2012 15th Euromicro Conference on Digital System Design 2012
DOI: 10.1109/dsd.2012.71
|View full text |Cite
|
Sign up to set email alerts
|

HEAP: A Highly Efficient Adaptive Multi-processor Framework

Abstract: Writing parallel code is difficult, especially when starting from a sequential reference implementation. Our research efforts, as demonstrated in this paper, face this challenge directly by providing an innovative toolset that helps software developers profile and parallelize an existing sequential implementation, by exploiting top-level pipeline-style parallelism. The innovation of our approach is based on the facts that a) we use both automatic and profiling-driven estimates of the available parallelism, b) … Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1

Citation Types

0
2
0
1

Year Published

2013
2013
2019
2019

Publication Types

Select...
3
3

Relationship

0
6

Authors

Journals

citations
Cited by 6 publications
(3 citation statements)
references
References 10 publications
0
2
0
1
Order By: Relevance
“…NNs can be implemented in SW (running on either general (CPU) or specialized (GPU) processors), or in HW (either reconfigurable (FPGA) or application-specific (ASIC)). FPGAs can bring significant speed and energy improvements for both high-end implementations [39]- [42] and for energy-and processing-constrained embedded devices [27], [43], [44], while preserving programmability. Since most NN computations are embarrassingly parallel, they can considerably benefit from FPGAs flexibility (e.g., resource allocation, scheduling, data flow, data width).…”
Section: Neural Network Implementationmentioning
confidence: 99%
“…NNs can be implemented in SW (running on either general (CPU) or specialized (GPU) processors), or in HW (either reconfigurable (FPGA) or application-specific (ASIC)). FPGAs can bring significant speed and energy improvements for both high-end implementations [39]- [42] and for energy-and processing-constrained embedded devices [27], [43], [44], while preserving programmability. Since most NN computations are embarrassingly parallel, they can considerably benefit from FPGAs flexibility (e.g., resource allocation, scheduling, data flow, data width).…”
Section: Neural Network Implementationmentioning
confidence: 99%
“…Then, the effectiveness of the use of the toolset will be demonstrated, both in terms of simplification of the parallelization task for low skill users as well as the acceleration obtained on a stereo vision application of practical interest. [16,15] is a free software project designed to support the developers of various skill levels to parallelize legacy sequential C code that can include complex control structures, pointer operations, and dynamic memory allocation. ParTools was designed to facilitate the discovery of both task and data parallelization opportunities and can be used for any parallelization technique.…”
Section: Pharaon Workflow For Paral-lelizationmentioning
confidence: 99%
“…A lo largo del proyecto se han realizado numerosas publicaciones en la literatura científica que pueden consultarse para tener un mejor conocimiento del mismo, todas ellas están disponibles en el sitio web del proyecto [HEAP], algunas de las más destacadas son [LLP + 13], donde se presentan el conjutno de herramientas desarrolladas dentro del proyecto, y [RK12], donde se presenta un protocolo para el manteniemiento de la información almacenada en las memorias de los sistemas multinúcleo.…”
Section: Proyecto Heapunclassified