Publications

During my bachelor and master, I published some paper while “working” at NECSTLab, a laboratory inside Politecnico di Milano which focuses on systems (OS and hardware design) and security.

I mostly worked on the acceleration of algorithms (for computer vision, such as filtering and CNNs), and in the end did a thesis on distributed systems and serverless acceleration using FPGAs.

I don’t do research anymore (didn’t pursue a PhD afterall), but here you can find all links to my previous works.

Master Thesis

BlastFunction: an FPGA-as-a-Service system for accelerated serverless computing

For my master thesis I developed a distributed system for enabling the acceleration of serverless functions using FPGAs. The project included:

• a custom OpenCL implementation for executing workloads on remote nodes
• a module for interacting with Altera’s FPGAs using the remote OpenCL library above, deployed on every node in the cluster
• a kubernetes operator for allocating FPGAs to serverless functions (using OpenFAAS) in order to share the same FPGA between multiple functions (using the same hardware module)

I learned a lot on low-level programming (interacting with shared memory and threads to allow OpenCL remoting) and kubernetes. The most rewarding aspect was seeing the project working on a 3-nodes Kubernetes cluster, in which multiple serverless functions used the same FPGAs concurrently to share the workload, improving the overall throughput of the system!

Articles

Blastfunction: A full-stack framework bringing fpga hardware acceleration to cloud-native applications
Andrea Damiani , Giorgia Fiscaletti , Marco Bacis , Rolando Brondolin , Marco Domenico Santambrogio
    link

BlastFunction: an FPGA-as-a-Service system for accelerated serverless computing
Marco Bacis , Rolando Brondolin , Marco Domenico Santambrogio
  pdf   link

A Framework with Cloud Integration for CNN Acceleration on FPGA Devices
Niccolò Raspa , Giuseppe Natale , Marco Bacis , Marco Domenico Santambrogio
  pdf   link

Convolutional Neural Networks on Dataflow Engines
Nils Voss , Marco Bacis , Oskar Mencer , Georgi Gaydadjiev , Wayne Luk
  pdf   link

On how to design dataflow FPGA-based accelerators for Convolutional Neural Networks
Giuseppe Natale , Marco Bacis , Marco Domenico Santambrogio
  pdf   link

A Pipelined and Scalable Dataflow Implementation of Convolutional Neural Networks on FPGA
Marco Bacis , Giuseppe Natale , Emanuele Del Sozzo , Marco Domenico Santambrogio
  pdf   link

Software Implementation and Hardware Acceleration of Retinal Vessel Segmentation for Diabetic Retinopathy Screening Tests
Lara Cavinato , Irene Fidone , Marco Bacis , Emanuele Del Sozzo , Gianluca Carlo Durelli , Marco Domenico Santambrogio
poster   pdf   link

Patents

An FPGA-As-A-Service System for Accelerated Serverless Computing
International Publication Number: WO2020234792A1
Authors: Marco Bacis, Rolando Brondolin, Marco Domenico Santambrogio