Page 158 - My FlipBook
P. 158
研
人 Research Fellow
員 吳真貞 Jan-Jan Wu
Faculty Ph.D., Computer Science, Yale University, United States
T +886-2-2788-3799 ext. 1610 E wuj@iis.sinica.edu.tw
F +886-2-2782-4814 W www.iis.sinica.edu.tw/pages/wuj
・ Research Fellow, Institute of Information Science, Academia Sinica (2011-present)
・ Associate Research Fellow, Institute of Information Science, Academia Sinica (2001-
2011)
・ Assistant Research Fellow, Institute of Information Science, Academia Sinica (1996-2001)
・ Associate Software Engineer, Institute of Information Industry (1987-1989)
・ Ph.D., Computer Science, Yale University (1995)
・ M.S., Computer Science and Information Engineering, National Taiwan University (1987)
・ B.S., Computer Science and Information Engineering, National Taiwan University (1985)
Research Description
My research interests include resource management in parallel and cloud computing, parallel and distributed processing for big data, and
dynamic binary translation for multicores/manycores. In resource management, we study dynamic provision, scheduling and management
of virtual machines, automatic scaling of system resources for application service requirements, and dynamic resource management for
performance/energy tradeo . In big data processing, we develop e cient data partitioning strategies for NoSQL databases, data caching and
replacement techniques for in-memory cluster computing, and distributed algorithms for large- scale graph computing.
In dynamic binary translation (DBT), we developed a system emulator, HQEMU, which supports e cient simulation of ARM binary execution
on x86 architectures. We also extend our research to address important DBT issues in architectures with SIMD (single instruction, multiple
data) extensions. Hardware manufacturers have adopted many distinct strategies for microprocessor design to improve parallelism, including
multi-cores, many-cores, GPGPU, SIMD, and others. However, these parallel architectures have very di erent parallel execution models and
thus substantial problems are encountered when migrating applications from one architecture to another: (1) application developers have
to re-write programs based on the target execution model, which increases the time to market (2) legacy applications are poorly optimized
due to under-utilization of parallelism in the target hardware, and thus, only a small fraction of the potential performance gain is realized.
To overcome these problems, we developed an e cient and retargetable dynamic binary translator to transparently transform application
binaries among di erent parallel execution models. In our current work, the DBT dynamically transforms binaries of short-SIMD loops to
equivalent long-SIMD loops, in order to exploit the wider SIMD lanes of the hosts.
Publications 7. Cing-Fu Jhu, Pangfeng Liu and Jan-Jan Wu, "Data Pinning and
Back Propagation Memory Optimization for Deep Learning on
1. Yu-Ping Liu, Ding-Yong Hong, Jan-Jan Wu, Sheng-Yu Fu, Wei- GPU," International Symposium on Computing and Networking,
Chung Hsu, "Exploiting SIMD Asymmetry in ARM-to-x86 Takayama, Japan, November 2018, (Outstanding Paper Award)
Dynamic Binary Translation," ACM Transactions on Architecture
and Code Optimization, volume 16, number 1, pages 2:1-2:24, 8. Li-Yung Ho, Jan-Jan Wu and Pangfeng Liu, "Adaptive
February 2019. Communication for Distributed Deep Learning on Commodity
GPU Cluster," IEEE/ACM International Symposium on Cluster,
2. Ding-Yong Hong, Jan-Jan Wu, Yu-Ping Liu, Sheng-Yu Fu, Wei- Cloud and Grid Computing (CCGrid) , Washington DC, USA,
Chung Hsu, "Processor-Tracing Guided Region Formation in May 2018.
Dynamic Binary Translation," ACM Transactions on Architecture
and Code Optimization, volume 15, number 4, pages 52:1-52:25, 9. Olivier Valery, Pangfeng Liu, Jan-Jan Wu, "Low precision
November 2018. deep learning training on mobile heterogeneous platform," 26th
Euromicro International Conference on Parallel, Distributed, and
3. Olivier Valery, Pangfeng Liu, Jan-Jan Wu, "A collaborative Network-Based Processing (PDP 2018), Cambridge, UK, March
CPU-GPU approach for deep learning on mobile devices," 2018.
Concurrency and Computation: Practice and Experience, volume
31, number 17, pages published online, September 2019. 10. Sheng-Yu Fu, Chih-Min Lin, Ding-Yong Hong, Yu-Ping Liu, Jan-
Jan Wu, Wei-Chung Hsu, "Exploiting SIMD Capability in an
4. Ding-Yong Hong, Yu-Ping Liu, Sheng-Yu Fu, Jan-Jan Wu, Wei- ARMv7-to-ARMv8 Dynamic Binary Translator," International
Chung Hsu, "Improving SIMD Parallelism via Dynamic Binary Conference on Compilers, Architectures and Synthesis for
Translation," ACM Transactions on Embedded Computing Embedded Systems (CASES), Turin, Italy, September 2018.
Systems (TECS), volume 17, number 3, pages 61:1-61:27,
February 2018. 11. Yu-Tung Hsieh, Chuan-Yu Lee, Ching-Chi Lin, Pangfeng
Liu, and Jan-Jan Wu, "A Bicameralism Voting Framework for
5. Olivier Valery, Pangfeng Liu, Jan-Jan Wu, "A Collaborative CPU- Combining Knowledge from Clients into Better Prediction,"
GPU Approach for Principal Component Analysis on Mobile IEEE International Conference on Big Data, Los Angeles, CA,
Heterogeneous Platform," Journal of Parallel and Distributed USA, December 2019.
Computing (JPDC), volume 120, pages 44-61, October 2018.
6. Po-Yen Wu, Pangfeng Liu, and Jan-Jan Wu:, "Versatile
Communication Optimization for Deep Learning by Modularized
Parameter Server," IEEE International Conference on Big Data ,
Seattle, USA, December 2018.
156
人 Research Fellow
員 吳真貞 Jan-Jan Wu
Faculty Ph.D., Computer Science, Yale University, United States
T +886-2-2788-3799 ext. 1610 E wuj@iis.sinica.edu.tw
F +886-2-2782-4814 W www.iis.sinica.edu.tw/pages/wuj
・ Research Fellow, Institute of Information Science, Academia Sinica (2011-present)
・ Associate Research Fellow, Institute of Information Science, Academia Sinica (2001-
2011)
・ Assistant Research Fellow, Institute of Information Science, Academia Sinica (1996-2001)
・ Associate Software Engineer, Institute of Information Industry (1987-1989)
・ Ph.D., Computer Science, Yale University (1995)
・ M.S., Computer Science and Information Engineering, National Taiwan University (1987)
・ B.S., Computer Science and Information Engineering, National Taiwan University (1985)
Research Description
My research interests include resource management in parallel and cloud computing, parallel and distributed processing for big data, and
dynamic binary translation for multicores/manycores. In resource management, we study dynamic provision, scheduling and management
of virtual machines, automatic scaling of system resources for application service requirements, and dynamic resource management for
performance/energy tradeo . In big data processing, we develop e cient data partitioning strategies for NoSQL databases, data caching and
replacement techniques for in-memory cluster computing, and distributed algorithms for large- scale graph computing.
In dynamic binary translation (DBT), we developed a system emulator, HQEMU, which supports e cient simulation of ARM binary execution
on x86 architectures. We also extend our research to address important DBT issues in architectures with SIMD (single instruction, multiple
data) extensions. Hardware manufacturers have adopted many distinct strategies for microprocessor design to improve parallelism, including
multi-cores, many-cores, GPGPU, SIMD, and others. However, these parallel architectures have very di erent parallel execution models and
thus substantial problems are encountered when migrating applications from one architecture to another: (1) application developers have
to re-write programs based on the target execution model, which increases the time to market (2) legacy applications are poorly optimized
due to under-utilization of parallelism in the target hardware, and thus, only a small fraction of the potential performance gain is realized.
To overcome these problems, we developed an e cient and retargetable dynamic binary translator to transparently transform application
binaries among di erent parallel execution models. In our current work, the DBT dynamically transforms binaries of short-SIMD loops to
equivalent long-SIMD loops, in order to exploit the wider SIMD lanes of the hosts.
Publications 7. Cing-Fu Jhu, Pangfeng Liu and Jan-Jan Wu, "Data Pinning and
Back Propagation Memory Optimization for Deep Learning on
1. Yu-Ping Liu, Ding-Yong Hong, Jan-Jan Wu, Sheng-Yu Fu, Wei- GPU," International Symposium on Computing and Networking,
Chung Hsu, "Exploiting SIMD Asymmetry in ARM-to-x86 Takayama, Japan, November 2018, (Outstanding Paper Award)
Dynamic Binary Translation," ACM Transactions on Architecture
and Code Optimization, volume 16, number 1, pages 2:1-2:24, 8. Li-Yung Ho, Jan-Jan Wu and Pangfeng Liu, "Adaptive
February 2019. Communication for Distributed Deep Learning on Commodity
GPU Cluster," IEEE/ACM International Symposium on Cluster,
2. Ding-Yong Hong, Jan-Jan Wu, Yu-Ping Liu, Sheng-Yu Fu, Wei- Cloud and Grid Computing (CCGrid) , Washington DC, USA,
Chung Hsu, "Processor-Tracing Guided Region Formation in May 2018.
Dynamic Binary Translation," ACM Transactions on Architecture
and Code Optimization, volume 15, number 4, pages 52:1-52:25, 9. Olivier Valery, Pangfeng Liu, Jan-Jan Wu, "Low precision
November 2018. deep learning training on mobile heterogeneous platform," 26th
Euromicro International Conference on Parallel, Distributed, and
3. Olivier Valery, Pangfeng Liu, Jan-Jan Wu, "A collaborative Network-Based Processing (PDP 2018), Cambridge, UK, March
CPU-GPU approach for deep learning on mobile devices," 2018.
Concurrency and Computation: Practice and Experience, volume
31, number 17, pages published online, September 2019. 10. Sheng-Yu Fu, Chih-Min Lin, Ding-Yong Hong, Yu-Ping Liu, Jan-
Jan Wu, Wei-Chung Hsu, "Exploiting SIMD Capability in an
4. Ding-Yong Hong, Yu-Ping Liu, Sheng-Yu Fu, Jan-Jan Wu, Wei- ARMv7-to-ARMv8 Dynamic Binary Translator," International
Chung Hsu, "Improving SIMD Parallelism via Dynamic Binary Conference on Compilers, Architectures and Synthesis for
Translation," ACM Transactions on Embedded Computing Embedded Systems (CASES), Turin, Italy, September 2018.
Systems (TECS), volume 17, number 3, pages 61:1-61:27,
February 2018. 11. Yu-Tung Hsieh, Chuan-Yu Lee, Ching-Chi Lin, Pangfeng
Liu, and Jan-Jan Wu, "A Bicameralism Voting Framework for
5. Olivier Valery, Pangfeng Liu, Jan-Jan Wu, "A Collaborative CPU- Combining Knowledge from Clients into Better Prediction,"
GPU Approach for Principal Component Analysis on Mobile IEEE International Conference on Big Data, Los Angeles, CA,
Heterogeneous Platform," Journal of Parallel and Distributed USA, December 2019.
Computing (JPDC), volume 120, pages 44-61, October 2018.
6. Po-Yen Wu, Pangfeng Liu, and Jan-Jan Wu:, "Versatile
Communication Optimization for Deep Learning by Modularized
Parameter Server," IEEE International Conference on Big Data ,
Seattle, USA, December 2018.
156
