2D/3D IC Energy/Carbon Evaluation Tool

This is the artifact of the following papers:

• '3D IC Architecture Evaluation and Optimization with Digital Compute-in-Memory Designs' presented in ACM/IEEE International Symposium on Low Power Electronics and Design (ISLPED), 2024.
• 'Energy-/Carbon-Aware Evaluation and Optimization of 3-D IC Architecture with Digital Compute-in-Memory Designs' presented in IEEE Journal of Exploratory Solid-State Computational Devices and Circuits (JxCDC), 2024.

Updates

• [Aug. 06, 2024]: Initial version of the artifact is published together with the camera-ready version of our ISLPED paper.
• [Jan. 16, 2025]: New version of the artifact is published with embodied carbon estimation feature, which is presented in JXCDC paper.

How to use?

• To get the argument information, type python FILENAME.py -h for the following python scripts.
• In the workload folder, 8 different AI workloads are loaded as csv files. You can add your own model dimensions following the other file formats. Refer to workload.py which reads and makes class from the csv file.
• In the parameters folder, there are parameters related to embodied carbon estimation, which is derived from ACT and 3D-Carbon.
• Following scripts are the examples to represent the simulation results. You can also get more detailed data from the simulator in terms of energy, area, cycles and utilizations.

e_breakdown.py

Generates energy breakdown of hardware given PE array size, buffer size and workload.

• Usage: python e_breakdown.py -x xdim -y ydim --buf bufsize --wk workload

cycle_util.py

Generates cycles and utilization of hardware given PE array size, buffer size and workload.

• Usage: python cycle_util.py -x xdim -y ydim --buf bufsize --wk workload

energy_sweep.py

Generates energy breakdown across 8 different AI models and different sizes of PE array and buffer.

• Usage: python energy_sweep.py
• If you change the number of workloads, then the output figure might crash. In that case, change the dimension of the subplots at the end of the script.

c_breakdown.py

Generates carbon breakdown across 6 different systolic array/DCIM based 2D/3D architectures.

• Usage: python c_breakdown.py -x xdim -y ydim --buf bufsize

Authors

Hyung Joon Byun, Udit Gupta, and Jae-sun Seo. School of ECE, Cornell Tech.

Cite Us

• 3D IC Architecture Evaluation and Optimization with Digital Compute-in-Memory Designs (Byun et al., ISLPED, 2024).
• Energy-/Carbon-Aware Evaluation and Optimization of 3-D IC Architecture with Digital Compute-in-Memory Designs (Byun et al., JXCDC, 2024).

Acknowledgement

This work was supported in part by the Center for the Co-Design of Cognitive Systems (CoCoSys) in JUMP 2.0, a Semiconductor Research Corporation (SRC) Program sponsored by the Defense Advanced Research Projects Agency (DARPA).

References

• U. Gupta et al., "ACT: Designing sustainable computer systems with an architectural carbon modeling tool", Proc. ISCA, pp. 784-799, 2022.
• Y. Zhao, Y. Zhao, C. Wan and Y. Lin, "3D-carbon: An analytical carbon modeling tool for 3D and 2.5D integrated circuits", Proc. 61st ACM/IEEE Design Automation Conference, 2024.