Just started Udacity's "Introduction to Parallel Programming". The first Unit is online. It explains the barebones of how to launch a simple CUDA kernel for computing squares and cubes, as well as the CUDA parallel computation model. I ended Coursera's "Heterogeneous Parallel Programming" a couple of weeks ago, and so the impressions I get from this first class (compared to HPP) are the following:
1) First unit is really the barebones of CUDA (and explains a bit of the necessary C). The videos are short and concise, many quizzes take you gently to the knowledge (Udacity's style...).
2) The instructor and the layout of the slides and of the explanations in Udacity are much better than the approach followed in Coursera's HPP (Prof. Hwu waving in front of slides, without ever seeing a pen writing or underlining important info in the slides).
3) The code submission process seems to be much more effective in Udacity (I don't know about the number of users when I submitted, neither the number of available devices - GPUs). It looks exactly like the Python submission infrastructure we are used to in Udacity. The response time perceived by the user is similar.
Overall I just say that now I'm hooked to finish Udacity's "Introduction to Parallel Programming" even after finishing Coursera's HPP. From the pedagogical POV, Udacity's offer seems much better (at least in the 1st unit) than Coursera's. Also we will do a lot of image processing in the quizzes and exercises. Nice.
My take: go and see if you like. In fact, parallel processing and programming is here to go along with us (forever ?)...
asked 05 Feb '13, 11:22
answered 05 Feb '13, 11:42