Resilient Chip Multiprocessors with Mixed-Grained Reconfigurability

Sourdis, Ioannis; Khan, Danish Anis; Malek, Alirad; Tzilis, Stavros; Smaragdos, Georgios; Strydis, Christos

doi:10.1109/mm.2015.7

Cited by 8 publications

(1 citation statement)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In summary, the relevance of this topic is caused by the need of the development of new or of improving the known tools of tasks mapping on the computational structure of dynamically reconfigured computational systems, which include hardware, constructive and technological limitations In the field of reconfigurable computing systems, the problems of varying the granularity level are described in the following topics: [7] -tasks mapping is done based on the idea of defining the needed amount of computational resources from the global reconfigurable computing space, which is shared among cores, which consists of predefined sets of finegrained and coarse-grained modules; [8,9] -the principle of reconfiguration based on the usage of hardware instruction set extensions (ISEs) to accelerate functional core;…”

Section: Introductionmentioning

confidence: 99%

The development of means of definition of the optimum ratio of computational algorithm and the reconfigurable structure

Klymenko

Holovko

Hilliaka

et al. 2016

EEJET

View full text Add to dashboard Cite

of the FPGA chips and can be effectively used to solve tasks of big size. Analysis of literature and the given problemsThere are a lot of methods and technologies for reconfiguration overhead reduction. The most known are resource reuse [1], caching of configuration data [2], forward reconfiguration [4], hardware tools of input/output acceleration [3] and optimization of the virtual structure of configuration data [5]. Each of them is based on the maximum possible reconfiguration acceleration of type "Best Effort" without any optimization of space solution and excluding hardware and technological limitations of the FPGA. Overcoming the space limitations of the FPGA is done with standard tools, for example, defragmentation of the computational surface of the FPGA [1, 2], loading out non-critical configuration [1, 2, 4], which brings an additional overhead to the reconfiguration.Granularity is considered as an effective space solution in the field of parallel computation [6]. The reconfiguration possibilities of the computing structure give perspectives to build an ideal computing structure for each task by varying the granularity level. Well-known methods and tools to vary

show abstract

Section: Introductionmentioning

confidence: 99%