Mitigating Silent Data Corruptions in Integer Matrix Products: Toward Reliable Multimedia Computing on Unreliable Hardware
Mitigating Silent Data Corruptions in Integer Matrix Products: Toward Reliable Multimedia Computing on Unreliable Hardware
复制标题
DOI：
10.1109/tcsvt.2016.2589622
复制
发表时间：
2017-11
影响因子：
8.4
通讯作者：
Ijeoma Anarado;M. A. Anam;F. Verdicchio;Y. Andreopoulos
中科院分区：
文献类型：
--
作者：
Ijeoma Anarado;M. A. Anam;F. Verdicchio;Y. Andreopoulos
通用矩阵乘法(GEMM)例程由许多处理整数输入的信息检索、机器学习和对象识别系统的计算和内存密集型部分组成。因此，确保整数GEMM计算对静默数据损坏(SDC)保持稳健至关重要，SDC源于意外的电压或频率超标，或其他硬件非理想情况。本文介绍了一种新的基于数值填充概念的SDC抑制方法。我们的方法与所有现有方法之间的关键区别是在输出的数字表示中产生多余的结果，而不是作为一组单独的校验和。重要的是，与众所周知的基于算法的GEMM容错方法不同，该方法能够可靠地检测GEMM计算结果中绝大多数可能的SDC的位置。在Intel i7-4578U 3 GHz处理器上运行的最先进的图像和视频检索算法中，针对视觉描述符匹配的电压缩放整数GEMM计算的实验研究表明，与所有其他替代方案相比，基于数字打包的SDC缓解导致了类似或更低的执行和能耗开销。
The generic matrix multiply (GEMM) routine comprises the compute and memory-intensive parts of many information retrieval, machine learning, and object recognition systems that process integer inputs. Therefore, it is of paramount importance to ensure that integer GEMM computations remain robust to silent data corruptions (SDCs), which stem from accidental voltage or frequency overscaling, or other hardware nonidealities. In this paper, we introduce a new method for SDC mitigation based on the concept of numerical packing. The key difference between our approach and all existing methods is the production of redundant results within the numerical representation of the outputs, rather than as a separate set of checksums. Importantly, unlike well-known algorithm-based fault tolerance approaches for GEMM, the proposed approach can reliably detect the locations of the vast majority of all possible SDCs in the results of GEMM computations. An experimental investigation of voltage-scaled integer GEMM computations for visual descriptor matching within state-of-the-art image and video retrieval algorithms running on an Intel i7-4578U 3 GHz processor shows that SDC mitigation based on numerical packing leads to comparable or lower execution and energy-consumption overhead in comparison with all other alternatives.