Frequently-Asked Questions (FAQ)

General

What is LibreLane?

Originally started as version 2.0 of OpenLane, LibreLane is a piece of software for ASIC implementation, initially developed by Efabless Corporation, but its currently maintained under the stewardship of the FOSSi Foundation.

Version 1.0 of OpenLane was a simple but stable and battle-tested flow, primarily intended for implementing designs for Efabless’s MPW programs.

LibreLane reimagines it as not just a single flow, but as an infrastructure by which flows can be implemented. An OpenLane 1-compatible flow named “Classic” exists, but is in beta pending further silicon validation.

Should I use OpenLane or LibreLane?

Since the liquidation of Efabless Corporation, the authors of OpenLane have lost control of the trademark, and OpenLane’s original developers currently maintain LibreLane under the stewardship of the FOSSi Foundation.

OpenLane 1.0 as hosted under The OpenROAD Project is a super-stable version that was used for countless tapeouts. It is in maintenance mode, but is not recommended at all for new designs and only exists to enable old designs to be re-taped out.

How is LibreLane different from OpenROAD?

OpenROAD is one of many utilities used by LibreLane, which integrates it and many other tools in order to achieve a full RTL-to-GDSII flow.

OpenROAD is primarily developed by The OpenROAD Project, which involves many corporations and academic institutions (primarily the University of California, San Diego, Parallax Software, and Precision Innovations). LibreLane, on the other hand, was primarily developed at Efabless Corporation and is currently maintained by the community under the stewardship of the FOSSi Foundation.

Can I use LibreLane with my (company’s) proprietary PDK?

In general, yes, but you would have to create LibreLane configuration files for said PDK. See Using non-Ciel PDKs for more info.

Is LibreLane silicon-proven?

OpenLane 1.0 has been used for countless verified tapeouts, including more or less every open-source design on the Google MPW shuttles.

LibreLane/OpenLane 2.0 has been silicon-proven in a relatively more limited capacity, having been used for all Tiny Tapeout shuttles since 3.5 and a number of internal tape-outs at Efabless Corporation.

Why should I use LibreLane over other open-source RTL-to-GDS-II flows?

Point of Comparison

OpenROAD Flow Scripts

SiliconCompiler

OpenLane

LibreLane

Architecture

Monolithic

Plugin-based

Monolithic

Plugin-based

Configuration

Tcl Files

Python Files

JSON/Tcl Files

JSON/Tcl/Python Files

Programming Language

GNU Make

Python

Tcl

Type-checked Python

Maintainer

The OpenROAD Project

ZeroASIC

Efabless

FOSSi Foundation

Dependencies

Separate (Build Scripts)

Separate (Build Scripts)

Bundled

Bundled

Cloud Service

No

Yes

No

No (Planned)

Proprietary Tool Support

No

Yes

No

Yes (with Plugins)

Pre-built Binaries

.deb (x86-64)

N/A

Docker (x86-64, ARM64)

* Natively through Nix: Linux and macOS (x86-64, ARM64)
* Docker (x86-64, ARM64)

Open-Source PDK Support

sky130, gf180mcu, nangate45, asap7, ihp-sg13g2

sky130, gf180mcu, asap7

sky130, gf180mcu

sky130, gf180mcu, ihp-sg13g2 (beta)

Community Examples

Limited

Limited

9+ public multi-project wafer shuttles with Efabless

Backwards Compatible with OL Examples

Setup

Why does running LibreLane on Windows require the Windows Subsystem for Linux (WSL)?

In short, a lot of the open-source EDA tools LibreLane relies on presume a Linux-based environment, so they would be non-trivial to port to Windows as we’d have to make sure every tool both compiles and behaves as expected on Windows.

Why do you use Nix?

Nix allows us to create a near-perfectly reproducible environment on macOS and all Linux distributions with just a single set of scripts, and the rich community ecosystem surrounding it also enables us to distribute these environments in their entirety to end-users.

Similar to Docker, this mostly eliminates variables related to the user’s environment, although unlike Docker, it maintains integration with the user’s filesystem, doesn’t add a virtualization penalty on macOS, and does not require the entire image to be re-downloaded every time an update occurs.