Floorplanning is a critical early stage of VLSI physical design, and its quality directly impacts interconnect wirelength, chip performance, and downstream design efficiency. This article presents a multi-stage fixed-outline floorplanning framework that combines non-convex and convex optimization. The framework operates in three successive phases. In the initial floorplanning phase, a quadratic placement optimizes module connectivity to produce a topology-aware starting configuration with strong net clustering, albeit with significant module overlaps. In the global floorplanning phase, Adam-based projected gradient method is employed to solve non-convex optimization that minimizes wirelength and overlap. In the legalization phase, a log-transformed model exploiting horizontal and vertical constraint graphs eliminates residual overlaps under a convex formulation. Experiments on the MCNC, GSRC and HB+ benchmarks demonstrate that the framework achieves state-of-the-art wirelength quality, reducing average HPWL by at least 1\% and 5\% compared to state-of-the-art floorplanners on tested benchmarks.
Multi-task inference models share a single backbone across diverse tasks, yet execute identical computation regardless of which task is active - wasting energy and cycles on task-irrelevant operations. We observe that the task command, typically available before inference begins,…
Structured State Space Models (SSMs), such as Mamba, enable efficient long-sequence modeling with linear time complexity. Recent implementations realize this capability through Structured State Space Duality (SSD), which transforms recursive state evolution into matrix-form compu…
The increasing deployment of large language models (LLMs) has magnified the computational and memory bottlenecks of autoregressive decoding, where low compute intensity and bandwidth-bound kernels dominate inference cost. Weight pruning offers a promising remedy, but existing met…
With the rapid adoption of long-context large language models (LLMs), the continuously growing KV cache during decoding has become the critical memory bottleneck. To tackle this challenge, we propose HiKV, a novel algorithm-hardware co-design that exploits KV cache redundancy thr…
Transformers are widely used across many domains, including natural language processing, computer vision, web search, and DNA sequence analysis. Given their broad applicability, improving the performance of transformer models is critical. However, the high volume of data movement…
By leveraging standard RISC-V extensions, namely Zfh (scalar float16) and Zvfh (vector float16), this work proposes an open-source framework to enable complete on-device training on resource-constrained RISC-V single-core. Our approach allows memory footprint reduction by about 5…