensure performance has not degraded.

      For further performance confirmation;
      1. Create a M or L course depending on your hardware.
      2. Backup that course
      3. Restore that course
      4. Apply the patch for MDL-41674
      5. Restore that course again.
      6. Compare performance of the two runs to ensure we have improved.

      Show
      1. Backup a course with users and questions 2. Restore the course and confirm the users and question information is restored correctly. If you have XHProf installed and can restore the course prior to the patch, do that and save the result as a reference run. Then restore the course again and compare the results to ensure performance has not degraded. For further performance confirmation; 1. Create a M or L course depending on your hardware. 2. Backup that course 3. Restore that course 4. Apply the patch for MDL-41674 5. Restore that course again. 6. Compare performance of the two runs to ensure we have improved.

      Following on from the changes in MDL-40585, where a cache has been added to the get_parent_path function, I found from profiling that a significant time in restore is still spent on grouped_parser_processor::parent_exists.

      Adding a cache directly to this function (in addition to the existing get_parent_path cache) appears to improve performance on a large test course by about one sixth with about a 2MB memory cost.

      I did two runs before/after my code patch (alternating) using the 'M' size standard test course, with profiling turned off again. Results:

      Before: 359.5s (forgot to note memory use)
      After: 308.4s (85.6 MB)
      Before: 373.7s (83.4MB)
      After: 297.7s (85.6MB)

      Average is a 17% improvement.

      This needs review as maybe I'm missing something silly.

      If successful, it may be possible to backport this change to 2.4 and 2.5.

            mr-russ Russell Smith
            quen Sam Marshall
            Tim Hunt Tim Hunt
            Eloy Lafuente (stronk7) Eloy Lafuente (stronk7)
            Ankit Agarwal Ankit Agarwal
            Votes:
            2 Vote for this issue
            Watchers:
            8 Start watching this issue

              Created:
              Updated:
              Resolved:

                Error rendering 'clockify-timesheets-time-tracking-reports:timer-sidebar'. Please contact your Jira administrators.