How does the selection of the multi-threaded app occur? I was thinking about how running an MT app along with projects that are only single-threaded can create wasted time when there are fewer MT apps running than available cores. Could the MT app be scalable to take advantage of other cores as they become available in such a scenario?

---

Click here to see My Detailed BOINC Stats

I think that was part of the original idea, however, BOINC currently seems to always run mt apps on all the cores in a system - although at most 64. The science applications implement the required api to limit the core count but I think the scheduling algorithm to do it ended up being too complex.