The first 15 tasks would finish quickly and wait idle for the last task to finish execution (Unless there are several other tasks in the work queue which the cores can pick and execute).
nemo
Sorting the tasks in decreasing order of computation time before dynamic scheduling can solve the problem being depicted in this slide.
ctabrizi
@nemo, are you referring to Longest Processing Time?
https://en.wikipedia.org/wiki/Multiprocessor_scheduling
ask
Another potential approach to solve is to have multiple task queues based on task size/execution time and assign the queues priorities to ensure that tasks of a larger class are executed first. This kind of scheduling policy will have a linear scheduling time as opposed to sorting which is order nlogn.
The first 15 tasks would finish quickly and wait idle for the last task to finish execution (Unless there are several other tasks in the work queue which the cores can pick and execute).
Sorting the tasks in decreasing order of computation time before dynamic scheduling can solve the problem being depicted in this slide.
@nemo, are you referring to Longest Processing Time? https://en.wikipedia.org/wiki/Multiprocessor_scheduling
Another potential approach to solve is to have multiple task queues based on task size/execution time and assign the queues priorities to ensure that tasks of a larger class are executed first. This kind of scheduling policy will have a linear scheduling time as opposed to sorting which is order nlogn.
@ctabrizi Yes! That's what I was referring to.