A Framework for FPGA-based Accelerators with Maximum Memory Performance
Project/Area Number |
16K16026
|
Research Category |
Grant-in-Aid for Young Scientists (B)
|
Allocation Type | Multi-year Fund |
Research Field |
Computer system
|
Research Institution | Hokkaido University |
Principal Investigator |
|
Project Period (FY) |
2016-04-01 – 2018-03-31
|
Project Status |
Completed (Fiscal Year 2017)
|
Budget Amount *help |
¥3,900,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥900,000)
Fiscal Year 2017: ¥1,950,000 (Direct Cost: ¥1,500,000、Indirect Cost: ¥450,000)
Fiscal Year 2016: ¥1,950,000 (Direct Cost: ¥1,500,000、Indirect Cost: ¥450,000)
|
Keywords | FPGA / 高位合成 / Python / コンパイラ / 深層学習 / ディープニューラルネットワーク / 計算機システム |
Outline of Final Research Achievements |
We developed a multi-paradigm high-level hardware design framework that easily exploits on-chip memory blocks and memory bandwidth of an FPGA. The framework is based on Veriloggen, a Python-based domain-specific language for hardware design. The newly developed framework supports 3 different programming paradigms; The compiler supports Sequential, Stream, and RTL. In addition to the framework, we developed a highly-abstracted dataflow-based hardware compiler for deep neural networks.
|
Report
(3 results)
Research Products
(18 results)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
[Presentation] Customizable Hardware Abstraction2016
Author(s)
Shinya Takamaeda-Yamazaki
Organizer
16th International Forum on MPSoC for Software-defined Hardware (MPSoC 2016)
Place of Presentation
Nara Hotel, Nara, Nara, Japan
Year and Date
2016-07-11
Related Report
Int'l Joint Research / Invited
-
-