Come to think of it, even if you can't specify the number "in progress" of any other science but CEP2, you can still control how many of each run for all sciences, to the point that if the buffer has at least one waiting "ready to start", you'd be able to do a per-core science contribution. Worth experimenting, there always being the risk of idle cores, or there eventually building up so many of one science, that there ain't enough time to complete them all. A function long wished for now added. :D

It was rather curious to see that an early testing adopter has added a line to the app_config.xml (which really belongs in one place only, the projects folder). He had the <user_friendly_name> tag added, which is not mentioned in the wiki as applying to this config feature (see Application Configuration section of official wiki http://boinc.berkeley.edu/wiki/Client_configuration#Application_configuration ). I'd rather not experiment with non documented tags and risk something goes foul.

Still running with CEP2 to only run 1 thread of 8, and it's been doing that nicely 24/7, uninterrupted, because I've got always CEP2 work that is older than the size of the buffer **. In effect this could also be used to control projects X on 1 thread, project Y on 1 thread, etc. What I've not tested is, if there is an idle core [due absence of jobs for an overworked project], will it fetch work from WCG? The developers have said that the new control messes up "Round-Robin", the system that does the project weight/resource share management. Users employing the new control have to experiment and not be surprised that one or the other or all does not *seem* to be run anymore on a "fairness" basis... but what to expect if you stick a spanner in the wheel ;>)

The coding to identify intel_gpu is progressing nicely, but it will be OpenCL as I understand it. Who'll be the first to port a science up in BOINC world... All those xxx Bridges that may come to use (and maybe all those jerky mouses getting added ;P)

** Meaning, each time a CEP2 task finishes, there's another CEP2 task older than the rest of the cached work to take the slot. In FIFO (First in, First Out], the oldest task of a project gets first turn, when that project has share-rights to crunch.