Inline vector compression for computational physics
نویسندگان
چکیده
A novel inline data compression method is presented for single-precision vectors in three dimensions. The primary application of the accelerating computational physics calculations where throughput bound by memory bandwidth. scheme employs spherical polar coordinates, angle quantisation, and a bespoke floating-point representation magnitude to achieve fixed ratio 1.5. anisotropy this considered, along with companding fractional splitting techniques improve efficiency representation. We evaluate numerically within context high-order fluid dynamics. For both isentropic convecting vortex Taylor–Green test cases, results are found be comparable those without compression. Performance evaluated vector addition kernel on an NVIDIA Titan V GPU; it demonstrated that speedup 1.5 can achieved.
منابع مشابه
Computational RAM Implementation of Vector Quantization for Image Compression
Vector Quantization (VQ) is a powerful techreported in the literature. One class of architectures is nique for low-bit rate image and video compression. However, VQ is a computational intensive technique. In this paper, a Computational RAM (C*RAM) implementation of VQ for real-time image compression is presented. The C*RAM architecture enables parallel search in the direction of codewords. The ...
متن کاملStatistical and Computational Physics
(1) Exact universal amplitude ratios for two-dimensional Ising model and a quantum spin chain. (2) Critical behaviour of semi-infinite quenched dilute Ising-like systems in three dimensions: Ordinary transition. (3) Universal scaling functions for bond percolation on planar random and square lattices with multiple percolating clusters. (4) Polydispersity effect and universality of finite-size s...
متن کاملComputational complexity in physics
More generally, computational complexity is about drawing qualitative distinctions between problems based on how hard it is to solve them. In a sense, computational complexity is descended from the concerns of the constructivists on hearing from Cantor that some infinities are bigger than others, e.g. that there are more subsets of the integers than their are integers. The constructivists denie...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Computer Physics Communications
سال: 2021
ISSN: ['1879-2944', '0010-4655']
DOI: https://doi.org/10.1016/j.cpc.2020.107562