Course Evaluation

Of the 17 participants, 11 filled out the course evaluation. Almost all came from a university or a research laboratory, and more than half of these were research staff, plus two undergraduate and three graduate students. Their disciplines varied widely, from Physics to Bio-sciences to Fluid Dynamics, with many coming from Computing Sciences. Only three respondents were CSCS users.

Surprisingly 8 of 11 respondents felt the course had appropriate duration (2 days), even though we felt we were packing a great deal of information into those two days (to make it fit in with the subsequent week's conference). Only 3 felt the duration was too short.

The individual units in the two-day course were given for evaluation. The units were given almost uniformly good grades:

1. "OpenCL / CUDA Introduction" received grades of 4.3 to 4.7 out of 5.0 for teacher preparation and knowledge. The contents and scope grades ranged from 3.9 to 4.3. Only 2 responded that the teaching speed was too fast, the rest "just right".

2. "NVIDIA libraries and tools" received grades of 4.4 to 4.8 for teacher preparation and knowledge, and 4.0 to 4.2 for content. Again only 2 responded that the teaching speed was too fast.

3. "Thrust, CUSP, ViennaCL and higher level libraries" received a uniform grade of 4.6 for teacher knowledge and preparation, and 4.1 to 4.3 for content. Again only 2 responded that the teaching speed was too fast.

4. "Introduction to Paralution" had grades of 4.2 to 4.6 for teacher knowledge and preparation, and 3.7 to 4.0 for content. Here 4 responded that the teaching speed was too fast, 4 just right. 3 people did not respond.

5. "Introduction to MAGMA" received grades of 4.2 to 4.6 for teacher knowledge and preparation, and 4.2 to 4.3 for content. 3 responded that it was too fast, 5 just right, and 3 did not respond.

For all units it was commented that it would have been nice to have "maybe less material more time for exercises".

Overall, 5 felt the course met expectations, 1 that it exceeded them, 3 felt it "nearly met" them and none felt it missed expectations. 2 did not respond. All the respondents felt that this was either very important (3) or "important" (6) to their profession (2 did not respond).

There were two additional comments:

There were too many libraries been discussed and the difficulty to decide which one to use. It will be great deal to have stated the strengths and weaknesses of these libraries in comparison. I do accept this is the task of the individual attendees but a general overview of comparison would have suffice. One important phrase i took with me was, "try to use existing libraries if you can.." Great!

Finally:

some hands-on sessions did require knowledge of the numerical methods, rather than just focussing on the programming. this is not ideal if you don't know it