Normally, we are all in favor of increasing the concurrency of our software – how many parallel threads it is doing at the same time. Typically, a reasonable rule of thumb is two times the number of cores plus 2. For a quad core chip that means 10 threads. If Intel Hyper-Threading is available on a quad core chip we push the work to 18 threads. When things are going well significant savings are realized in clock time: if it takes 18 minutes for one thread to finish a set amount of work it might take 2 minutes for 18 threads to do the work in parallel.

However, when it comes to having live students get kwan daos, less is better. There’s way too much noise and confusion when 16 students are fetching long weapons. We are experimenting  with four groups of four students each. A little slower, but safer.