Institution profile

Advanced Micro Devices, Inc.

Industry researchnorthamerica · us
Official website
Research library254linked papers
Opportunities0open roles
Selected work

Representative Papers

KernelCraft: Benchmarking for Agentic Close-to-Metal Kernel Generation on Emerging Hardware

Feb 10, 2026

This work addresses the challenge of prolonged and error-prone manual kernel development for emerging AI accelerators, which stems from their use of specialized instruction set architectures (ISAs) and hinders cross-platform portability. To overcome this, the paper introduces the first agent-driven benchmark for kernel generation tailored to novel hardware, featuring a large language model (LLM)-based feedback optimization framework. This framework leverages function calling and iterative refinement to automatically synthesize efficient and correct low-level kernels. Evaluation across more than twenty machine learning tasks on three distinct emerging accelerators demonstrates that the approach rapidly generates high-performance kernel code—often matching or surpassing compiler-generated baselines—even for previously unseen ISAs, thereby significantly accelerating the hardware development cycle.

3 citationsRead paper

CD4LM: Consistency Distillation and aDaptive Decoding for Diffusion Language Models

Jan 05, 2026arXiv.org

This work addresses the inefficiency of parallel decoding in diffusion language models, which stems from the mismatch between static scheduling during training and dynamic generation during inference. To overcome this limitation, the authors propose Discrete Space Consistency Distillation (DSCD), which renders the model invariant to denoising trajectories, and introduce a Confidence-Adaptive Decoding (CAD) strategy that dynamically allocates computational resources. This approach breaks free from the fixed scheduling constraints of conventional diffusion models, achieving significantly accelerated parallel inference while maintaining or even improving generation quality. Empirical results demonstrate that the method matches baseline accuracy on GSM8K with a 5.18× speedup, and achieves an average 3.62× acceleration across code and mathematical reasoning tasks with higher accuracy, consistently outperforming existing approaches.

2 citationsRead paper

MonoGS++: Fast and Accurate Monocular RGB Gaussian SLAM

Apr 03, 2025

To address the lack of depth priors, poor reconstruction in textureless regions, and Gaussian redundancy in monocular RGB SLAM, this paper proposes a real-time, high-accuracy SLAM framework that requires no depth sensor. Methodologically, it tightly integrates online visual odometry (VO) with 3D Gaussian Splatting, incorporating a dynamic Gaussian insertion mechanism, a sharpness-driven adaptive densification module, and plane-based geometric regularization—collectively enhancing dense reconstruction quality and geometric consistency under sparse feature constraints. Evaluated on Replica and TUM-RGBD datasets, the method achieves state-of-the-art tracking accuracy and attains millisecond-level real-time performance, with inference speed 5.57× faster than the baseline. Key contributions include: (i) the first dynamic Gaussian management strategy tailored for monocular SLAM, and (ii) a multi-level geometric regularization paradigm, effectively mitigating drift in textureless scenes and Gaussian over-saturation.

2 citationsRead paper

DiffBench Meets DiffAgent: End-to-End LLM-Driven Diffusion Acceleration Code Generation

Jan 06, 2026arXiv.org

Diffusion models suffer from high computational costs during inference and deployment challenges, compounded by the absence of a unified framework to automatically integrate diverse acceleration techniques. This work proposes DiffBench, a benchmark suite, and DiffAgent, an intelligent agent that uniquely combines large language models (LLMs) with genetic algorithms to establish a closed-loop automated workflow. This framework enables the generation, code synthesis, and iterative optimization of acceleration strategies tailored to any diffusion model. Leveraging a three-stage evaluation pipeline and an integrated code debugging mechanism, DiffAgent substantially outperforms existing LLM-based approaches, achieving consistently high performance across various hardware architectures and deployment scenarios.

1 citationsRead paper

Integrating Quantum Software Tools with(in) MLIR

Jan 05, 2026SCA/HPC Asia

This work addresses the lack of interoperability among current quantum compilation tools, which hinders the development of a unified software stack. It presents the first systematic integration of PennyLane and the Munich Quantum Toolkit (MQT) through the MLIR framework, leveraging MLIR’s intermediate representation and LLVM infrastructure to establish reusable integration pathways and best practices. By lowering the barrier to entry for quantum software engineers adopting MLIR, this approach enables seamless interoperability between distinct quantum toolchains. The study provides both a practical roadmap and actionable strategies for constructing a modular, interoperable quantum software ecosystem, thereby advancing the standardization and composability of quantum programming tools.

1 citationsRead paper
Recent publications

Latest Papers