Hardware accelerators are an effective solution to increase the performance of algorithms in a wide array of disciplines, from Data Science to Scientific Calculus. However, data scientists and mathematicians often do not have the required knowledge or time to fully exploit these accelerators, and they perceive them as difficult and frustrating to use.
Furthermore, Artificial Neural Networks are becoming the base of many of these applications, both in embedded and in server-class contexts. While Graphics Processing Units (GPUs) are predominantly used for training, solutions for inference often rely on Field Programmable Gate Arrays (FPGAs) since they are more flexible and cost-efficient in many scenarios.
The main goal is employing FPGA systems for processing huge quantities of data, that may be analyzed in real time, in order to provide a ready-to-use valuable solution to data scientists and mathematicians and applying FPGA systems to machine learning and AI applications to have a power efficient solution for complicated analytics problems.
Linux Systems Programming: Semaphores, Shared Memory, and Message Queues
ReWArDS - NECSTTechTalk 11/06/2020
1. ReWArDS
Reconfigurable hardWare for Artificial
intelligence and Data Science
Luca Stornaiuolo
luca.stornaiuolo@polimi.it
11/06/2020
NECST Talk – Sala Seminari
20. 51
FaceDetection on multi-PYNQ
Anna Maria Nestorov, Alberto Scolari,
Enrico Reggiani, Marco D. Santambrogio
Xilinx
Pynq
FPG
A
Xilinx
Pynq
FPG
A
Xilinx
Pynq
FPG
A
Xilinx
Pynq
FPG
A
Stage1 Stage2_2 Stage3 Stage4 Stage5
Xilinx
Pynq
FPGA
Xilinx
Pynq
FPGA
Stage2_1
Xilinx
Pynq
FPGA
Xilinx
Pynq
FPGA
Xilinx
Pynq
FPGA
Xilinx
Pynq
FPGA
21. 52
FaceDetection on multi-PYNQ
FPGA Single PYNQ vs ARM Single PYNQ: 14.7×
FPGA Distributed System vs ARM Distributed System: 39.5×
FPGA Distributed System vs GPU: 0,082×
22. 53
FaceDetection on multi-PYNQ
FPGA Single PYNQ vs ARM Single PYNQ: 16.2×
FPGA Distributed System vs ARM Distributed System: 44×
FPGA Distributed System vs GPU: 1.3×
FPGA Single PYNQ vs ARM Single PYNQ: 14.7×
FPGA Distributed System vs ARM Distributed System: 39.5×
FPGA Distributed System vs GPU: 0,082×
40. ReWArDS
Reconfigurable hardWare for Artificial
intelligence and Data Science
https://necst.it/
https://www.slideshare.net/necstlab
11/06/2020
NECST Talk – Sala Seminari