Accelerating Large Scale Image Analyses on Parallel, CPU-GPU Equipped Systems

Teodoro, George; Kurç, Tahsin; Pan, Tony; Cooper, Lee; Kong, Jun; Widener, Patrick; Saltz, Joel

doi:10.1109/ipdps.2012.101

Cited by 35 publications

(34 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The specification of a parallel task consists of a set of input and output parameters with a corresponding data type (lines [16][17][32][33]) and a list of implementation variants. Each implementation variant defines the data distributions for the input and output parameters (lines [19][20][35][36] and either defines a runtime prediction (basic parallel tasks that are not further decomposed) or includes an entire task graph (complex parallel tasks). In the case of a basic parallel task, a set of symbolic runtime formulas is given (lines 22-25).…”

Section: ) Application Specificationmentioning

confidence: 99%

“…8 shows an example schedule for a hybrid platform. A schedule consists of communication operations (lines 2-12) and the execution of parallel tasks (lines [13][14][15][16][17][18][19] where each operation has an associated start time and an associated finish time. A communication operation additionally contains the identifiers of the source and target nodes in the underlying task graph and the definition of the source and target execution units.…”

Section: ) Schedule Specificationmentioning

confidence: 99%

See 1 more Smart Citation

Extending the Scheduling Toolkit SEParAT to Support Hybrid Parallel Platforms

Dümmler¹,

Schulze²

2016

IJCTE

View full text Add to dashboard Cite

Abstract-Current parallel platforms are increasingly equipped with additional accelerators leading to hybrid system architectures. Parallel applications for these platforms can be implemented using a task-based programming approach. Such an approach facilitates the exploitation of all available execution units including the processor cores and the accelerators. The execution of a task-based application requires scheduling decisions, which may be provided by a suitable scheduling tool.This article discusses the extensions of the scheduling toolkit SEParAT to support hybrid cluster architectures. In particular, it first defines the extended programming model for hybrid platforms and the corresponding scheduling problem. The second part of the article describes the integration of this model into SEParAT. A particular focus lies on the extension of SEParAT's input and output interfaces.

show abstract

Section: ) Application Specificationmentioning

confidence: 99%

Section: ) Schedule Specificationmentioning

confidence: 99%

Extending the Scheduling Toolkit SEParAT to Support Hybrid Parallel Platforms

Dümmler¹,

Schulze²

2016

IJCTE

View full text Add to dashboard Cite

show abstract

“…Most of the research on power management for the CPU-GPU subsystems has been geared towards optimizing for general purpose GPU (GPGPU) computing [4,15,3,8,17,11]. Suda and Ren [13] proposed a method to estimate the power consumption of CUDA kernels in discrete GPU subsystems for the purpose of energy optimization.…”

Section: Related Workmentioning

confidence: 99%

“…Unlike the performance controller output described in (14) and (15), this algorithm does not take into account any performance jitter since it does not use the reference constraint on the number of packets in the system queues. In addition, it considers the GPU and the CPU in isolation and does not guarantee selection of the frequency for the producer (CPU) that would enable sufficient performance to the consumer (GPU).…”

Section: Adding Energy Optimization To the Controlmentioning

confidence: 99%

A control-theoretic approach for energy efficient CPU-GPU subsystem in mobile platforms

Kadjo

Ayoub

Kishinevsky

et al. 2015

Proceedings of the 52nd Annual Design Automation Conference

View full text Add to dashboard Cite

This paper presents a control-theoretic approach to optimize the energy consumption of integrated CPU and GPU subsystems for graphic applications. It achieves this via a dynamic management of the CPU and GPU frequencies. To this end, we first model the interaction between the GPU and CPU as a queuing system. Second, we formulate a Multi-InputMulti-Output state-space closed loop control to ensure robustness and stability. We evaluated this control on an Intel Baytrail-based Android platform. Experimental evaluations show energy savings of 17.4% in the CPU-GPU subsystem with a low performance impact of 0.9%.

show abstract

“…The feature computations on objects are generally more regular and compute intensive than the operations in the segmentation stage. This characteristics of the feature computation operations lead to better GPU acceleration [13]. …”

Section: Application Parallelization For High Throughput Executionmentioning

confidence: 99%

High-throughput Analysis of Large Microscopy Image Datasets on CPU-GPU Cluster Platforms

Teodoro

Pan

Kurç

et al. 2013

2013 IEEE 27th International Symposium on Parallel and Distributed Processing

Self Cite

View full text Add to dashboard Cite

Analysis of large pathology image datasets offers significant opportunities for the investigation of disease morphology, but the resource requirements of analysis pipelines limit the scale of such studies. Motivated by a brain cancer study, we propose and evaluate a parallel image analysis application pipeline for high throughput computation of large datasets of high resolution pathology tissue images on distributed CPU-GPU platforms. To achieve efficient execution on these hybrid systems, we have built runtime support that allows us to express the cancer image analysis application as a hierarchical data processing pipeline. The application is implemented as a coarse-grain pipeline of stages, where each stage may be further partitioned into another pipeline of fine-grain operations. The fine-grain operations are efficiently managed and scheduled for computation on CPUs and GPUs using performance aware scheduling techniques along with several optimizations, including architecture aware process placement, data locality conscious task assignment, data prefetching, and asynchronous data copy. These optimizations are employed to maximize the utilization of the aggregate computing power of CPUs and GPUs and minimize data copy overheads. Our experimental evaluation shows that the cooperative use of CPUs and GPUs achieves significant improvements on top of GPU-only versions (up to 1.6×) and that the execution of the application as a set of fine-grain operations provides more opportunities for runtime optimizations and attains better performance than coarser-grain, monolithic implementations used in other works. An implementation of the cancer image analysis pipeline using the runtime support was able to process an image dataset consisting of 36,848 4Kx4K-pixel image tiles (about 1.8TB uncompressed) in less than 4 minutes (150 tiles/second) on 100 nodes of a state-of-the-art hybrid cluster system.

show abstract

Accelerating Large Scale Image Analyses on Parallel, CPU-GPU Equipped Systems

Cited by 35 publications

References 22 publications

Extending the Scheduling Toolkit SEParAT to Support Hybrid Parallel Platforms

Extending the Scheduling Toolkit SEParAT to Support Hybrid Parallel Platforms

A control-theoretic approach for energy efficient CPU-GPU subsystem in mobile platforms

High-throughput Analysis of Large Microscopy Image Datasets on CPU-GPU Cluster Platforms

Contact Info

Product

Resources

About