Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userAccelerating GPU-Based Out-of-Core Stencil Computation with On-the-Fly Compression
144
0
0.0
(
0
)
Added by
Jingcheng Shen
Publication date
2021
fields
Informatics Engineering
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
Stencil computation is an important class of scientific applications that can be efficiently executed by graphics processing units (GPUs). Out-of-core approach helps run large scale stencil codes that process data with sizes larger than the limited capacity of GPU memory. However, the performance of the GPU-based out-of-core stencil computation is always limited by the data transfer between the CPU and GPU. Many optimizations have been explored to reduce such data transfer, but the study on the use of on-the-fly compression techniques is far from sufficient. In this study, we propose a method that accelerates the GPU-based out-of-core stencil computation with on-the-fly compression. We introduce a novel data compression approach that solves the data dependency between two contiguous decomposed data blocks. We also modify a widely used GPU-based compression library to support pipelining that overlaps CPU/GPU data transfer with GPU computation. Experimental results show that the proposed method achieved a speedup of 1.2x compared the method without compression. Moreover, although the precision loss involved by compression increased with the number of time steps, the precision loss was trivial up to 4,320 time steps, demonstrating the usefulness of the proposed method.
rate research
Read More
The growth of data to be processed in the Oil & Gas industry matches the requirements imposed by evolving algorithms based on stencil computations, such as Full Waveform Inversion and Reverse Time Migration. Graphical processing units (GPUs) are an attractive architectural target for stencil computations because of its high degree of data parallelism. However, the rapid architectural and technological progression makes it difficult for even the most proficient programmers to remain up-to-date with the technological advances at a micro-architectural level. In this work, we present an extension for an open source compiler designed to produce highly optimized finite difference kernels for use in inversion methods named Devito. We embed it with the Oxford Parallel Domain Specific Language (OP-DSL) in order to enable automatic code generation for GPU architectures from a high-level representation. We aim to enable users coding in a symbolic representation level to effortlessly get their implementations leveraged by the processing capacities of GPU architectures. The implemented backend is evaluated on a NVIDIA GTX Titan Z, and on a NVIDIA Tesla V100 in terms of operational intensity through the roof-line model for varying space-order discretization levels of 3D acoustic isotropic wave propagation stencil kernels with and without symbolic optimizations. It achieves approximately 63% of V100s peak performance and 24% of Titan Zs peak performance for stencil kernels over grids with 256 points. Our study reveals that improving memory usage should be the most efficient strategy for leveraging the performance of the implemented solution on the evaluated architectures.
Stencil computation is one of the most important kernels in various scientific and engineering applications. A variety of work has focused on vectorization and tiling techniques, aiming at exploiting the in-core data parallelism and data locality respectively. In this paper, the downsides of existing vectorization schemes are analyzed. Briefly, they either incur data alignment conflicts or hurt the data locality when integrated with tiling. Then we propose a novel transpose layout to preserve the data locality for tiling and reduce the data reorganization overhead for vectorization simultaneously. To further improve the data reuse at the register level, a time loop unroll-and-jam strategy is designed to perform multistep stencil computation along the time dimension. Experimental results on the AVX-2 and AVX-512 CPUs show that our approach obtains a competitive performance.
GAMER is a GPU-accelerated Adaptive-MEsh-Refinement code for astrophysical simulations. In this work, two further extensions of the code are reported. First, we have implemented the MUSCL-Hancock method with the Roes Riemann solver for the hydrodynamic evolution, by which the accuracy, overall performance and the GPU versus CPU speed-up factor are improved. Second, we have implemented the out-of-core computation, which utilizes the large storage space of multiple hard disks as the additional run-time virtual memory and permits an extremely large problem to be solved in a relatively small-size GPU cluster. The communication overhead associated with the data transfer between the parallel hard disks and the main memory is carefully reduced by overlapping it with the CPU/GPU computations.
Over the years, accelerating neural networks with quantization has been widely studied. Unfortunately, prior efforts with diverse precisions (e.g., 1-bit weights and 2-bit activations) are usually restricted by limited precision support on GPUs (e.g., int1 and int4). To break such restrictions, we introduce the first Arbitrary Precision Neural Network framework (APNN-TC) to fully exploit quantization benefits on Ampere GPU Tensor Cores. Specifically, APNN-TC first incorporates a novel emulation algorithm to support arbitrary short bit-width computation with int1 compute primitives and XOR/AND Boolean operations. Second, APNN-TC integrates arbitrary precision layer designs to efficiently map our emulation algorithm to Tensor Cores with novel batching strategies and specialized memory organization. Third, APNN-TC embodies a novel arbitrary precision NN design to minimize memory access across layers and further improve performance. Extensive evaluations show that APNN-TC can achieve significant speedup over CUTLASS kernels and various NN models, such as ResNet and VGG.
We propose a distance between continuous-time Markov chains (CTMCs) and study the problem of computing it by comparing three different algorithmic methodologies: iterative, linear program, and on-the-fly. In a work presented at FoSSaCS12, Chen et al. characterized the bisimilarity distance of Desharnais et al. between discrete-time Markov chains as an optimal solution of a linear program that can be solved by using the ellipsoid method. Inspired by their result, we propose a novel linear program characterization to compute the distance in the continuous-time setting. Differently from previous proposals, ours has a number of constraints that is bounded by a polynomial in the size of the CTMC. This, in particular, proves that the distance we propose can be computed in polynomial time. Despite its theoretical importance, the proposed linear program characterization turns out to be inefficient in practice. Nevertheless, driven by the encouraging results of our previous work presented at TACAS13, we propose an efficient on-the-fly algorithm, which, unlike the other mentioned solutions, computes the distances between two given states avoiding an exhaustive exploration of the state space. This technique works by successively refining over-approximations of the target distances using a greedy strategy, which ensures that the state space is further explored only when the current approximations are improved. Tests performed on a consistent set of (pseudo)randomly generated CTMCs show that our algorithm improves, on average, the efficiency of the corresponding iterative and linear program methods with orders of magnitude.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
