TetrisG-SDK: Efficient Convolutional Layer Mapping with Adaptive Windows and Grouped Convolutions for Fast In-Memory Computing

2026-04-28

AI Summary

This work addresses the limitations of existing CIM SDK mapping approaches, which are confined to single-macro optimization and fail to exploit inter-macro parallelism, resulting in suboptimal efficiency. To overcome these challenges, we propose TetrisG-SDK, a novel framework that synergistically integrates an adaptive windowing mechanism, grouped convolutions, and a multi-CIM-macro parallel mapping strategy to jointly optimize array utilization and computational latency under fixed hardware budgets. System-level evaluation using a validated CIM simulator demonstrates that, compared to VWC-SDK, TetrisG-SDK achieves 1.2–1.3× higher inference throughput, reduces system latency and energy consumption by up to 2.4× and 1.7×, respectively, and lowers the EDAP metric by 36%–70%, all while preserving near-lossless model accuracy.

Abstract

Shifted-and-Duplicated-Kernel (SDK) mapping has emerged as an effective strategy to accelerate convolutional layers on compute-in-memory (CIM) hardware. However, existing SDK variants (e.g., VWC-SDK) merely optimize mapping for a single CIM macro, leaving inter-macro parallelism unexplored. Moreover, their mapping methodologies are still suboptimal. To address these limitations, we present TetrisG-SDK, a novel framework that employs adaptive windows to boost mapping performance. The proposed windows accommodate more input channels, increase array utilization at marginal space, and adapt to different channel depths. More importantly, TetrisG-SDK reduces compute latency by searching for optimal window configurations across multiple CIM macros with a fixed hardware budget. Besides, it incorporates grouped convolution to further decrease computing cycles while maintaining near-lossless model accuracy. In addition, TetrisG-SDK integrates a validated CIM hardware simulator to provide accurate system-/application-level estimations of latency, area and energy. Compared to the single-macro VWC-SDK, the proposed framework achieves a speed-up by 1.2x, 1.3x, and 1.3x for CNN8, GoogLeNet Inception, and DenseNet40 models, respectively. When deployed on the simulator, it reduces system-level latency and energy by 2.4x and 1.7x for CNN8, 1.3x and 1.2x for Inception, and 1.3x and 1.6x for DenseNet40, respectively. When leveraging macro-level parallelism, TetrisG-SDK reduces the Energy-Delay-Area-Product (EDAP) by 70% for CNN8, 68% for Inception, and 36% for DenseNet40 compared to its non-grouped counterpart. These results manifest that TetrisG-SDK is a promising solution to efficiently mapping convolutional layers on CIM hardware.

Problem

• Compute-in-Memory
• Convolutional Layer Mapping
• Inter-macro Parallelism
• SDK Mapping
• Hardware Utilization

Innovation

• adaptive windows
• grouped convolutions
• compute-in-memory
• convolutional layer mapping
• macro-level parallelism