Machine Learning Approach for the Predicting Performance of SpMV on GPU

Benatia, Akrem; Ji, Weixing; Wang, Yizhuo; Feng, Shi

doi:10.1109/icpads.2016.0120

Cited by 13 publications

(17 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent studies such as [18] have explored the concept of using ML techniques such as multi-layer perceptron (MLP) and support vector regression (SVR) to model SpMV performance. We draw inspiration from these studies and extend them by using an ensemble of MLP for the performance modeling task.…”

Section: Results: Spmv Performance Modelingmentioning

confidence: 99%

See 1 more Smart Citation

Effective Machine Learning Based Format Selection and Performance Modeling for SpMV on GPUs

Nisa

Siegel

Sukumaran-Rajam

et al. 2018

EasyChair Preprints

View full text Add to dashboard Cite

Abstract-Sparse Matrix-Vector multiplication (SpMV) is a key kernel for many applications in computational science and data analytics. Several efforts have addressed the optimization of SpMV on GPUs, and a number of compact sparse-matrix representations have been considered for it. It has been observed that the sparsity pattern of non-zero elements in a sparse matrix has a significant impact on achieving SpMV performance. Further, no single sparse-matrix format is consistently the best across the range of sparse matrices encountered in practice. In this paper, we perform a comprehensive study that explores the use of Machine Learning to answer two questions: 1) Given an unseen sparse matrix, can we effectively predict the best format for SpMV on GPUs? 2) Can SpMV execution time for that matrix be predicted, for different matrix formats?By identifying a small set of sparse matrix features to use in training the ML models, we demonstrate that efficient prediction of the best format with ≈ 88% accuracy can be achieved when selecting between six well known sparsematrix formats on two GPU architectures (NVIDIA Pascal and Kepler), and ≈ 10% relative mean error (RME) with execution time prediction.

show abstract

Section: Results: Spmv Performance Modelingmentioning

confidence: 99%

“…Benatia et al [18] proposed to use multi-layer perceptron (MLP) and support vector regression (SVR) to predict the performance number of a SpMV operation. On average, it achieves low prediction error of 7% to 14% on a dataset of 1800 matrices.…”

Section: Related Workmentioning

confidence: 99%

Effective Machine Learning Based Format Selection and Performance Modeling for SpMV on GPUs

Nisa

Siegel

Sukumaran-Rajam

et al. 2018

EasyChair Preprints

View full text Add to dashboard Cite

show abstract

“…To address this issue, we used a machine learning approach to build performance models for the SpMV kernel under different sparse formats both on the CPU side and the GPU side. This approach was presented in our previous work (Benatia et al, 2016) for the sparse formats COO, CSR, ELL, and HYB on the GPU side. We found that it is straightforward to use the same approach to train performance models for the SpMV kernel using the COO and CSR formats on the CPU side.…”

Section: Sparse Matrix Partitioning For Spmv On Cpu-gpu Heterogenementioning

confidence: 99%

“…The off-line learning stage consists in training the performance models of the SpMV kernel under different sparse formats on the available processing units (CPU and GPU). The details of this stage have been previously presented in our work (Benatia et al, 2016) of which we summarize the main points in the following.…”

Section: Sparse Matrix Partitioning For Spmv On Cpu-gpu Heterogenementioning

confidence: 99%

“…For the CPU side, we used the implementation of the SpMV kernel under COO and CSR available in the Intel Math Kernel Library (MKL) (Gustafson and Greer, 2007). Given the rising complexity of multicore CPUs and GPUs hardware, we used a machine learning-based performance models of the SpMV kernel under different sparse formats built in our previous work (Benatia et al, 2016). These models are based only on very easy to compute sparsity features of the input sparse matrix.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Sparse matrix partitioning for optimizing SpMV on CPU-GPU heterogeneous platforms

Benatia

Wang

et al. 2019

The International Journal of High Performance Computing Applica

Self Cite

View full text Add to dashboard Cite

Sparse matrix–vector multiplication (SpMV) kernel dominates the computing cost in numerous applications. Most of the existing studies dedicated to improving this kernel have been targeting just one type of processing units, mainly multicore CPUs or graphics processing units (GPUs), and have not explored the potential of the recent, rapidly emerging, CPU-GPU heterogeneous platforms. To take full advantage of these heterogeneous systems, the input sparse matrix has to be partitioned on different available processing units. The partitioning problem is more challenging with the existence of many sparse formats whose performances depend both on the sparsity of the input matrix and the used hardware. Thus, the best performance does not only depend on how to partition the input sparse matrix but also on which sparse format to use for each partition. To address this challenge, we propose in this article a new CPU-GPU heterogeneous method for computing the SpMV kernel that combines between different sparse formats to achieve better performance and better utilization of CPU-GPU heterogeneous platforms. The proposed solution horizontally partitions the input matrix into multiple block-rows and predicts their best sparse formats using machine learning-based performance models. A mapping algorithm is then used to assign the block-rows to the CPU and GPU(s) available in the system. Our experimental results using real-world large unstructured sparse matrices on two different machines show a noticeable performance improvement.

show abstract

ZAKI: A Smart Method and Tool for Automatic Performance Optimization of Parallel SpMV Computations on Distributed Memory Machines

et al. 2019

View full text Add to dashboard Cite

Machine Learning Approach for the Predicting Performance of SpMV on GPU

Cited by 13 publications

References 18 publications

Effective Machine Learning Based Format Selection and Performance Modeling for SpMV on GPUs

Effective Machine Learning Based Format Selection and Performance Modeling for SpMV on GPUs

Sparse matrix partitioning for optimizing SpMV on CPU-GPU heterogeneous platforms

ZAKI: A Smart Method and Tool for Automatic Performance Optimization of Parallel SpMV Computations on Distributed Memory Machines

Contact Info

Product

Resources

About