The rise in GPU compute speed has outpaced improvements in host-to-device memory transfer speeds, despite the advent of shared-memory superchips. Consequently, memory transfer times now constitute an increasingly large fraction of total time-to-solution, compelling developers to compress GPU kernel input and output data into compact, minimal formats prior to GPU-offloading. Complementary to existing work on GPU- and compute-friendly data arrangements, this work argues that significant attention should furthermore be paid to the data organisation on the host such that quick data rearrangements and offloading become feasible. We study a Smoothed Particle Hydrodynamics solver and propose memory layout strategies for host-side particle data that are particularly well-suited to GPU-offloading. Specifically, we advocate splitting classic array-of-struct data structures into a struct-of-arrays-of-structs arrangement, in which each logical struct decomposes into substructs determined by kernel read/write access patterns and attribute types. Splitting a monolithic particle struct into several bespoke, finer-grained structs can reduce the time required to pack data to and from buffers by ~20%-40%, lowering total time spent on GPU-offloading by ~12%-25%.

翻译：暂无翻译