TritonCTS

------------------------------- WARNING --------------------------------

‼️ This version of TritonCTS has been retired from the OpenROAD flow. ‼️

‼️ You can find TritonCTS 2.0 here. ‼️

--------------------------------------------------------------------------

Original work

K. Han, A. B. Kahng and J. Li, "Optimal Generalized H-Tree Topology and Buffering for High-Performance and Low-Power Clock Distribution", IEEE Trans. on CAD (2018), doi:10.1109/TCAD.2018.2889756.

Many subsequent changes for open-sourcing were made by Mateus Fogaça.

Inputs and outputs

TritonCTS requires 5 input files and produces 2 output files. Refer to OpenROAD Flow and Notes for more information.

Inputs:

• Library characterization files
• Verilog with gate-level netlist
• Placed DEF with netlist
• Configuration file

Outputs:

• Placed DEF with clock buffers
• Verilog with clock buffers

Supported features / assumptions

• 1 clock source;

Validation

TritonCTS has been validated for the following list of platforms, tools and enablements.

Feature	Support
Operating system	CentOS 6
Compiler	GCC 4.4.7
Enablements	ST28 and TSMC16
TCL	Version 8.4.20
Python	Version 2.7.12
Lemon	Version 1.3.1

Tutorials

• Run the library characterization
• Compile and run TritonCTS

Example

Below we display the clock nets for the design jpeg_encoder implemented in TSMC65 with 24K instances, followed by TritonCTS routing topology.

TritonCTS generated clock nets for jpeg_encoder

TritonCTS clock tree topology for jpeg_encoder