Construction of Performance Model of Tile CAQR and Performance Result of the Implementation
Construction of Performance Model of Tile CAQR and Performance Result of the Implementation
复制标题
Tile CAQR性能模型构建及实施性能结果
DOI：
10.1109/mcsoc.2017.18
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
Tomohiro Suzuki
中科院分区：
文献类型：
--
作者：
M. Takayanagi;Tomohiro Suzuki
可以通过异步执行许多细粒度任务来利用高度并行的计算资源。矩阵分解的tile算法可以生成许多细粒度的任务，因此适合现代多核/多核架构。然而，该算法的性能在很大程度上取决于贴图的大小。我们在集群系统上以OpenMP/MPI混合方式实现了平铺算法，并构建了一个性能模型，该模型通过测量实现中简单计算内核的性能来调整平铺大小。在本报告中，我们在K计算机上测试了我们的高矩阵和瘦矩阵的避免通信的平铺QR实现，并演示了性能模型的适用性。
Highly parallel computational resources can be exploited by asynchronously executing many fine-grained tasks. The tile algorithm for matrix decomposition can generate many fine-grained tasks, so is suitable for modern multicore/manycore architectures. However, the performance of this algorithm significantly depends on the tile size. We implement the tile algorithm in OpenMP/MPI hybrid fashion on a cluster system and construct a performance model that tunes the tile size by measuring the performance of simple computational kernels in our implementation. In this report, we test our communication-avoiding tile QR implementation for tall and skinny matrices on the K computer, and demonstrate the applicability of the performance model.