Optimizing register transfer level (RTL) code is of vital importance in hardware design. Large language models (LLMs) provide new methods for the automatic generation and optimization of RTL code. However, existing methods for generating RTL code often focus on model fine-tuning and the use of various expansion techniques to enhance the RTL code generation capabilities, lacking attention to the functional correctness. To address this issue, we propose EstCoder, an LLM-powered collaborative agent framework for RTL code generation based on static functional score estimation. EstCoder operates a three-stage paradigm: Generation, Estimation and Correction. During the stages, the functional estimation agent statically evaluates the generated code based on score and assessment results, and decides whether to output the code directly, return it for regeneration, or forward it to the code correction agent. This famework can be applied to various LLMs that designed for RTL code generation, further enhancing the correctness of the generated code. By providing quantitative scores and human-readable requirements comparisons, it improves the transparency of AI-assisted RTL code generation. Experiments show that EstCoder significantly improves the correctness of RTL code generation by generic LLM by 3.2%-9.0%, demonstrating the practical value of our system.

The alignment between RNA sequences and structures in foundation models (FMs) has yet to be thoroughly investigated. Existing FMs have struggled to establish sequence-structure alignment, hindering the seamless flow of genomic information between RNA sequences and structures. In this study, we introduce OmniGenome, an RNA FM trained to align RNA sequences with respect to secondary structures through structure-contextualized modelling. This alignment enables free and bidirectional mappings between sequences and structures by utilizing a flexible RNA modelling paradigm that supports versatile input and output modalities, i.e., sequence and/or structure as input/output. We implement RNA design and zero-shot secondary structure prediction as case studies to evaluate the Seq2Str and Str2Seq mapping capabilities of OmniGenome. Results on the EternaV2 benchmark show that OmniGenome solved 74% of puzzles, whereas existing FMs solved only up to 3% of the puzzles due to the lack of sequence-structure alignment. We leverage four comprehensive in-silico genome modelling benchmarks to evaluate performance across a diverse set of downstream genome tasks, where the results show that OmniGenome achieves state-of-the-art performance on RNA and DNA benchmarks, even without any training on DNA genomes.

Spiking neural networks (SNNs) are promisingly considered as energy-efficient alternatives to traditional deep neural networks. At the same time, neuromorphic processors have garnered increasing development to support the efficient execution of large-scale SNNs. However, current works always separate their design to primarily prioritize a single criterion. Hardware-algorithm co-design allows for the simultaneous consideration of hardware and algorithm characteristics during the design process, effectively reducing resource usage while optimizing the algorithm's performance. In light of this, we developed a hardware-algorithm co-design framework named ASNPC for SNNs and neuromorphic processors. Considering the vast mixed-variable co-design space and the time-expensive function evaluations, we employed the surrogate-based multi-objective optimization algorithm MOTPE to identify Pareto solutions that balance algorithm performance and hardware costs. To rapidly obtain hardware results, we designed an end-to-end methodology that can automatically generate the Register-Transfer Level (RTL) code for neuromorphic processors corresponding to each candidate using templates from the hardware library. The evaluated hardware metrics, such as hardware resource and power consumption, are then fed back to MOTPE for the next candidate selection. Compared to existing works, the proposed approach can find better Pareto solutions within a limited search budget, making it effectively adapted to various application scenarios. Additionally, under the same hardware configuration, the neuromorphic processor we generated achieves lower hardware resource usage and higher throughput.

The automatic generation of Verilog code using Large Language Models (LLMs) presents a compelling solution to enhance the efficiency of hardware design flow. However, the state-of-the-art performance of LLMs in Verilog generation remains limited compared to programming languages such as Python. Previous research, Chain of Thought (CoT), has demonstrated that incorporating intermediate reasoning steps can significantly improve the performance of LLMs in code generation. In this paper, we propose the Verilog Tree of Thoughts (VToT) method. This structured prompting technique addresses the abstraction gap between Verilog and CoT by embedding hierarchical design constraints within the prompt. Experimental results on the VerilogEval and RTLLM benchmarks demonstrate that VToT prompting enhances both the syntactic and functional correctness of the generated code. Specifically, according to the RTLLM benchmark, VToT achieved a correctness rate of 75.9% at pass@5, representing an improvement of 10.4%. Furthermore, in the VerilogEval benchmark, VToT achieved state-of-the-art performance with a correctness rate of 52.4% at pass@1 (an increase of 8.9%) and 65.4% at pass@5 (an increase of 9.6%).

The rapid advancement of large language models (LLMs) has enabled automated Register Transfer Level (RTL) code generation, accelerating chip design workflows. However, existing benchmarks focus mainly on syntax and functionality, overlooking critical engineering aspects such as lint compliance, readability, and coding style. To address this gap, we propose RTLBench, a benchmark suite of 160 copyright-free RTL cases sourced from textbooks and open-source projects. RTLBench features a multi-dimensional evaluation framework covering syntax, functionality, lint compliance, readability, and style consistency. To assess subjective code quality metrics, it also incorporates an LLM-as-a-judge mechanism. We evaluated 24 state-of-the-art LLMs using RTLBench, finding that while several models perform well in syntax and functionality, most fall short on engineering quality. To address this, we propose Log2BetterRTL, a log-driven feedback system that transforms EDA tool diagnostics into iterative improvement prompts. It improves syntax correctness by up to 18.13 %, boosts functional correctness by 14.38 %, reduces lint violations by up to 229, and raises clarity scores by 0.51. These results demonstrate RTLBench's effectiveness in evaluating and enhancing LLMgenerated RTL, bridging the gap between generative AI and industrial-grade hardware design. The suite and scripts are available at: https://fangzhigang32.github.io/RTLBench.

Spiking Neural Networks (SNNs) are increasingly becoming a research hotspot in next-generation intelligent computing architectures due to their potential in energy efficiency, which also imposes higher demands on their structural design. However, manually designing SNN architectures becomes increasingly complex, making it difficult to balance design efficiency and optimal performance. Neural Architecture Search (NAS) provides a new approach for automatically constructing high-performance SNNs. Yet, existing SNN-NAS methods primarily focus on accuracy as the main objective, neglecting the optimization of critical hardware efficiency metrics such as latency, which limits their practical deployment value. This paper proposes a hardware-aware NAS-based method for SNNs, called SHL-NAS, which combines Constrained Bayesian Optimization (CBO) with a deployed SNN Hardware Latency (SHL) model to identify the optimal architecture under user-specified latency constraints. The SHL model can directly evaluate latency based on hardware parameters and architectural parameters, thereby reducing reliance on resource-intensive training and hardware-specific measurements, significantly improving efficiency. Experiments demonstrate that on the CIFAR-10 and CIFAR-100 datasets, SHL-NAS reduces the search iterations by 80% while discovering architectures that satisfy the latency constraints. Specifically, under 30ms and 40ms latency constraints on CIFAR-100, it achieves accuracy improvements of 0.2% and 1.1%, respectively, compared to state-of-the-art methods.

The increasing complexity of processor design demands higher requirements for simulation-based verification, particularly in test case generation. Existing random test generators often struggle to thoroughly validate the deep processor states or fail to provide sufficient coverage diversity. In this paper, we propose PerturbGen, an evolutionary algorithm-inspired population-based perturbation test generation method, designed to enhance traditional random test generators. Per-turbGen introduces crossover and mutation operators from evolutionary algorithm, applying perturbations at both the population and member levels to generate high-quality tests. Our approach also includes a coverage-guided feedback loop for iteratively filtering members to guide the exploration of uncovered areas in the processor. We evaluated PerturbGen on an open-source RISC-V processor and compared it with two widely-used random test generators. Our method achieves relative improvements of 3.56%, 5.70%, and 14.01% in three key coverage metrics, respectively, proving the superiority of PerturbGen. Our code is open-sourced at the anonymous link: https://anonymous.4open.science/r/PerturbGen-2B07.

With the increasing complexity of the hardware design, conducting verification before the tapeout is of utmost importance. Simulation-based verification remains the primary method owing to its scalability and flexibility. A comprehensive verification of modern processors usually requires numerous effective tests to cover all possible conditions and use cases, leading to significant time, resource, and manual effort even with the EDA. Moreover, novel domain specific architecture (DSA), such as neuromorphic processors, will exacerbate the challenge of verification. Fortunately, emerging large language models (LLMs) have been demonstrating a powerful ability to complete specific tasks assigned by human instructions. In this paper, we explore the challenges and opportunities encountered when using the LLMs to accelerate the DSA verification using the proposed LLM-based workflow consisting of test generation, compilation&simulation, and result collection&processing. By verifying a RISC-V core and a neuromorphic processor, we examine the capabilities and limitations of the LLMs when using them for the function verification of traditional processors and emerging DSA. In the experiment, 36$C$programs and 128 assembly snippets for the RISC-V core and the neuromorphic processor are generated using an advanced LLM to demonstrate our claim. The experimental results show that the code coverage based on the LLM test generation can reach 89% and 91% for the above two architectures respectively, showing a promising research direction for the future processor verification in the new golden age for computer architecture.

Design verification (DV) has existed for decades and is crucial for identifying potential bugs before chip tape- out. Hand-crafting test cases is time-consuming and error-prone, even for experienced verification engineers. Prior work has attempted to lighten this burden by rule-guided random test case generation. However, this approach does not eliminate the manual effort required to write rules that describe detailed hardware behavior. Motivated by advances in large language models (LLMs), we explore their potential to capture register transfer level (RTL) behavior and construct prompts for test case generation based on RTL behavior. First, we introduce a prompt framework, LLM - Driven Test Generation (LLM - TG), to generate test cases, thereby enhancing LLMs' test generation capabilities. Additionally, we provide an open-source prompt library that offers a set of standardized prompts for processor verification, aiming to improve test generation efficiency. Lastly, we use an LLM to verify a 12-stage, multi-issue, out-of-order RV64GC processor, achieving at least an 8.34 % increase in block coverage and at least a 5.8 % increase in expression coverage compared to the state-of-the-art (SOTA) methods, LLM4DV and RISCV- DV. The prompt library is available at https://github.com/LLM-TGIPrompt_Library.

Task migration maximizes performance while maintaining thermal safety in many-cores systems. Existing techniques exploit offline learning which requires tremendous training data and fixed-cycle migration which causes threads to miss the optimal migration timing. This paper presents Online Learning based State-Aware Task Migration (OLSATM). It pretrains a neural network (NN) with a small set of data and updates the model online to substitute the laborious data collection and model training of offline learning. It is state-aware and detects the timing when migration is needed, overcoming the shortcomings of periodical migration. Experimental results show that OLSATM enhances the performance by 3.7% and reduces the number of migration judgments by 26 % on average compared to the state-of-the-art task migration.

Recent advances in hardware/algorithm co-design for spiking neural networks have demonstrated its potential for jointly optimizing algorithmic performance while minimizing hardware overhead. However, the gigantic mixed-variable hard-ware/algorithm co-design space and time-consuming hardware verification still pose an intractable challenge for solutions exploration. To tackle these problems, 1) we propose a generic three-phase hardware/algorithm co-design framework. In this framework, 2) we target a reconfigurable neuromorphic processor, and parameterize the hardware and network architecture in a unified design space. 3) We propose a generic analytical model to estimate the parameter size and power consumption, which can support fast candidate evaluation during the exploration. 4) We extend vanilla TPE (a single-objective optimization algorithm) to MOTPE/D, a generic Multi-objective optimization (MOO) algorithm, by introducing a decomposition strategy.

Obstacle avoidance is a crucial task in unmanned aerial vehicles (UAV) motion planning. The accuracy and consistency of real-time visual information affect the gener-ation of obstacle avoidance commands, raising higher safety demands for obstacle avoidance. The neuromorphic computing-based obstacle avoidance solution can address these challenges. Dynamic vision sensors (DVS) exhibit low latency, low power consumption, and high dynamic range as novel neuromorphic sensors. Spiking neural networks (SNN) also leverage the same mechanism to efficiently process asynchronous and sparse event data generated by DVS, offering latency and energy efficiency advantages. Additionally, the optimal estimation method effectively mitigates the impact of noise and interference within the system, reducing the influence of errors on the algorithm and enhancing safety. Based on these considerations, this paper proposes a fast and safe obstacle avoidance framework. DVS is used to acquire event data from the environment, and a hardware-compatible lightweight SNN is employed to extract dynamic obstacle position information from the data. Compared to baseline methods, this approach reduces latency by 85%. Furthermore, two estimation methods are used to predict the movement of obstacles, ensuring flight safety by generating different UAV obstacle avoidance actions based on confidence intervals, even in the presence of obstacle information errors and omissions.

Liquid State Machine (LSM) is a brain-inspired computational model that has proven highly effective in various applications, owing to its intrinsic capability to process spatiotemporal information and its minimal training complexity. However, the performance of LSMs significantly depends on the design of their network architecture, which is overly reliant on existing human experience. Furthermore, as the network scale increases, the computing resources required for deployment and operation also increase, so we regarded the network design as a multi-objective problem. To address these challenges, we introduced an effective surrogate-assisted multi-objective evolutionary neural architecture search algorithm that balanced the accuracy and network scale. Our approach utilized parameter sensitivity analysis followed by the upper confidence bound algorithm to reduce the search space. Experimental results demonstrate that we successfully reduced the dimensions of the search space by 11% and the size of the entire search space by 75%. Compared to the state-of-the-art, our approach offered better trade-off solutions, such as a solution that reduced network scale by 32.5% while maintaining the same accuracy, and another that improved accuracy by 1.4% without changing the network scale. Furthermore, the knee point reduced network scale by 25 % and simultaneously increased accuracy by 0.7%. The source code can be accessed at https://github.com/XinSida/MOENAS-PSA.

Many real-world problems are computationally costly and the objective functions evolve over time. Data-driven, a.k.a. surrogate-assisted, evolutionary optimization has been recognized as an effective approach to tackle expensive black-box optimization problems in a static environment whereas it has rarely been studied under dynamic environments. This paper proposes a simple yet effective transfer learning framework to empower data-driven evolutionary optimization to solve expensive dynamic optimization problems. Specifically, a hierarchical multi-output Gaussian process is proposed to capture the correlation among data collected from different time steps with a linearly increased number of hyperparameters. Furthermore, an adaptive source task selection along with a bespoke warm staring initialization mechanisms are proposed to better leverage the knowledge extracted from previous optimization processes. By doing so, the data-driven evolutionary optimization can jump start the optimization in the new environment with a very limited computational budget. Experiments on synthetic benchmark test problems and a real-world case study demonstrate the effectiveness of our proposed algorithm in comparison with nine state-of-the-art peer algorithms.

Real-world optimization problems are often computationally expensive and feature multi-modal objective functions. Surrogate-assisted evolutionary optimization has proven to be an effective approach for addressing expensive black-box optimization challenges, but the technique has not been adequately studied in multi-modal situations. In this paper, we propose a simple but effective multi-output surrogate-based approach for empowering surrogate-assisted evolutionary optimization to address expensive multi-modal optimization problems. Specifically, our proposed approach employs a multi-output Gaussian process to capture correlations between data collected from different local areas. Experiments on synthetic benchmark test problems demonstrate the effectiveness of our proposed algorithm against five state-of-the-art peer algorithms.

The replacement policy contributes to enhancing the cache hit ratio, affecting the performance of the processor indirectly. Prior proposed static replacement policies are limited to certain classes of workload types, failing to achieve high hit rates in various benchmarks. In this work, we propose a self-adaptive replacement algorithm. The algorithm detects the drop points of the hit ratio online based on Bayesian inference and selects a new policy from a policy pool containing multiple policies according to its weight at the change points. Besides, the algorithm updates the weights based on the performance of the selected policy. Compared to the static replacement policy, our algorithm is able to apply to more access patterns and achieve a high hit ratio. We choose 15 benchmarks from DPC3 and concatenate them to generate a total of 13 composite benchmarks. We run our algorithm using a 2MB last-level cache (LLC) and show that our algorithm improves the hit rate by 2.6% over LRU and 10.6% over Random.

Multiobjective optimization problems are ubiquitous in real-world science, engineering, and design optimization problems. It is not uncommon that the objective functions are as a black box, the evaluation of which usually involve time-consuming and/or costly physical experiments. Data-driven evolutionary optimization can be used to search for a set of nondominated tradeoff solutions, where the expensive objective functions are approximated as a surrogate model. In this article, we propose a framework for implementing batched data-driven evolutionary multiobjective optimization (EMO). It is so general that any off-the-shelf EMO algorithms can be applied in a plug-in manner. There are two unique components: 1) based on the Karush–Kuhn–Tucker conditions, a manifold interpolation approach that explores more diversified solutions with a convergence guarantee along the manifold of the approximated Pareto-optimal set and 2) a batch recommendation approach that reduces the computational time of the data-driven evolutionary optimization process by evaluating multiple samples at a time in parallel. Comparing against seven state-of-the-art surrogate-assisted evolutionary algorithms, experiments on 168 benchmark test problem instances with various properties and a real-world application on hyper-parameter optimization fully demonstrate the effectiveness and superiority of our proposed framework, which is featured with a faster convergence and a stronger resilience to various Pareto-optimal front shapes.

The ultimate goal of multi-objective optimisation is to help decision makers (DMs) identify solution(s) of interest. However, providing the DMs with a large amount of the trade-off alternatives not only increase their workload, but also add irrelevant noise to the decision-making process. Without any prior knowledge, knee points, characterised as their smallest trade-off loss at all objectives, are attractive to decision makers in multi-criterion decision-making. In this paper, we propose a simple but effective knee point identification method based on Voronoi diagram. It divides the objective space into several Voronoi cells to capture the geometric characteristics of the underlying trade-off solution set. Thereafter, the knee points are identified as those having a local Voronoi distance. Empirical results demonstrate that our proposed method is able to identify knee points located in both convex and concave part of the corresponding Pareto-optimal front.

Expensive black-box optimization in dynamic environments is a challenging but important task since many real-world problems are changing over time and are computationally costly. Bayesian optimization has been widely recognized as an effective approach for tackling expensive black-box optimization in a static environment whereas it has rarely been studied for in dynamic environments. This paper proposes a simple but effective method to empower Bayesian optimization to solve dynamic optimization problems. It augments the covariance function with the measurement of the relationship between historical observations and the current ones. By doing so, the Bayesian optimization is able to leverage the observations from the previous time step to jump start the optimization in the new environment with a strictly limited computational budget. Experiments on synthetic benchmark test problems and a real-world case study demonstrate the effectiveness of our proposed algorithm.

When solving constrained multiobjective optimization problems, an important issue is how to balance convergence, diversity, and feasibility simultaneously. To address this issue, this paper proposes a parameter-free constraint handling technique, a two-archive evolutionary algorithm, for constrained multiobjective optimization. It maintains two collaborative archives simultaneously: one, denoted as the convergence-oriented archive (CA), is the driving force to push the population toward the Pareto front; the other one, denoted as the diversity-oriented archive (DA), mainly tends to maintain the population diversity. In particular, to complement the behavior of the CA and provide as much diversified information as possible, the DA aims at exploring areas under-exploited by the CA including the infeasible regions. To leverage the complementary effects of both archives, we develop a restricted mating selection mechanism that adaptively chooses appropriate mating parents from them according to their evolution status. Comprehensive experiments on a series of benchmark problems and a real-world case study fully demonstrate the competitiveness of our proposed algorithm, in comparison to five state-of-the-art constrained evolutionary multiobjective optimizers.

Decomposition has become an increasingly popular technique for evolutionary multiobjective optimization (EMO). A decomposition-based EMO algorithm is usually designed to approximate a whole Pareto-optimal front (PF). However, in practice, a decision maker (DM) might only be concerned in her/his region of interest (ROI), i.e., a part of the PF. Solutions outside that might be useless or even noisy to the decision-making procedure. Furthermore, there is no guarantee that the preferred solutions will be found when many-objective problems. This paper develops an interactive framework for the decomposition-based EMO algorithm to lead a DM to the preferred solutions of her/his choice. It consists of three modules, i.e., consultation, preference elicitation, and optimization. Specifically, after every several generations, the DM is asked to score a few candidate solutions in a consultation session. Thereafter, an approximated value function, which models the DM's preference information, is progressively learned from the DM's behavior. In the preference elicitation session, the preference information learned in the consultation module is translated into the form that can be used in a decomposition-based EMO algorithm, i.e., a set of reference points that are biased toward the ROI. The optimization module, which can be any decomposition-based EMO algorithm in principle, utilizes the biased reference points to guide its search process. Extensive experiments on benchmark problems with three to ten objectives fully demonstrate the effectiveness of our proposed method for finding the DM's preferred solutions.

Rather than a whole Pareto-optimal front, which demands too many points (especially in a high-dimensional space), the decision maker (DM) may only be interested in a partial region, called the region of interest (ROI). In this case, solutions outside this region can be noisy to the decision-making procedure. Even worse, there is no guarantee that we can find the preferred solutions when tackling problems with complicated properties or many objectives. In this paper, we develop a systematic way to incorporate the DM's preference information into the decomposition-based evolutionary multiobjective optimization methods. Generally speaking, our basic idea is a nonuniform mapping scheme by which the originally evenly distributed reference points on a canonical simplex can be mapped to new positions close to the aspiration-level vector supplied by the DM. By this means, we are able to steer the search process toward the ROI either directly or interactively and also handle many objectives. Meanwhile, solutions lying on the boundary can be approximated as well given the DM's requirements. Furthermore, the extent of the ROI is intuitively understandable and controllable in a closed form. Extensive experiments on a variety of benchmark problems with 2 to 10 objectives, fully demonstrate the effectiveness of our proposed method for approximating the preferred solutions in the ROI.

Existing studies on dynamic multiobjective optimization (DMO) focus on problems with time-dependent objective functions, while the ones with a changing number of objectives have rarely been considered in the literature. Instead of changing the shape or position of the Pareto-optimal front/set (PF/PS) when having time-dependent objective functions, increasing or decreasing the number of objectives usually leads to the expansion or contraction of the dimension of the PF/PS manifold. Unfortunately, most existing dynamic handling techniques can hardly be adapted to this type of dynamics. In this paper, we report our attempt toward tackling the DMO problems with a changing number of objectives. We implement a dynamic two-archive evolutionary algorithm which maintains two co-evolving populations simultaneously. In particular, these two populations are complementary to each other: one concerns more about the convergence while the other concerns more about the diversity. The compositions of these two populations are adaptively reconstructed once the environment changes. In addition, these two populations interact with each other via a mating selection mechanism. Comprehensive experiments are conducted on various benchmark problems with a time-dependent number of objectives. Empirical results fully demonstrate the effectiveness of our proposed algorithm.