Activity-aware, frequency-prioritised register placement: controlled-replication data (SCL CP8D 0.8 µm and SKY130)
Description
Designs, stimulus, switching-activity data, placement scripts and measured results for a controlled replication of activity-aware, frequency-prioritised register placement. Three flows that differ only in register placement are compared: baseline (no region constraint), proposed (frequency-prioritised regions with the eight most active registers packed first) and a control with the frequency order scrambled. Stage I is a four-bit adder on SCL CP8D 0.8 µm and SkyWater SKY130; Stage II is a five-domain PDM decimation front end (464 registers) on SCL CP8D, replicated at three placement densities with two ablations. The dataset accompanies the article "Activity-Aware, Frequency-Prioritised Register Placement Reduces Clock-Network Power but Not Total Power: A Controlled Replication With a Scrambled Control at Two Design Scales". The SCL CP8D PDK is the property of the Semiconductor Laboratory, Mohali, and is used under its PDK agreement. No PDK file and no file carrying PDK technology data is included: the CP8D layouts, netlists, parasitic files, tool logs and library cell names are withheld, and holders of the PDK may request them from the corresponding author. For CP8D the dataset gives the measured metrics, the power reports and a results table; the SKY130 layouts are included in full. README.md maps each folder to the results of the article.
Files
Steps to reproduce
1. Read README.md. 2. Simulate the RTL with Icarus Verilog using tb/ and the stimulus in sim2/, and compare with the golden outputs in sim2/. 3. Extract register activity with scripts/extract_activity*.py. 4. Run the placement method (scripts/place_regions.py: modes pack, group, project) in OpenROAD; the SKY130 Stage I flows use scripts/run_pnr.py. 5. The CP8D flows need the SCL CP8D PDK and the withheld files; set the placeholders SCL_DFF_CELL and SCL_BUF_CELL to the library cells. 6. Every Stage II number of the article is in results/stage2_cp8d/results_summary.csv; the percentages behind the figures are in paper/fig/numbers.json.