2017 Fiscal Year Research-status Report
ヘテロジニアス計算機環境における並列探索アルゴリズムの研究
Project/Area Number |
17K00296
|
Research Institution | The University of Tokyo |
Principal Investigator |
福永 ALEX 東京大学, 大学院総合文化研究科, 教授 (90452002)
|
Project Period (FY) |
2017-04-01 – 2020-03-31
|
Keywords | 人工知能 / 探索アルゴリズム / 並列探索アルゴリズム |
Outline of Annual Research Achievements |
We investigated graph search methods for the GPU. We first investigated GPU-based implementations of the standard Iterative Deepening A* algorithm. We showed that straightforward thread-based parallelization techniques which were previously proposed for massively parallel SIMD processors perform poorly due to warp divergence and load imbalance. We developed Block-Parallel IDA* (BPIDA*), which assigns the search of a subtree to a block (a group of threads with access to fast shared memory) rather than a thread, and showed that BPIDA* achieves significant speedups compared to thread-based parallelization.
|
Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
As proposed in the initial research plan, this year, we focused on developing efficient search algorithms for GPUs. We successfully developed Block-Parallel Iterative Deepening A*, a GPU-based graph, limited-memory graph search algorithm which significantly outperforms standard thread-based implementations of Iterative Deepening A*. We presented this result in a peer-reviewed paper at the Sympotium on Combinatorial Search.
|
Strategy for Future Research Activity |
The overall research plan remains unchanged. We will continue to investigate additional GPU-based search algorithms. In particular, during the next year, we will focus on search algorithms which search for satisficing solutions rather than optimal solutions. In addition, we will begin to investigate heterogeneous search algorithms which simultaneously use all available GPU and CPU resources to perform search.
|
Causes of Carryover |
(理由)In this fiscal year, the majority of the research funds were used to fund travel to international conferences for the purpose of presenting our research results from this project, as well as discuss ongoing work on this project with international researchers. Due to unpredictability of exact travel costs, we incurred a surplus of 146,727.
(仕様計画) We plan to use the surplus funds to pay for travel expenses to present and discuss the research results with both international an national researchers.
|