Cost-aware function migration in heterogeneous systems

Kicherer, Mario; Buchty, Rainer; Karl, Wolfgang

doi:10.1145/1944862.1944883

Cited by 14 publications

(21 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As we can see, a PEPPHER container not only keeps track of data copies on different memory units but also helps in minimizing the data communication between different memory units by delaying the communication until it becomes necessary. In this case, only 2 copy operations of data are made in the shown program execution instead of 7 copy operations which are required if one considers each component call independently, as done in, e.g., Kicherer et al [132,133].…”

Section: Inter-component Parallelismmentioning

confidence: 99%

“…In this case, the prediction function predict returns the expected best implementation name, managing the complete selection process internally. Many performance models can fit in here, including models based on decision tree, SVM and one presented in [132]. When comparing different variants for implementation selection, potential data transfer cost must also be considered besides the computation cost predicted for each variant.…”

Section: Performance Modelling Apimentioning

confidence: 99%

“…It is based on the model from Kicherer et al [132]. The model aims for execution contexts with limited or no repetitions in performance relevant properties (i.e., almost every execution has a different value for performance relevant properties).…”

Section: Convex Modelmentioning

confidence: 99%

“…Examples of such bulk operations include applying sorting, image filters, or compressions/decompressions over multiple files in a loop. In these applications, performance models can be learned at runtime [127,132]. Our framework provides support for online feedback which allows performance models to be learned/improved during a program execution.…”

Section: Online Learningmentioning

confidence: 99%

“…The selection is normally done by estimating the execution time with the help of a performance model for each component implementation. The performance model could be trained empirically either offline [132] or generated online [20] by the runtime system; it could also be an analytical model supplied by the implementation writer [131]. There exist two variations of this greedy local scheduling and selection technique:…”

Section: Greedy Scheduling and Selection Using Heftmentioning

confidence: 99%

See 4 more Smart Citations

Performance-aware component composition for GPU-based systems

Dastgeer¹

2014

View full text Add to dashboard Cite

This thesis adresses issues associated with efficiently programming modern heterogeneous GPU-based systems, containing multicore CPUs and one or more programmable Graphics Processing Units (GPUs). We use ideas from component-based programming to address programming, performance and portability issues of these heterogeneous systems. Specifically, we present three approaches that all use the idea of having multiple implementations for each computation; performance is achieved/retained either a) by selecting a suitable implementation for each computation on a given platform or b) by dividing the computation work across different implementations running on CPU and GPU devices in parallel.In the first approach, we work on a skeleton programming library (SkePU) that provides high-level abstraction while making intelligent implementation selection decisions underneath either before or during the actual program execution. In the second approach, we develop a composition tool that parses extra information (metadata) from XML files, makes certain decisions offline, and, in the end, generates code for making the final decisions at runtime. The third approach is a framework that uses source-code annotations and program analysis to generate code for the runtime library to make the selection decision at runtime. With a generic performance modeling API alongside program analysis capabilities, it supports online tuning as well as complex program transformations.These approaches differ in terms of genericity, intrusiveness, capabilities and knowledge about the program source-code; however, they all demonstrate usefulness of component programming techniques for programming GPU-based systems. With experimental evaluation, we demonstrate how all three approaches, although different in their own way, provide good performance on different GPU-based systems for a variety of applications.This work has been supported by two EU FP7 projects (PEP-PHER, EXCESS) and by SeRC. Populärvetenskaplig sammanfattningAtt få varje generation av datorer att fungera snabbareär viktigt för samhä-llets utveckling och tillväxt. Traditionellt hade de flesta datorer bara en general-purpose processor (den så kallade CPU:n) som bara kunde exekvera en beräkningsuppgift i taget. Under det senasteårtiondet har dock flerkärniga och mångkärniga processorer blivit vanliga, och datorer har också blivit mer heterogena. Ett modernt datorsystem innehåller vanligtvis flerän en CPU, tillsammans med specialprocessorer såsom grafikprocessorer (GPU:er) som ar anpassade för att kunna exekvera vissa typer av beräkningar effektivarë an CPU:er. Vi kallar ett sådant system med en eller flera GPU:er för ett GPU-baserat system. GPU:er i sådana system har sitt eget separata minne, och för att kunna köra en beräkning på en GPU så behöver man vanligtvis flytta all indata till GPU:ns minne och sedan hämta tillbaka resultatet när beräkningenär klar.Programmeringen av GPU-baserade systemär icke-trivialt av flera anledningar: (1) CPU:er och GPU:er kräver olika programmeringsexper...

show abstract

Section: Inter-component Parallelismmentioning

confidence: 99%

Section: Performance Modelling Apimentioning

confidence: 99%

Section: Convex Modelmentioning

confidence: 99%

Section: Online Learningmentioning

confidence: 99%

Section: Greedy Scheduling and Selection Using Heftmentioning

confidence: 99%

See 3 more Smart Citations

Performance-aware component composition for GPU-based systems

Dastgeer¹

2014

View full text Add to dashboard Cite

show abstract

A survey on hardware‐aware and heterogeneous computing on multicore processors and accelerators

Buchty

Heuveline

Karl

et al. 2011

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

Abstract. The paradigm shift towards multicore technologies is offering a great potential of computational power for scientific and industrial applications. It is, however, posing considerable challenges to software development. This problem is impaired by increasing heterogeneity of hardware platforms on both, processor level, and by adding dedicated accelerators. Performance gains for data-and compute-intensive applications can currently only be achieved by exploiting coarse-and fine-grained parallelism on all system levels, and improved scalability with respect to constantly increasing core counts. A key methodology is hardware-aware computing where in all production steps explicit knowledge of processor, memory, and network details is profitably utilized. In this work we provide a survey on current multicore and accelerator technologies. We outline architectural features and show, how these features are exposed to the programmer and how they can be beneficially utilized in the application-mapping process. In particular, we characterize the discrepancy to conventional parallel platforms with respect to hierarchical memory sub-systems, fine-grained parallelism on several system levels, and chip-and system-level heterogeneity. We motivate the necessity of hardware-aware computing and summarize the challenges arising from high-performance heterogeneous computing. Furthermore, we investigate the interaction of hardware and application characteristics for selected applications in numerical simulation.

show abstract