Presentation | 2018-06-14 Acceleration of Analytical Placement by Wire Length Prediction using Machine Learning Tatsuki Hoshiba, Yukihide Kohira, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In recent LSI design, it is difficult to obtain a placement that satisfies both design constraints and specifications due to increase of circuit size and progress of manufacturing technology. Analytical placement methods have been proposed to obtain a placement with short wire length for placement problem with many blocks. Analytical placement methods formulate placement problem to mathematical programming problems and obtain a placement by using their solvers. The analytical placement methods have advantages such that conditions and constraints are added easily and existing solvers for mathematical programming problems can be utilized. The analytical methods using quasi-Newton method have good convergence and they can be applied to problems with large scale circuits. However, since the analytical placement methods using quasi-Newton method depend on initial placements, they are repeatedly applied to obtain a placement with short wire length. In this paper, we propose a placement method that makes a machine learning model to predict wire length of the placement obtained by an analytical placement method using quasi-Newton method from a placement, predicts wire length after applying the analytical placement method using quasi-Newton method by using the model, and applies the analytical placement method using quasi-Newton method to only placements whose predicted wire lengths are short. We evaluate effectiveness of the proposed method in computational experiments. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Placement problem / analytical placement / machine learning / wire length prediction |
Paper # | CAS2018-14,VLD2018-17,SIP2018-34,MSS2018-14 |
Date of Issue | 2018-06-07 (CAS, VLD, SIP, MSS) |
Conference Information | |
Committee | CAS / SIP / MSS / VLD |
---|---|
Conference Date | 2018/6/14(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Hokkaido Univ. (Frontier Research in Applied Sciences Build.) |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | System and Signal Processing, etc |
Chair | Hideaki Okazaki(Shonan Inst. of Tech.) / Shogo Muramatsu(Niigata Univ.) / Morikazu Nakamura(Univ. of Ryukyus) / Noriyuki Minegishi(Mitsubishi Electric) |
Vice Chair | Taizo Yamawaki(Hitachi) / Naoyuki Aikawa(TUS) / Kazunori Hayashi(Osaka City Univ) / Shigemasa Takai(Osaka Univ.) / Nozomu Togawa(Waseda Univ.) |
Secretary | Taizo Yamawaki(Shonan Inst. of Tech.) / Naoyuki Aikawa(Hitachi) / Kazunori Hayashi(Takushoku Univ.) / Shigemasa Takai(Hiroshima Univ.) / Nozomu Togawa(Toshiba) |
Assistant | Motoi Yamaguchi(Renesas Electronics) / / Hideki Kinjo(Okinawa Univ.) |
Paper Information | |
Registration To | Technical Committee on Circuits and Systems / Technical Committee on Signal Processing / Technical Committee on Mathematical Systems Science and its applications / Technical Committee on VLSI Design Technologies |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Acceleration of Analytical Placement by Wire Length Prediction using Machine Learning |
Sub Title (in English) | |
Keyword(1) | Placement problem |
Keyword(2) | analytical placement |
Keyword(3) | machine learning |
Keyword(4) | wire length prediction |
1st Author's Name | Tatsuki Hoshiba |
1st Author's Affiliation | The University of Aizu(Univ. of Aizu) |
2nd Author's Name | Yukihide Kohira |
2nd Author's Affiliation | The University of Aizu(Univ. of Aizu) |
Date | 2018-06-14 |
Paper # | CAS2018-14,VLD2018-17,SIP2018-34,MSS2018-14 |
Volume (vol) | vol.118 |
Number (no) | CAS-82,VLD-83,SIP-84,MSS-85 |
Page | pp.pp.75-80(CAS), pp.75-80(VLD), pp.75-80(SIP), pp.75-80(MSS), |
#Pages | 6 |
Date of Issue | 2018-06-07 (CAS, VLD, SIP, MSS) |