For those interested in parallel dynamic scheduling, here is an interesting talk by Guy Blelloch on various scheduling algorithms. Link
As background, the randomized variant of the work-stealing algorithm implemented with continuation stealing and deques is quite common, as the time of execution is, in expectation, k*Tmin, where Tmin is the theoretical min run time and space usage is O(S*P), where S is the stack usage if the computation was run on a single core and P is the number of processors. The talk outlines another algorithm called Parallel Depth-first scheduling, which tends to have better space usage.
For those interested in parallel dynamic scheduling, here is an interesting talk by Guy Blelloch on various scheduling algorithms. Link
As background, the randomized variant of the work-stealing algorithm implemented with continuation stealing and deques is quite common, as the time of execution is, in expectation, k*Tmin, where Tmin is the theoretical min run time and space usage is O(S*P), where S is the stack usage if the computation was run on a single core and P is the number of processors. The talk outlines another algorithm called Parallel Depth-first scheduling, which tends to have better space usage.