Search for: All records

Integrated circuit design is a highly complex and time-consuming process. Leveraging large language models (LLMs) for automating hardware design generation is receiving increasing attention. A prominent challenge is that the inherent structure of the text is overlooked during the training process. Existing efforts focus on supervised fine-tuning LLMs to acquire specialized knowledge in hardware design, without considering the conflict between LLMs' linear data processing and the structural nature inherent in hardware design. In this work, we propose a novel LLM-based reinforcement learning (RL) framework that integrates Abstract Syntax Trees (ASTs) and Data Flow Graphs (DFGs). Our approach enhances the accuracy of generated hardware code by capturing the syntactic and semantic structures of hardware designs. Experimental results show that the SFT-RL model integrated with Text, AST, and DFG achieves notable improvements: a 12.57% increase on VerilogEval-Human and a 5.49% increase on VerilogEval-Machine, outperforming GPT-4; a 14.29% improvement on RTLLM, approaching GPT-4. 

The increasing complexity of integrated circuit design requires customizing Power, Performance, and Area (PPA) metrics according to different application demands. However, most engineers cannot anticipate requirements early in the design process, often discovering mismatches only after synthesis, necessitating iterative optimization or redesign. Some works have shown the promising capabilities of large language models (LLMs) in hardware design generation tasks, but they fail to tackle the PPA trade-off problem. In this work, we propose an LLM-based reinforcement learning framework, PPA-RTL, aiming to introduce LLMs as a cutting-edge automation tool by directly incorporating post-synthesis metrics PPA into the hardware design generation phase. We design PPA metrics as reward feedback to guide the model in producing designs aligned with specific optimization objectives across various scenarios. The experimental results demonstrate that PPA-RTL models, optimized for Power, Performance, Area, or their various combinations, significantly improve in achieving the desired trade-offs, making PPA-RTL applicable to a variety of application scenarios and project constraints. 

Abstract Incorporation of ferric iron in mantle silicates stabilizes different crystal structures and changes phase transition conditions, thus impacting seismic wave speeds and discontinuities. In MgSiO3-Fe2O3 mixtures, recent experiments indicate the coexistence of fully oxidized iron-rich (Mg0.5Fe0.53+)(Fe0.53+Si0.5)O3 with Fe-poor silicate (wadsleyite or bridgmanite) and stishovite at 15 to 27 GPa and 1773 to 2000 K, conditions relevant to subducted lithosphere in the Earth’s transition zone and uppermost lower mantle. X-ray diffraction measurements show that (Mg0.5Fe0.53+)(Fe0.53+Si0.5)O3 recovered from these conditions adopts the R3c LiNbO3-type structure, which transforms to the bridgmanite structure again between 18.3 GPa and 24.7 GPa at 300 K. Diffraction observations are used to obtain the equation of state of the LiNbO3-type phase up to 18.3 GPa. These observations combined with multi-anvil experiments suggest that the stable phase of (Mg0.5Fe0.53+)(Fe0.53+Si0.5)O3 is bridgmanite at 15-27 GPa, which transforms on decompression to LiNbO3-type structure. Our calculation revealed that ordering of the ferric ion reduces the kinetic energy barrier of the transition between (Mg0.5Fe0.53+)(Fe0.53+Si0.5)O3 LiNbO3 structure and bridgmanite relative to the MgSiO3 akimotoite-bridgmanite system. Dense Fe3+-rich bridgmanite structure is thus stable at substantially shallower depths than MgSiO3 bridgmanite and would promote subduction. 

ABSTRACT While many plant lineages display remarkable diversity in morphological form, our understanding of how phenotypic diversity, or disparity, arises in relation to genomic evolution over geologic scales remains poorly understood. Here, we investigated the relationship between phenotypic and genomic evolution in the Fagales, a lineage of woody plants that has been a dominant component of temperate and subtropical forests since the Late Cretaceous. We examine newly generated transcriptomic and trait datasets representing most extant genera and a rich diversity of Cretaceous fossil representatives. Our phylogenomic analyses identify recurrent hotspots of gene duplication and genomic conflict across the order. Our phenotypic analyses showed that the morphospace occupied by Fagales was largely filled by the early Cenozoic, and rates of evolution were highest during the early radiation of the Fagales crown and its major families. These results suggest that Fagales conforms to an “early‐burst” model of disparification, with morphospace being filled early in the order's diversification history, and that elevated levels of phenotypic evolution also often correspond to hotspots of gene duplication. Species diversification appears decoupled from patterns of both phenotypic and genomic evolution, highlighting the multidimensional nature of the evolution of plant diversity across geological timescales. 

Careful placement of a distributed computational application within a target device cluster is critical for achieving low application completion time. The problem is challenging due to its NP-hardness and combinatorial nature. In recent years, learning-based approaches have been proposed to learn a placement policy that can be applied to unseen applications, motivated by the problem of placing a neural network across cloud servers. These approaches, however, generally assume the device cluster is fixed, which is not the case in mobile or edge computing settings, where heterogeneous devices move in and out of range for a particular application. To address the challenge of scaling to different-sized device clusters and adapting to the addition of new devices, we propose a new learning approach called GiPH, which learns policies that generalize to dynamic device clusters via 1) a novel graph representation gpNet that efficiently encodes the information needed for choosing a good placement, and 2) a scalable graph neural network (GNN) that learns a summary of the gpNet information. GiPH turns the placement problem into that of finding a sequence of placement improvements, learning a policy for selecting this sequence that scales to problems of arbitrary size. We evaluate GiPH with a wide range of task graphs and device clusters and show that our learned policy rapidly finds good placements for new problem instances. GiPH finds placements that achieve up to 30.5% better makespan, searching up to 3× faster than other search-based placement policies.