US 12,443,785 B1
Placing hard macros using machine learning predictions trained on different circuit designs
Yi-Min Jiang, Campbell, CA (US); Xiang Gao, San Jose, CA (US); Lixin Shao, Sunnyvale, CA (US); and Pedja Raspopovic, Cary, NC (US)
Assigned to Synopsys, Inc., Sunnyvale, CA (US)
Filed by Synopsys, Inc., Mountain View, CA (US)
Filed on Sep. 26, 2022, as Appl. No. 17/953,110.
Claims priority of provisional application 63/248,953, filed on Sep. 27, 2021.
Int. Cl. G06F 30/398 (2020.01); G06F 30/31 (2020.01); G06F 30/392 (2020.01)
CPC G06F 30/398 (2020.01) [G06F 30/31 (2020.01); G06F 30/392 (2020.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
accessing a set of candidate macro placements of hard macros within a circuit design; and
estimating, by a processing device, a quality metric for each candidate macro placement, comprising:
predicting a plurality of model-specific estimates of the quality metric, comprising applying different machine learning models to predict each of the different model-specific estimates, wherein the machine learning models applied for different model-specific estimates are trained using sets of source macro placements for different source circuit designs; and
combining the model-specific estimates of the quality metric, wherein the combining is based on an applicability of (a) the set of source macro placements for each model-specific estimate to (b) the set of candidate macro placements.