Evaluating the Impact of SDC on the GMRES Iterative Solver
Evaluating the Impact of SDC on the GMRES Iterative Solver
复制标题
评估 SDC 对 GMRES 迭代求解器的影响
DOI：
10.1109/ipdps.2014.123
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
F. Mueller
中科院分区：
文献类型：
--
作者：
James Elliott;M. Hoemmen;F. Mueller
不断增加的并行性和晶体管密度，以及越来越严格的能量和峰值功率限制，可能会迫使应用程序代码偶尔暴露不正确的计算或存储。静默数据损坏（SDC）可能很少发生，但一个SDC足以使迭代线性求解器等数值算法停止向正确答案前进。因此，我们重点研究迭代线性求解器GMRES对单个瞬态SDC的弹性。我们推导出廉价的检查来检测GMRES中SDC的影响，这些检查适用于更一般的SDC模型，而不是假设位翻转。我们的实验表明，当使用GMRES作为内外迭代的内部求解器时，它可以在计算密集的正交化阶段“运行”几乎任何大小的SDC。也就是说，它使用错误的数据得到正确的答案，而不需要任何回滚。那些它无法通过的SDCs，会被我们的检测方案捕捉到。
Increasing parallelism and transistor density, along with increasingly tighter energy and peak power constraints, may force exposure of occasionally incorrect computation or storage to application codes. Silent data corruption (SDC) will likely be infrequent, yet one SDC suffices to make numerical algorithms like iterative linear solvers cease progress towards the correct answer. Thus, we focus on resilience of the iterative linear solver GMRES to a single transient SDC. We derive inexpensive checks to detect the effects of an SDC in GMRES that work for a more general SDC model than presuming a bit flip. Our experiments show that when GMRES is used as the inner solver of an inner-outer iteration, it can "run through" SDC of almost any magnitude in the computationally intensive orthogonalization phase. That is, it gets the right answer using faulty data without any required roll back. Those SDCs which it cannot run through, get caught by our detection scheme.