Skip to content

StanfordAHA/canal

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Build Status

Design Philosophy

In canal, the hardware is represented as a directed graph (DiGraph). In principle each node in the graph will be turn into a hardware component based on the property of the node. The edge will be transformed into a wire.

One thing to notice is that staged generation happens on the graph level. That is, passes have to be operated on the DiGraph, rather than the hardware generator itself. This allows better analysis and reduce the computation overhead since magma is slow when constructing large hardware. In canal, different bit width routing network will have different DiGraph by design. This allows us to share the common passes on different DiGraph while constructing on the same interconnect

How to install

pip install .

Due to the magma backend, Python 3.7.2+ is required to install canal

Relationship with CGRA_PNR (Thunder/Cyclone)

The internal DiGraph representation is identical to Cyclone. Due to the requirement of pure-Python implementation, canal re-implemented the graph representation with more utility functions. The PnR information is interchangeable with Thunder and Cyclone and thus can be used in place route directly.

There is a python utility package called archipelago, that can take Interconnect object in canal and perform place and route directly. To install archipelago, you can use

pip install archipelago

It will use pre-built Python native wheels to speed up place and route.


A new kratos-based of canal is here.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages