Last update: 2019-11-17
FP21 is our proposed 21-bit floating-point number format, which has the advantage of the same dynamic range as FP32 and a better accuracy than FP16. An example of usage of FP21 in OpenACC is provided in [FP21AXPY source code].
Takuma Yamaguchi, Kohei Fujita, Tsuyoshi Ichimura, Akira Naruse, Muneo Hori, and Maddegedara Lalith, gGPU Implementation of a Sophisticated Implicit Low-Order Finite Element Solver with FP21-32-64 Computation Using OpenACC,h In Proceedings of Sixth Workshop on Accelerator Programming Using Directives (WACCPD), IEEE, 2019. [paper]
TC-enhanced Matched Filtering
TC-enhanced Matched Filtering is a fast Matched Filter method using Tensor Core operations on NVIDIA GPUs. [source code]
Takuma Yamaguchi, Tsuyoshi Ichimura, Kohei Fujita, Aitaro Kato, and Shigeki Nakagawa, gMatched Filtering Accelerated by Tensor Cores on Volta GPUs with Improved Accuracy using Half Precision Variables,h IEEE Signal Processing Letters, 2019. [paper]
Copyright@The University of Tokyo, all rights reserved.