完全模拟代用于解决矩阵方程与内存计算
Jiancong Li1, Yibai Xue1, Yi Li1,2
1School of Integrated Circuits, Hubei Key Laboratory for Advanced Memories, Huazhong University of Science and Technology, Wuhan 430074, China.
本研究提出了一种用于解决矩阵方程的内存式内存计算方法,在科学计算任务中实现高精度和显著的速度和能量改进.
科学领域:
- 计算机科学 计算机科学
- 材料科学 材料科学 材料科学
- 应用数学 应用数学 应用数学
背景情况:
- 记忆内存计算为科学计算提供了潜力,但模拟不准确性给高精度,低能耗解决方案带来了挑战.
- 解决矩阵方程是许多科学和工程学科的基础,通常需要大量的计算资源.
研究的目的:
- 引入一种新型的记忆矩阵方程解答器,它利用模拟计算来加快解决方案.
- 展示一种混合模拟-数字方法,以降低开销,实现高精度的结果.
- 用实验性的异质计算系统验证解决者在复杂科学问题上的表现.
主要方法:
- 开发了一个模拟矩阵方程解答器,在模拟领域执行数学代.
- 实施了混合方法,将快速模拟近似与数字精细化相结合,以最大限度地减少数字到模拟转换.
- 在异质计算系统上实验验验证了溶解器,模拟了扩散方程和PN连接平衡.
主要成果:
- 当与数字精细化相结合时,记忆式解决器实现了与软件相当的精度 (10^-12误差).
- 与传统的数字处理单元相比,解决方案速度提高了128倍.
- 实现了160倍的计算能源消耗降低.
结论:
- 这种memristive解决器通过在模拟领域进行代来显著加速科学计算.
- 混合方法有效平衡速度,能源效率和高精度,克服模拟不准确性.
- 在未来的高性能科学计算应用中利用不精确的模拟设备的基础.
更多相关视频
06:45Design and Application of a Fault Detection Method Based on Adaptive Filters and Rotational Speed Estimation for an Electro-Hydrostatic Actuator
Published on: October 28, 2022
05:30Large Scale Energy Efficient Sensor Network Routing Using a Quantum Processor Unit
Published on: September 8, 2023
相关概念视频
Ampere-Maxwell's Law: Problem-Solving
To solve the problem, we can use the equations from the analysis of an RC circuit and Maxwell's version of Ampère's law.
For the first part of...
Ampere's Law: Problem-Solving
Specific steps need to be considered while calculating the symmetric magnetic field distribution...
Euler's Formula to Columns: Problem Solving
The system comprises two vertical rigid bars, AB and BC,...
Cartesian Form for Vector Formulation
Vector Algebra: Method of Components
In many applications, the magnitudes and directions of...
Biot-Savart Law: Problem-Solving
Consider a mobile phone battery bank as a source of steady current, which flows through the wire connected between the two. What is the magnitude of the magnetic field created by this current at a field point P?
To estimate the magnitude of the total magnetic field, we first consider a small current element of length dl, at a distance r from the field point. Now the following...
