Multicore-Optimized Wavefront Diamond Blocking for Optimizing Stencil Updates

Abstract

The importance of stencil-based algorithms in computational science has focused attention on optimized parallel implementations for multilevel cache-based processors. Temporal blocking schemes leverage the large bandwidth and low latency of caches to accelerate stencil updates and approach theoretical peak performance. A key ingredient is the reduction of data traffic across slow data paths, especially the main memory interface. In this work we combine the ideas of multicore wavefront temporal blocking and diamond tiling to arrive at stencil update schemes that show large reductions in memory pressure compared to existing approaches. The resulting schemes show performance advantages in bandwidth-starved situations, which are exacerbated by the high bytes per lattice update case of variable coefficients. Our thread groups concept provides a controllable trade-off between concurrency and memory usage, shifting the pressure between the memory interface and the CPU. We present performance results on a contemporary Intel processor.

Optimizations of Unstructured Aerodynamics Computations for Many-core Architectures

“…For example, seismic [63], stencil [64], [65], electromagnetic [66], molecular dynamics [67], Fast Multipole Methods [68], tensors [39], deep learning [69], [70], databases [49], [71], [72], big data [73], systems and graph engines [74], and many more.…”

Towards Fast Reverse Time Migration Kernels using Multi-threaded Wavefront Diamond Tiling

“…We use our cache blocking technique described in (Malas, Hager, Ltaief, Stengel, et al 2014). The implementation code is also available online (Malas 2015).…”

“…The parameter search space is reduced by our analytical model, which predicts the largest diamond size that fits in a given cache size. (Malas, Hager, Ltaief, Stengel, et al 2014)…”

“…We use the spatial blocking implementation described in (Malas, Hager, Ltaief, Stengel, et al 2014). Each lattice site update (LUP) in the spatially blocked code requires transferring 16Bytes from the main memory (i.e., update one array and read from two arrays, all in single precision).…”

DiamondTorre Algorithm for High-Performance Wave Modeling

“…The example of the latter is the wavefront diamond blocking method [16,17]. The advantage of DiamondTorre in comparison to these is an efficient use of the GPGPU architecture.…”