SBIR Phase I: Efficient Arithmetic on Quasi-Compressed Data for Performance Improvement

SBIR 第一阶段：准压缩数据的高效算法以提高性能

• 批准号: 2111696
• 负责人: David Chen
• 金额: $ 25.6万
• 依托单位: ARITH INC.
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-02-01 至 2023-10-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2111696&HistoricalAwards=false
• 关键词: SBIR; Phase; Efficient; Arithmetic; Quasi

The broader impact of this Small Business Innovation Research (SBIR) Phase I project is to improve the performance of semiconductor chips. Specifically, it develops a novel method to operate directly on compressed data, saving time, energy, and latency. This improved performance will affect computationally intensive applications such as medical imaging (e.g., CT/MRI), climate simulation, hurricane warnings, and earthquake alerts.This Small Business Innovation Research (SBIR) Phase I project develops a method to operate on compressed data. Today, computers apply data compression to identify and remove redundancy in the data in order to save storage space. Computers apply arithmetic to compute in integers or real numbers (usually represented internally as floating-point data). However, today's computers first decompress the data, compute, and then compress the computed result, consuming additional time and energy. This project develops an arithmetic and math hardware accelerator capable of processing compressed data directly to dramatically improve computation performance, storage effectiveness, and energy efficiency. This project combines data compression and floating-point engineering to deliver the first-ever Compressed Floating-Point Unit (CFPU) that minimizes semiconductor and energy usage and reduces computation latency.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

这个小企业创新研究（SBIR）第一阶段项目的更广泛影响是提高半导体芯片的性能。 具体来说，它开发了一种新的方法来直接对压缩数据进行操作，节省了时间，能源和延迟。 这种改进的性能将影响计算密集型应用，例如医学成像（例如，CT/MRI）、气候模拟、飓风警报和地震警报。这个小型企业创新研究（SBIR）第一阶段项目开发了一种对压缩数据进行操作的方法。今天，计算机应用数据压缩来识别和删除数据中的冗余，以节省存储空间。 计算机应用算术来计算整数或真实的数（通常在内部表示为浮点数据）。 然而，今天的计算机首先对数据进行加密，计算，然后压缩计算结果，消耗额外的时间和精力。 该项目开发了一种算术和数学硬件加速器，能够直接处理压缩数据，以显着提高计算性能，存储效率和能源效率。 该项目将数据压缩和浮点工程相结合，提供了首个压缩浮点单元（CFPU），最大限度地减少了半导体和能源的使用，并减少了计算延迟。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。