Heterogeneous Von Neumann/Dataflow Microprocessors.
General-purpose processors (GPPs), which traditionally rely on a Von Neumann-based execution model, incur burdensome power overheads, largely due to the need to dynamically extract parallelism and maintain precise state. Further, it is extremely difficult to improve their performance without increas...
| Publicado en: | Communications of the ACM Vol. 62; no. 6; pp. 82 - 91 |
|---|---|
| Autores principales: | , , |
| Formato: | Artículo |
| Publicado: |
Association for Computing Machinery
Jun2019
|
| Materias: | |
| Acceso en línea: | Ver este registro en EBSCOhost |
| fields | @attributes: recordID: 1 pdfLink: plink: https://search.ebscohost.com/login.aspx?direct=true&db=hlh&AN=137697934&site=ehost-live header: @attributes: shortDbName: hlh uiTerm: 137697934 longDbName: Humanities International Complete uiTag: AN controlInfo: bkinfo: jinfo: jid: 00010782 ACM jtl: Communications of the ACM issn: 00010782 maglogo: N pubinfo: dt: Jun2019 vid: 62 iid: 6 pid: 68 pub: Association for Computing Machinery artinfo: ui: 137697934 10.1145/3323923 ppf: 82 ppct: 9 formats: tig: atl: Heterogeneous Von Neumann/Dataflow Microprocessors. aug: au: Nowatzki, Tony Gangadhar, Vinay Sankaralingam, Karthikeyan affil: University of California, Los Angeles, Los Angeles, CA, USA University of Wisconsin - Madison, Madison, WI, USA su: Microprocessor design & construction Microprocessor performance Microprocessor energy consumption Von Neumann architecture (Computers) Data flow computing sug: subj: Microprocessor design & construction Microprocessor performance Microprocessor energy consumption Von Neumann architecture (Computers) Data flow computing ab: General-purpose processors (GPPs), which traditionally rely on a Von Neumann-based execution model, incur burdensome power overheads, largely due to the need to dynamically extract parallelism and maintain precise state. Further, it is extremely difficult to improve their performance without increasing energy usage. Decades-old explicit-dataflow architectures eliminate many Von Neumann overheads, but have not been successful as stand-alone alternatives because of poor performance on certain workloads, due to insufficient control speculation and communication overheads. We observe a synergy between out-of-order (OOO) and explicit-dataflow processors, whereby dynamically switching between them according to the behavior of program phases can greatly improve performance and energy efficiency. This work studies the potential of such a paradigm of heterogeneous execution models, by developing a specialization engine for explicit-dataflow (SEED) and integrating it with a standard out-of-order (OOO) core. When integrated with a dual-issue OOO, it becomes both faster (1.33×) and dramatically more energy efficient (1.70×). Integrated with an in-order core, it becomes faster than even a dual-issue OOO, with twice the energy efficiency. pubtype: Periodical doctype: Article src: R language: English refInfo: copyright: @attributes: flag: Y dt: @attributes: year: 2019 holdings: @attributes: islocal: N |
|---|