Research

  • Real-time sound field rendering

Real-time sound field rendering is computationally-intensive.This research investigates a hardware-accelerated sound field rendering system by co-designing algorithm and architecture to offer real-time rendering according to live sound signals, provide real acoustic characteristics for more accurate sound synthesis, and explore the architecture extension to develop a generic accelerated platform to solve sound/wave propagation problems in other domains. We mainly focus on the (1) System architecture design to speed up computation; (2) Rendering algorithm design and optimization for hardware implementation; (3)Specific sound field rendering processor to reduce power consumption and improve system scalability. The project was supported by the JSPS Kakenhi (General C) JP1 9 K 1 2 0 9 2 and JP2 2 K 1 2 1 2 3.

  • Computation with arbitrary precision and its acceleration

This research will investigate a generic acceleration system for computation with arbitrary precision. We mainly focus on the data representation with arbitrary precision, the processing element's architecture with arbitrary precision, system architecture of the accelerator, and the generation of the entire system. Matrix multiplication, stencil computation, and AI will be targeted.

  • CGRA

This research will study new CGRA for high performance computing. We mainly focus on the processing element's architecture, interconnection topology of tiles and switches, and system generation. This project is collaborated with the Processor Research Team in RIKEN Center for Computational Science, and it is supported by the NEDO.