Scholarlink

<p>In modern semiconductor manufacuring, computational lithogrpaphy plays more and more important role and has become a key enabler in keeping up with growing demands of the field. However, as wafer sizes shrink, and feature density increses, traditional forms of computational lithography has started facing limitations. Recent advancements in General-Purpose GPU (GPGPU) computing have transformed this landscape, enabling the transition from traditional CPU-based heuristics to massive, physics-based optimization. This paper provides a comprehensive survey of the GPU-accelerated computational lithography ecosystem. We first establish a multi-dimensional taxonomy covering application domains such as Inverse Lithography Technology (ILT) and EUV modeling alongside their underlying numerical algorithms. We then detail the GPU-resident computational workflow, illustrating how iterative optimization loops minimize host-device bottlenecks. Finally, we analyze specific GPU acceleration strategies, from SIMT parallelism and memory-aware tiling to emerging AI-physics hybrids. By synthesizing these elements, this review offers a roadmap for researchers to navigate the current state-of-the-art and identifies future trajectories in GPU-driven lithographic optimization.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GPU acceleration in computational lithography: Taxonomy, survey, and roadmap

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Shadi Alawneh,
  • Mayuresh Karandikar

摘要

In modern semiconductor manufacuring, computational lithogrpaphy plays more and more important role and has become a key enabler in keeping up with growing demands of the field. However, as wafer sizes shrink, and feature density increses, traditional forms of computational lithography has started facing limitations. Recent advancements in General-Purpose GPU (GPGPU) computing have transformed this landscape, enabling the transition from traditional CPU-based heuristics to massive, physics-based optimization. This paper provides a comprehensive survey of the GPU-accelerated computational lithography ecosystem. We first establish a multi-dimensional taxonomy covering application domains such as Inverse Lithography Technology (ILT) and EUV modeling alongside their underlying numerical algorithms. We then detail the GPU-resident computational workflow, illustrating how iterative optimization loops minimize host-device bottlenecks. Finally, we analyze specific GPU acceleration strategies, from SIMT parallelism and memory-aware tiling to emerging AI-physics hybrids. By synthesizing these elements, this review offers a roadmap for researchers to navigate the current state-of-the-art and identifies future trajectories in GPU-driven lithographic optimization.