# OpenACMv2: An Accuracy-Constrained Co-Optimization Framework for Approximate DCiM

Source: [OpenACMv2: An Accuracy-Constrained Co-Optimization Framework for Approximate DCiM](<https://www.sota2.com/research/paper/69b78e7a0636e5bf4afbdbc6>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-09T05:00:07.006Z

Authors: Yiqi Zhou, Yue Yuan, Yikai Wang, Bohao Liu, Qinxin Mei, Zhuohua Liu, Shan Shen, Wei Xing, Daying Sun, Li Li, Guozhu Liu

Published: 2026-03-13T14:50:52.000Z

## Abstract

Digital Compute-in-Memory (DCiM) accelerates neural networks by reducing data movement. Approximate DCiM can further improve power-performance-area (PPA), but demands accuracy-constrained co-optimization across coupled architecture and transistor-level choices. Building on OpenYield, we introduce Accuracy-Constrained Co-Optimization (ACCO) and present OpenACMv2, an open framework that operationalizes ACCO via two-level optimization: (1) accuracy-constrained architecture search of compressor combinations and SRAM macro parameters, driven by a fast GNN-based surrogate for PPA and error; and (2) variation- and PVT-aware transistor sizing for standard cells and SRAM bitcells using Monte Carlo. By decoupling ACCO into architecture-level exploration and circuit-level sizing, OpenACMv2 integrates classic single- and multi-objective optimizers to deliver strong PPA-accuracy tradeoffs and robust convergence. The workflow is compatible with FreePDK45 and OpenROAD, supporting reproducible evaluation and easy adoption. Experiments show that the proposed two-level ACCO framework achieves most of its accuracy-constrained efficiency gain at Level-1 through architecture exploration, delivering roughly 50%+ PDP reduction, while Level-2 transistor-level optimization provides a further single-digit PDP improvement while preserving accuracy, enabling rapid "what-if" exploration for approximate DCiM. The framework is available on GitHub (https://github.com/ShenShan123/OpenACM).

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Image Blending on Test0 Cameraman + Boat 1.0 (test)](<https://www.sota2.com/research/sota/image-blending-on-test0-cameraman-plus-boat-1-0-test.md>): Image Blending; Test0 Cameraman + Boat 1.0 (test); MRED: 0.00425.

- [Image Blending on Test Cameraman + Lake 1.0 (test)](<https://www.sota2.com/research/sota/image-blending-on-test-cameraman-plus-lake-1-0-test.md>): Image Blending; Test Cameraman + Lake 1.0 (test); MRED: 0.00425.

- [Image Classification on CIFAR-10 (test) (MRED and PDP Metrics)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-mred-and-pdp-metrics.md>): Image Classification; CIFAR-10 (test); MRED: 0.000177.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.13042>)

[PDF](<https://arxiv.org/pdf/2603.13042.pdf>)