This repository contains the source code of the architecture for following paper:
Martin Schoeberl, One-Way Shared Memory, accepted for publication, DATE 2018
We will continue to work on the project. Therefore, to reproduce the exact same results as presented in the DATE paper, use following tag:
03cacf682c01b35ceab72003488fadc7ae731cc2
This paper presents a novel idea for on-chip communication between processing cores. Each processor has a local memory (scratchpad memory (SPM)). A simple network-on-chip moves data from a region in the local SPM to a SPM in a remote processing core. This is done continually, which means that data is continually updated. This update is in one direction only, therefore we call it a one-way shared memory.
The implementation of the network-on-chip to support this one-way shared memory is cheap as there is no dynamic arbitration, no buffering, and no setup of messages to be sent. The hardware simply provides guarantees that data will be copied in a bounded time from one SPM to a remote SPM to support real-time systems. Higher level communication primitives can be built in software on top of this hardware primitive.
The evaluation of a prototype hardware implementation shows that the resource consumptions of the NoC routers and the network interface is very small compared to other NoC solutions. Furthermore, we provide bounds of the communication latency between processing cores.
The hardware is described in Chisel
and needs just a Java runtime and sbt installed. All other needed packages
will be automatically downloaded by sbt.
On a Mac with OS X sbt can be installed, assuming using homebrew
as your package management tool, with:
brew install sbt
On a Debian based Linux machine, such as Ubuntu, you can install sbt according to the instructions from sbt downloadwith:
For the Chisel based tests a compiler with gcc like interface is needed.
A small Makefile contains shortcuts to run different tasks.
A single router can be tested with:
make test
Verilog code can be generated with:
make hardware
The code is placed in folder generated and can by synthesized with your
favorite synthesize tool. A Quartus project is available in folder quartus.