Program optimization carving for GPU computing

Ryoo, Shane; Rodrigues, Christopher; Stone, Sam S.; Stratton, John A.; Ueng, Sain-Zee; Baghsorkhi, Sara S.; Hwu, Wen-mei W.

doi:10.1016/j.jpdc.2008.05.011

Cited by 98 publications

(47 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared to a set of recent studies on performance autotuning by empirical search [11,12,13,14,15], we provide an alternative optimization solution. Certainly search-based approaches are a powerful tool for optimization, but we note two disadvantages of such an approach.…”

Section: Introductionmentioning

confidence: 99%

A quantitative performance analysis model for GPU architectures

Zhang

Owens

2011

2011 IEEE 17th International Symposium on High Performance Computer Architecture

200

100

View full text Add to dashboard Cite

We develop a microbenchmark-based performance model for NVIDIA GeForce 200-series GPUs. Our model identifies GPU program bottlenecks and quantitatively analyzes performance, and thus allows programmers and architects to predict the benefits of potential program optimizations and architectural improvements. In particular, we use a microbenchmark-based approach to develop a throughput model for three major components of GPU execution time: the instruction pipeline, shared memory access, and global memory access. Because our model is based on the GPU's native instruction set, we can predict performance with a 5-15% error. To demonstrate the usefulness of the model, we analyze three representative real-world and already highly-optimized programs: dense matrix multiply, tridiagonal systems solver, and sparse matrix vector multiply. The model provides us detailed quantitative analysis on performance, allowing us to understand the configuration of the fastest dense matrix multiply implementation and to optimize the tridiagonal solver and sparse matrix vector multiply by 60% and 18% respectively. Furthermore, our model applied to analysis on these codes allows us to suggest architectural improvements on hardware resource allocation, avoiding bank conflicts, block scheduling, and memory transaction granularity.

show abstract

Section: Introductionmentioning

confidence: 99%

A quantitative performance analysis model for GPU architectures

Zhang

Owens

2011

2011 IEEE 17th International Symposium on High Performance Computer Architecture

200

100

View full text Add to dashboard Cite

show abstract

“…This empirical optimization with varying configurations is a typical approach in GPU programming because a general performance prediction model for a GPU architecture is not available due to the complexity of its parallel programming model [2,28,29]. Our experiments showed that the thread block with size 16 Â 26 yields the best performance in the block-level facet processing implementation.…”

Section: Thread-block Configurationmentioning

confidence: 91%

“…Later, Ryoo et al developed their ideas further on efficient GPU program optimization, and revealed that optimizing an application for maximum performance goes beyond simply application of a set of optimization techniques to code. The difficulty comes from the fact that the interactions among the underlying architectural and programming model constraints affect performance in a non-linear fashion [29]. They modeled GPU programming for maximum performance as a multivariable optimization problem.…”

Section: Related Workmentioning

confidence: 99%

Performance analysis of a novel GPU computation-to-core mapping scheme for robust facet image modeling

Park

Cao

Watson

et al. 2012

J Real-Time Image Proc

View full text Add to dashboard Cite

Modern graphics processing units (GPUs) are commodity data-parallel coprocessors capable of high performance computation and data throughput. It is well known that the GPUs are ideal implementation platforms for image processing applications. However, the level of efforts and expertise to optimize the application performance is still substantial. This paper investigates the computation-to-core mapping strategies to probe the efficiency and scalability of the robust facet image modeling algorithm on GPUs. Our fine-grained computation-to-core mapping scheme achieves a significant performance gain over the standard pixel-wise mapping scheme. With indepth performance comparisons across the two different mapping schemes, we analyze the impact of the level of parallelism on the GPU computation and suggest two principles for optimizing future image processing applications on the GPU platform.

show abstract

“…The limited research in kernel code optimization has resulted in the underutilization of these devices. Similar programming paradigms such as CUDA have been studied for many years to enable their use on these computing devices [8,9,16]. However, optimization techniques on CUDA are specified on heterogeneous multicore architectures with nVidia GPUs, which may be inappropriate for various OpenCL-compatible devices.…”

Section: Figmentioning

confidence: 99%

Methods for Optimizing OpenCL Applications on Heterogeneous Multicore Architectures

Chuand¹,

Hsiao²,

Li³

2013

Appl. Math. Inf. Sci.

View full text Add to dashboard Cite

Heterogeneous multicore architectures with CPU and add-on GPUs or streaming processors are now widely used in computer systems. These GPUs provide substantially more computation capability and memory bandwidth compared to traditional multi-cores. Also, because they are highly programmable, they provide the computational performance needed for realistic graphics rendering. Applications with general computations can also be leveraged onto these GPUs. This study discusses the architectures of these highly efficient GPUs and applies a unified programming standard called OpenCL to fully utilize their capabilities. Despite their great potential, applications of these GPUs are challenging because of their diverse underlying architectural characteristics. In this study, several optimizing techniques are applied on OpenCL-compatible heterogeneous multicore architectures to achieve thread-level and data-level parallelisms. The architectural implications of these techniques are discussed. Finally, optimization principles for these architectures will be are proposed. The experimental reveal average speedups of 24 and 430 for non-optimized and optimized kernels, respectively.

show abstract

Program optimization carving for GPU computing

Cited by 98 publications

References 17 publications

A quantitative performance analysis model for GPU architectures

A quantitative performance analysis model for GPU architectures

Performance analysis of a novel GPU computation-to-core mapping scheme for robust facet image modeling

Methods for Optimizing OpenCL Applications on Heterogeneous Multicore Architectures

Contact Info

Product

Resources

About