• Corpus ID: 38066628

Operating Systems Challenges for GPU Resource Management

@inproceedings{Kato2011OperatingSC,
  title={Operating Systems Challenges for GPU Resource Management},
  author={Shinpei Kato and Scott A. Brandt and Yutaka Ishikawa and Ragunathan Raj Rajkumar},
  year={2011}
}
The graphics processing unit (GPU) is becoming a very powerful platform to accelerate graphics and data-paralle l compute-intensive applications. [] Key Method A list of operating systems challenge s is also provided to highlight future directions of this rese arch domain, including specific ideas of GPU scheduling for realtime systems. Our preliminary evaluation demonstrates tha t the performance of open-source software is competitive wit h that of proprietary software, and hence operating systems r esearch…
GPU Virtualization and Scheduling Methods
TLDR
An extensive and in-depth survey of GPU virtualization techniques and their scheduling methods is presented and a perspective on the challenges and opportunities for virtualization of heterogeneous computing environments is delivered.
Towards adaptive GPU resource management for embedded real-time systems
TLDR
Two conceptual frameworks for GPU applications to adjust their task execution times based on total workload are presented, which enable smart GPU resource management when many applications share GPU resources while the workloads of those applications vary.
Leveraging Hybrid Hardware in New Ways - The GPU Paging Cache
TLDR
This paper presents an operating system extension that allows leveraging the GPU accelerator memory for operating system purposes, and utilizes graphics card memory as cache for virtual memory pages, which can improve the overall system responsiveness, especially under heavy load.
Leveraging Hybrid Hardware in New Ways - The GPU Paging Cache
TLDR
This paper presents an operating system extension that allows leveraging the GPU accelerator memory for operating system purposes, and utilizes graphics card memory as cache for virtual memory pages, which can improve the overall system responsiveness, especially under heavy load.
A GPU Kernel Transactionization Scheme for Preemptive Priority Scheduling
TLDR
This work proposes an approach to transactionize GPU kernels at the operating system (OS) level and limits the preemption delay to 18 μs in the 99.9th percentile with an average delay in execution time less than 10 % in most cases for high-priority tasks under a heavy load.
Towards using the Graphics Processing Unit (GPU) for embedded systems
TLDR
This paper looks at requirements inherent in the process and power industries where it is believed that the GPU has the potential to be a useful and natural element in future embedded system architectures.
Proceedings of the 4th Workshop on Adaptive and Reconfigurable Embedded Systems Workshop Organization Program Chairs Program Committee Steering Committee  Network Challenges in Cyber-physical Systems Session 1: Adaptive Resource Management  towards Adaptive Gpu Resource Management for Embedded Rea
TLDR
Two conceptual frameworks for GPU applications to adjust their task execution times based on total workload are presented, which enable smart GPU resource management when many applications share GPU resources while the workloads of those applications vary.
EDGE: Event-Driven GPU Execution
TLDR
An event-driven GPU execution model that enables non-CPU devices to directly launch preconfigured tasks on a GPU without CPU interaction is proposed, and it is estimated that EDGE can reduce the kernel launch latency by 4.4x compared to the baseline CPU-launched approach.
Accelerated test execution using GPUs
TLDR
This work proposes a technique that accelerates test execution, allowing test suites to run in a fraction of the original time, by parallel execution with a Graphics Processing Unit (GPU).
...
1
2
...

References

SHOWING 1-10 OF 39 REFERENCES
Exploring the multiple-GPU design space
  • Dana Schaa, D. Kaeli
  • Computer Science
    2009 IEEE International Symposium on Parallel & Distributed Processing
  • 2009
TLDR
This paper considers the benefits of running on multiple (parallel) GPUs to provide further orders of performance speedup and develops a methodology that allows developers to accurately predict execution time for GPU applications while varying the number and configuration of the GPUs, and the size of the input data set.
GViM: GPU-accelerated virtual machines
TLDR
GViM is presented, a system designed for virtualizing and managing the resources of a general purpose system accelerated by graphics processors and how such accelerators can be virtualized without additional hardware support.
TimeGraph: GPU Scheduling for Real-Time Multi-Tasking Environments
TLDR
TimeGraph is presented, a real-time GPU scheduler at the device-driver level for protecting important GPU workloads from performance interference and supports two priority-based scheduling policies in order to address the tradeoff between response times and throughput introduced by the asynchronous and non-preemptive nature of GPU processing.
Dynamic load balancing on single- and multi-GPU systems
TLDR
Experimental results show that the proposed task-based dynamic load-balancing solution can utilize the hardware more efficiently than the CUDA scheduler for unbalanced workload, and achieves near-linear speedup, load balance, and significant performance improvement over techniques based on standard CUDA APIs.
Resource Sharing in GPU-Accelerated Windowing Systems
TLDR
This paper proposes two protocols that enable application tasks to efficiently share the GPU resource in the X Window System and identifies and addresses resource-sharing problems raised in GPU-accelerated windowing systems.
Enabling Task Parallelism in the CUDA Scheduler
TLDR
An issue queue that merges workloads that would underutilize GPU processing resources such that they can be run concurrently on an NVIDIA GPU is proposed and throughput is increased in all cases where the GPU would have been underused by a single kernel.
StoreGPU: exploiting graphics processing units to accelerate distributed storage systems
TLDR
StoreGPU is designed, a library that accelerates a number of hashing based primitives popular in distributed storage system implementations that enable up to eight-fold performance gains on synthetic benchmarks as well as on a high-level application: the online similarity detection between large data files.
Real-Time Multiprocessor Systems with GPUs ∗
TLDR
This work presents two real-time analysis methods for such an integration into a soft real- time multiprocessor system and shows that a GPU can be exploited to achieve greater levels of total system performance.
A GPU accelerated storage system
TLDR
The design of a storage system prototype that uses GPU offloading to accelerate a number of computationally intensive primitives based on hashing is presented and the results show that this technique can bring tangible performance gains without negatively impacting the performance of concurrently running applications.
GPU virtualization on VMware's hosted I/O architecture
TLDR
This paper describes in detail the specific GPU virtualization architecture developed for VMware's hosted products (VMware Workstation and VMware Fusion) and finds that taking advantage of hardware acceleration significantly closes the gap between pure emulation and native, but that different implementations and host graphics stacks show distinct variation.
...
1
2
3
4
...