Hide menu

TDDC78 mid-term evaluation 2012

The course was mid-term-evaluated by the muddy card method before the break in the tenth lecture (after OpenMP and MPI advanced issues, by C.K.) on wednesday 18/4/2012 (10:15-12:00). 27 students attended the lecture, and I received 19 cards.
In the following I summarize those issues that were raised by several students, and comment as appropriate.

Subject

  • Interesting topic/contents (4)
  • Should take up more general-purpose computing (1)
    Comment: The focus of TDDC78 on scientific computing is intentional, as our course is also serving as a profile course for several natural science and engineering programs, and can be useful e.g. as preparation for thesis work in computational science and engineering, such as computational physics, chemistry and bioinformatics, which are important research fields for Linköping university.
    Note that we also offer a course in general purpose parallel (and GPU) programming in autumn, TDDD56 Multicore and GPU programming, which also has a more general purpose parallel programming profile for the labs; TDDD56 can be seen as a complement of TDDC78 (and vice versa).
  • OpenMP is useful even outside the course / HPC domain (1)

Lectures

  • Good / clear / interesting lectures (9)
  • Lectures should use more examples (2)
    more whiteboard (1),
    and assert that students really understand what is being said. (1)
  • There should be better synchronization between the different lecturers, there was some repetition and confusion (1)
  • Some repetition in the lectures. (1)
  • Sometimes too detailed (1)
  • Dislike Fortran, would prefer only C/C++. (1)

Course material

  • Good/clear/informative lecture slide material (3)
  • An example of MPI_Scatterv would be nice (1)

Lesson(s)

  • Well done, informative and pedagogical lesson (3)
  • Could have more time to explain labs in more detail (1)

Labs

  • Good labs / labs are fun / relevant / interesting (7)
  • Somewhat difficult (2)
  • Hard to get started (1) / Needs more detailed instructions (1)
  • Supplied makefile of lab 1 / of pthreads lab did not work on Neolith (2)
  • Lab compendium should be updated (1) /
    Lab instructions regarding MFLOPS calculation could use some clarification (1)
  • Sometimes hard to get even short time jobs on Neolith (1)
  • Labs 1 and 2 should use larger problem sets (bigger images) to make it easier to compare. (1)
  • Not enough time for labs, not enough help on labs (1)

Organization

  • Access to (NSC) supercomputer system is great (3)
  • Deadline for NSC accounts a little tight (1) /
    Passport copy requirement for labs is disturbing (1)
  • Should give a recommendation how to distribute the labs (5 labs on 9 scheduled opportunities) (1)
  • Information about grading would be nice - how important is performance? (1)
    Comment: Regarding the labs, ask your lab assistant. Some practical hints on program optimization will be given early in the next lecture. - Exam questions about parallel program design are usually explicit about the expected quality of the solution (e.g., time-optimal) and the points you still can get for a correct but inferior solution. See the demo exams on the example exams page.

Conclusion

By and large, the course seems to run very well.

I will try to make the remaining lectures more interactive. Now that we have gone through the major programming environments needed for the labs, we are less pressed by time. Please note that you also can help and contribute: please ask questions, also during lectures, and feel free to interrupt me whenever I am too fast or not clear enough.

Thanks a lot for your comments!

Christoph Kessler, course leader TDDC78

Page responsible: Webmaster
Last updated: 2012-04-23