| 000 | 00000nam c2200205 c 4500 | |
| 001 | 000045953628 | |
| 005 | 20230706174817 | |
| 007 | ta | |
| 008 | 180625s2018 ulkad bmAC 000c eng | |
| 040 | ▼a 211009 ▼c 211009 ▼d 211009 | |
| 085 | ▼a 0510 ▼2 KDCP | |
| 090 | ▼a 0510 ▼b 6D36 ▼c 1088 | |
| 100 | ▼a 박상현 ▼g 朴相炫 | |
| 245 | 1 0 | ▼a DQN-based OpenCL workload partition for performance optimization / ▼d Sanghyun Park |
| 246 | 1 1 | ▼a 성능 최적화를 위한 DQN 기반의 OpenCL 워크로드 분할 |
| 260 | ▼a Seoul : ▼b Graduate School, Korea University, ▼c 2018 | |
| 300 | ▼a iv, 47장 : ▼b 삽화, 도표 ; ▼c 26 cm | |
| 500 | ▼a 지도교수: 서태원 | |
| 502 | 0 | ▼a 학위논문(석사)-- ▼b 고려대학교 대학원, ▼c 컴퓨터·전파통신공학과, ▼d 2018. 8 |
| 504 | ▼a 참고문헌: 장 45-47 | |
| 530 | ▼a PDF 파일로도 이용가능; ▼c Requires PDF file reader(application/pdf) | |
| 653 | ▼a OpenCL ▼a Workload Partitioning ▼a DQN | |
| 776 | 0 | ▼t DQN-based OpenCL Workload Partition for Performance Optimization ▼w (DCOLL211009)000000081713 |
| 900 | 1 0 | ▼a Park, Sang-hyun, ▼e 저 |
| 900 | 1 0 | ▼a 서태원, ▼g 徐泰源, ▼d 1969-, ▼e 지도교수 ▼0 AUTH(211009)153243 |
| 945 | ▼a KLPA |
전자정보
소장정보
| No. | 소장처 | 청구기호 | 등록번호 | 도서상태 | 반납예정일 | 예약 | 서비스 |
|---|---|---|---|---|---|---|---|
| No. 1 | 소장처 과학도서관/학위논문서고/ | 청구기호 0510 6D36 1088 | 등록번호 123059637 | 도서상태 대출가능 | 반납예정일 | 예약 | 서비스 |
| No. 2 | 소장처 과학도서관/학위논문서고/ | 청구기호 0510 6D36 1088 | 등록번호 123059638 | 도서상태 대출가능 | 반납예정일 | 예약 | 서비스 |
컨텐츠정보
초록
This paper proposes a Deep Q Network (DQN)-based method for the workload partition problem in OpenCL. The DQN, a reinforcement learning algorithm, optimizes the workload partition for each PU by the self-training, based on the accumulated performance data on the computing environment. Our experiments reveal that the DQN-based partition provides the performance improvement by up to 62.2% and 6.9% in JPEG decoder, compared to the LuxMark-based and target-based partitions, respectively. The DQN is able to capture the low-level contention in slave devices such as caches and memory, and the communication bottleneck between devices, and reflect it to the workload partition ratio.
목차
1. Introduction 1 2. Related Works 4 3. Background 5 3.1 OpenCL 5 3.2 Deep Q Network 6 4. Workload Partition Optimization using DQN 8 4.1 Training Model for OpenCL Workload Partition 8 4.2 Training Algorithm 12 5. Experiments 15 5.1 Experiment Environment 15 5.2 LuxMark Based Partition (Baseline #1) 22 5.3 Target-Based Partition (Baseline #2) 22 5.4 DQN-based Partition 24 5.5 Performance Comparison 26 5.6 Analysis 29 6. Conclusion 44 References 45
