Constraint Solving on Hybrid Systems

Roque, Pedro; Pedro, Vasco

doi:10.1007/978-3-030-00801-7_1

Cited by 1 publication

(3 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [24] we described PHACT architecture, its load balancing techniques and compared its performance against Gecode (and PaCCS [21]) on solving some CSPs. Those CSPs were loaded to Gecode through FlatZinc models, but at the time, PHACT Figure 1.…”

Section: Resultsmentioning

confidence: 99%

“…As mentioned in section III and described in [24], PHACT is also capable of using GPUs, MICs and any other device compatible with OpenCL in order to speed the solving process up. The chart in Figure 4 shows the speedups achieved when using only an Intel I7-4870HQ CPU (4 cores, 8 threads), or this CPU and one Nvidia Geforce 980M GTX GPU (12 SMs).…”

Section: Resultsmentioning

confidence: 99%

“…PHACT implements a dynamic load balancing technique which tailors the size of the blocks of sub-search spaces to match the performance of each device solving the current problem, in relative terms, when compared to the performance of the other devices. This feature is extensively described in [24].…”

Section: Load Balancingmentioning

confidence: 92%

See 2 more Smart Citations

Improving Constraint Solving on Parallel Hybrid Systems

Roque¹,

Pedro²,

Díaz³

et al. 2018

2018 IEEE 30th International Conference on Tools With Artificial Intelligence (ICTAI)

Self Cite

View full text Add to dashboard Cite

Recently, we developed the Parallel Heterogeneous Architecture Constraint Toolkit (PHACT), which is a multithreaded constraint solver capable of using all the available devices which are compatible with OpenCL, in order to speed up the constraint satisfaction process. In this article, we introduce an evolution of PHACT which includes the ability to execute FlatZinc and MiniZinc models, as well as architectural improvements which boost the performance in solving CSPs, especially when using GPUs.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%