Synthesizing Number Generators for Stochastic Computing using Mixed Integer Programming
Synthesizing Number Generators for Stochastic Computing using Mixed Integer Programming
复制标题
使用混合整数规划综合随机计算的数字生成器
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
L. Ceze
中科院分区：
文献类型：
--
作者：
Vincent T. Lee;A. Elliott;Armin Alaghi;L. Ceze
随机计算(SC)是一种高密度、低功耗的计算技术，它将值编码为一元比特流，而不是二进制编码(BE)值。实际的SC实现需要确定性或伪随机数序列，它们是最佳相关的，以生成比特流并获得准确的结果。不幸的是，搜索空间的大小使得手动设计最佳关联的数字序列成为一项困难的任务。为了自动化这一设计负担，我们提出了一种使用混合整数规划的综合公式来自动生成最佳相关的数字序列。特别是，我们的综合公式将乘法和平方电路等算术运算的精度分别提高了2.5倍和20倍。我们还展示了如何将我们的技术扩展到更大的电路。
Stochastic computing (SC) is a high density, low-power computation technique which encodes values as unary bitstreams instead of binary-encoded (BE) values. Practical SC implementations require deterministic or pseudo-random number sequences which are optimally correlated to generate bitstreams and achieve accurate results. Unfortunately, the size of the search space makes manually designing optimally correlated number sequences a difficult task. To automate this design burden, we propose a synthesis formulation using mixed integer programming to automatically generate optimally correlated number sequences. In particular, our synthesis formulation improves the accuracy of arithmetic operations such as multiplication and squaring circuits by up to 2.5x and 20x respectively. We also show how our technique can be extended to scale to larger circuits.