一句话：提出支持 1–8 位任意位宽低精度类型的 Tile 级 GPGPU 编程语言 Tilus，经自动向量化与指令选择生成高效 GPU 内核。
LLM 推理对内存带宽与算力需求巨大，低精度计算是提升效率的关键，但现有低精度内核生成方法仅支持 2 的幂位宽，且受高层 GPU 编程抽象限制，难以进行细粒度寄存器管理与优化访存模式。为此，作者提出领域特定语言 Tilus，面向 GPGPU 计算，支持 1 至 8 位任意位宽的低精度数据类型，同时保持可编程性；其特性包括线程块级编程模型、分层内存空间、新颖的代数化布局系统以及对多种低精度类型的广泛支持。Tilus 程序通过自动向量化与指令选择被编译为高效 GPU 程序。实验表明，Tilus 能有效覆盖完整位宽谱系的低精度类型，性能较 Triton、Ladder、QuantLLM 与 Marlin 分别提升 1.75x、2.61x、1.29x 与 1.03x，并已开源。
Serving Large Language Models (LLMs) is critical for AI-powered applications, yet it demands substantial computational resources, particularly in memory bandwidth and computational throughput. Low-precision computation has emerged as a key technique to improve efficiency while reducing resource consumption. Existing approaches for generating low-precision kernels are limited to weight bit widths that are powers of two and suffer from suboptimal performance because of high-level GPU programming abstractions. These abstractions restrict critical optimizations, such as fine-grained register management and optimized memory access patterns, that are essential for efficient low-precision computations. In this paper, we introduce Tilus, a domain-specific language designed for General-Purpose GPU (GPGPU) computing that supports low-precision data types with arbitrary bit widths from 1 to 8 while maintaining GPU programmability. Tilus features a thread-block-level programming model, a hierarchical memory space, a novel algebraic layout system, and extensive support for diverse low-precision data types. Tilus programs are compiled into highly efficient GPU programs through automatic vectorization and instruction selection. Extensive experiments demonstrate that Tilus efficiently supports a full spectrum of low-precision data types, and outperforms state-of-the-art low-precision kernels. Compared to existing compilers such as Triton and Ladder, as well as hand-optimized kernels such as QuantLLM and Marlin, Tilus achieves performance improvements of: 1.75x, 2.61x, 1.29x and 1.03x, respectively. We open-source Tilus at https://github.com/NVIDIA/tilus.
DOI 原文 · arXiv · PDF（开放获取） · DBLP
@inproceedings{DingHZL0Y0P26,
title = {Tilus: A Tile-Level GPGPU Programming Language for Low-Precision Computation},
author = {Yaoyao Ding and Bohan Hou and Xiao Zhang and Allan Lin and Tianqi Chen and Cody Hao Yu and Yida Wang and Gennady Pekhimenko},
booktitle = {ASPLOS},
year = {2025},
doi = {10.1145/3760250.3762219}
}