CiteSeerX — Building a Basic Block Instruction Scheduler with Reinforcement Learning and Rollouts

Building a Basic Block Instruction Scheduler with Reinforcement Learning and Rollouts (1999) [10 citations — 1 self]

Popular Tags

No tags have been applied to this document.

BibTeX | Add To MetaCart

@INPROCEEDINGS{Mcgovern99buildinga,
    author = {Amy Mcgovern and Eliot Moss and Andrew G. Barto},
    title = {Building a Basic Block Instruction Scheduler with Reinforcement Learning and Rollouts},
    booktitle = {Machine Learning},
    year = {1999},
    pages = {2002}
}

Bookmarks

OpenURL

Abstract:

The execution order of a block of computer instructions on a pipelined machine can make a difference in running time by a factor of two or more. Compilers use heuristic schedulers appropriate to each specific architecture implementation to achieve the best possible program speed. However, these heuristic schedulers are time-consuming and expensive to build. We present empirical results using both rollouts and reinforcement learning to construct heuristics for scheduling basic blocks. In simulation, the rollout scheduler outperformed a commercial scheduler on all benchmarks tested, and the reinforcement learning scheduler outperformed the commercial scheduler on several benchmarks and performed well on the others. The combined reinforcement learning and rollout approach was also very successful. We present results of running the schedules on Compaq Alpha machines and show that the results from the simulator correspond well to the actual run-time results. Keywords: Reinforcemen...

Citations

1933	Reinforcement Learning: An introduction – Sutton, Barto - 1998
931	Learning to predict by the methods of temporal differences – Sutton - 1988
656	ATOM - A system for building customized program analysis tools – Srivastava, Eustace - 1994
93	Alpha Architecture Reference Manual – SITES, EDITOR - 1992
54	Approximate Dynamic Programming for Real-Time Control and Neural Modeling – Werbos - 1992
52	Two kinds of training information for evaluation function learning – Utgoff - 1991
26	On-line Policy Improvement Using Monte-Carlo Search – Tesauro, Galperin - 1996
22	Differential training of rollout policies – BERTSEKAS - 1997
20	Rollout Algorithms for Combinatorial Optimization – Bertsekas, Tsitsiklis, et al. - 1997
19	Advantage updating applied to a differential game – Baird, Harmon, et al. - 1994
17	Learning to schedule straight-line code – Moss, Utgoff, et al. - 1997
14	Scheduling straight-line code using reinforcement learning and rollouts – McGovern, Moss - 1998
8	Least-cost instruction selection in DAGs is NP-complete. http://research.microsoft.com/ toddpro /papers/proof.htm – Proebsting - 1998
6	DECchip 21064-AA Microprocessor Hardware Reference Manual – DEC - 1992
2	The character of the instruction scheduling problem – Stefanovic - 1997
1	Darko Stefanovic. Applying reinforcement learning to instruction scheduling within basic blocks – Scheeff, Brodley, et al. - 1997