Previous | Next --- Slide 10 of 38
Back to Lecture Thumbnails
askumar

Line 6: Each thread now only allocates an array corresponding to the size of its block, plus two extra rows (to copy data from other blocks that it needs). Since the threads do not share memory now, we don't need to worry about ensuring mutually exclusive array modifications.

akashr

All the threads other than thread 0 send their computation result to thread 0 (line 25b). Thread 0 receives (25f) and sums up all the floating numbers (25g) it gets from every other thread. Depending on the value, thread 0 tells every other thread whether the computation is done or not. According to what was sent from thread 0, computation will either stop or continue.