Proceedings of the 2016 International Conference on Parallel Architectures and Compilation 2016
DOI: 10.1145/2967938.2967963
|View full text |Cite
|
Sign up to set email alerts
|

Abstract: System designers typically use well-studied benchmarks to evaluate and improve new architectures and compilers. We design tomorrow's systems based on yesterday's applications. In this paper we investigate an emerging application, 3D scene understanding, likely to be signi cant in the mobile space in the near future. Until now, this application could only run in real-time on desktop GPUs. In this work, we examine how it can be mapped to power constrained embedded systems. Key to our approach is the idea of incr… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

2
57
0

Year Published

2017
2017
2023
2023

Publication Types

Select...
5
3

Relationship

2
6

Authors

Journals

citations
Cited by 36 publications
(59 citation statements)
references
References 31 publications
2
57
0
Order By: Relevance
“…Offline control: Prior work by Bodin et al in PACT'2016 has studied approximation in SLAM under the assumption that the entire trajectory is known before the agent starts to move. Design space exploration for the given trajectory is performed by executing actual trials and the results are used to select good knob settings that are used for the entire trajectory [24]. This is an example of offline control since knob settings are determined once and for all before the computation begins.…”
Section: A Approximating Slammentioning
confidence: 99%
See 1 more Smart Citation
“…Offline control: Prior work by Bodin et al in PACT'2016 has studied approximation in SLAM under the assumption that the entire trajectory is known before the agent starts to move. Design space exploration for the given trajectory is performed by executing actual trials and the results are used to select good knob settings that are used for the entire trajectory [24]. This is an example of offline control since knob settings are determined once and for all before the computation begins.…”
Section: A Approximating Slammentioning
confidence: 99%
“…They utilize all frame pixels for reconstruction (in contrast, sparse SLAM algorithms utilize only a subset of features [11]) so there is a lot of data to process. The mapping phase requires repeated application of floating-point-intensive kernels such as stencil computations and filters [9], [12], [23] for reducing noise in incoming frames when operating in real-world environments [24], [25]. Computationally intensive algorithms such as iterative closest point [9] and Gauss-Newton minimization [14] are used in the localization process.…”
Section: Introductionmentioning
confidence: 99%
“…Moreover, building cost models with a large number of parameters may be expensive. Alternatively, active learning [61], [62], [60] can trade off exploration and exploitation mechanisms to give an approximated optimal configuration for workload execution in Polystore++ systems.…”
Section: Optimization Challengesmentioning
confidence: 99%
“…Bodin et al proposed the idea of design space exploration (DSE) which tries to optimise the hardware and software parameters to achieve some of the desired performance metrics, including ATE, ENE, and EXT [7]. The methodology of their work is based on quantifying these indices by playing the KinectFusion algorithm using the ICL-NUIM dataset on two different platforms and exploring the design space parameters.…”
Section: B Design Space Explorationmentioning
confidence: 99%
“…A SLAM algorithm which is useful for a high accuracy industrial mapping application is almost certainly not the right choice for a low power embedded platform like a drone. This has started to open up research on Design Space Exploration (DSE) [7], where a high level search is made through the possible operating parameters of a SLAM system, in order to find the combinations which work best in terms of an appropriate compromise between accuracy and efficiency. In general, the results of DSE are represented by a Pareto front of possible operating points, where each point on the front represents an optimum set of parameters given the desired performance metrics.…”
Section: Introductionmentioning
confidence: 99%