CUDA work estimates put host into EDF mode

[Eric Korpela]

The scheduler uses CUDA work estimates when assigning work, but the host still uses CPU base work estimates for scheduling. So upon downloading a large number of CUDA results, the host gets into EDF mode. If the S@H deadlines are before the deadlines for other projects, it tries to schedule the S@H jobs on the CPUs, but can't because they need to use the GPU. So every time the scheduler loop executes it schedules a different non-CUDA job on each core.

From SETI@home bug report: "Looks like this may be tied to the problem someone else mentioned about the units running at high priority. Basically, Manager decides that the CUDA WU's are going to miss deadline, even though they aren't, and goes into EDF mode. Then, it tries to switch tasks to allow the CUDA jobs to run. But since there isn't another GPU available, it gets stuck in an infinite loop of task switching between the non-CUDA jobs."

[davea]

Fixed in 6.4.5