The HPC-DAG Task Model for Heterogeneous Real-Time Systems

Zahaf, Houssam-Eddine; Capodieci, Nicola; Cavicchioli, Roberto; Lipari, Giuseppe; Bertogna, Marko

doi:10.1109/tc.2020.3023169

Cited by 24 publications

(17 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We plan to further extend this work by considering the interference and the scheduling constraints that GPU kernels might experience when other are tasks running within the different processors/accelerators in the same computing platform. The goal is to integrate this work within the scheduling framework for heterogeneous architectures developed in [6]. We also plan to define more accurate kernels' categorization by accounting for architectural metrics such as cache-hit/miss rates, as profiled for each kernel.…”

Section: Discussionmentioning

confidence: 99%

Contention-Aware GPU Partitioning and Task-to-Partition Allocation for Real-Time Workloads

Zahaf¹,

Olmedo²,

Singh³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

In order to satisfy timing constraints, modern real-time applications require massively parallel accelerators such as General Purpose Graphic Processing Units (GPGPUs). Generation after generation, the number of computing clusters made available in novel GPU architectures is steadily increasing, hence, investigating suitable scheduling approaches is now mandatory. Such scheduling approaches are related to mapping different and concurrent compute kernels within the GPU computing clusters, hence grouping GPU computing clusters into schedulable partitions. In this paper we propose novel techniques to define GPU partitions; this allows us to define suitable task-to-partition allocation mechanisms in which tasks are GPU compute kernels featuring different timing requirements. Such mechanisms will take into account the interference that GPU kernels experience when running in overlapping time windows. Hence, an effective and simple way to quantify the magnitude of such interference is also presented. We demonstrate the efficiency of the proposed approaches against the classical techniques that considered the GPU as a single, nonpartitionable resource.

show abstract

Section: Discussionmentioning

confidence: 99%

Contention-Aware GPU Partitioning and Task-to-Partition Allocation for Real-Time Workloads

Zahaf¹,

Olmedo²,

Singh³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…This can be useful when modeling sub-tasks than can be executed on different engines with different execution costs. The selected edge may differ from an activation to another according the system state when the job is executed, in contrast to our previous work [1] where the same alternative configuration is selected during the system lifetime.…”

Section: B the Hpc-dag Task Modelmentioning

confidence: 96%

“…Cyber-physical embedded systems are increasingly complex and demand more and more powerful computational hardware platforms. COTS 1 vendors provide hardware platforms featuring multi-core CPU hosts with a number hardware accelerators, in order to support timing constraints of complex real-time applications with machine learning and image processing software modules.…”

Section: Introductionmentioning

confidence: 99%

“…[Houssam : Some bla bla about how good it is to express alternative, and conditionals]. [Houssam : Say that it is complex 1 Commercial Off The Shielf 2 The cost varies from some micro-seconds to couple of milliseconds.…”

Section: Introductionmentioning

confidence: 99%

“…task activation, deadlines, priority, etc. therefore the schedule is not necessarily known before run-time [1]. In contrast time-triggered systems, the system designer precompute/or totally compute in offline static execution schedule, i.e.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Building Time-Triggered Schedules for typed-DAG Tasks with alternative implementations

Zahaf¹,

Capodieci²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Hard real-time systems like image processing, autonomous driving, etc. require an increasing need of computational power that classical multi-core platforms can not provide, to fulfill with their timing constraints. Heterogeneous Instruction Set Architecture (ISA) platforms allow accelerating real-time workloads on application-specific cores (e.g. GPU, DSP, ASICs) etc. and are suitable for these applications. In addition, these platforms provide larger design choices as a given functionnality can be implemented onto several types of compute elements.HPC-DAG (Heterogeneous Parallel Directed Acyclic Graph) task model has been recently proposed to capture real-time workload execution on heterogeneous platforms. It expresses the ISA heterogeneity, and some specific characteristics of hardware accelerators, as the absence of preemption or costly preemption, alternative implementations and on-line conditional execution.In this paper, we propose a time-table scheduling approach to allocate and schedule a set of HPC-DAG tasks onto a set of heterogeneous cores, by the mean Integer Linear Programming (ILP). Our design allows to handle heterogeniety of resources, on-line execution costs, and a faster solving time, by exploring gradually the design space.

show abstract

An ILP representation of a DAG scheduling problem

Baruah

2021

Real-Time Syst

View full text Add to dashboard Cite

The HPC-DAG Task Model for Heterogeneous Real-Time Systems

Cited by 24 publications

References 30 publications

Contention-Aware GPU Partitioning and Task-to-Partition Allocation for Real-Time Workloads

Contention-Aware GPU Partitioning and Task-to-Partition Allocation for Real-Time Workloads

Building Time-Triggered Schedules for typed-DAG Tasks with alternative implementations

An ILP representation of a DAG scheduling problem

Contact Info

Product

Resources

About