CUDA Tile用简单代码实现高效AI计算，但性能依赖硬件架构。
Evaluating CUDA Tile for AI Workloads on Hopper and Blackwell GPUs
- 用Python编写简洁的分块内核，自动利用Tensor Core和TMA加速。
- 在Blackwell B200上，注意力计算达1007 TFLOP/s，比FlashAttention-2快2.5倍。
- 适合快速开发原型，但跨架构兼容性弱，需针对不同芯片调优。
NVIDIA的CUDA Tile（CuTile）引入一种基于Python的、以分块为中心的GPU内核编程抽象，在保留张量核心和张量内存加速器（TMA）效率的同时简化开发。我们首次对CuTile在三款横跨Hopper与Blackwell架构的NVIDIA GPU（H100 NVL、B200、RTX PRO 6000 Blackwell Server Edition）上进行了独立跨架构评估，对比cuBLAS、Triton、WMMA及原生SIMT等主流方法。测试涵盖GEMM、融合多头注意力及端到端大模型推理（BF16/FP16精度）。结果表明，CuTile效果高度依赖工作负载与架构：在数据中心级Blackwell（B200）上，融合注意力最高达1007 TFLOP/s，较FlashAttention-2提升2.5倍，仅需60行Python代码；而对GEMM，其性能达到cuBLAS的52%-79%（22行代码，相比WMMA的123行），可替代手工编写内核，但尚未超越厂商优化库。然而，同一注意力内核在RTX PRO 6000（sm_120）上仅达FlashAttention-2的53%，暴露显著跨架构优化差距。相比之下，Triton在所有平台保持62%-101%的cuBLAS性能，无需架构特化调优，展现出更强可移植性。
原文摘要 · Abstract (English)
NVIDIA's CUDA Tile (CuTile) introduces a Python-based, tile-centric abstraction for GPU kernel development that aims to simplify programming while retaining Tensor Core and Tensor Memory Accelerator (TMA) efficiency on modern GPUs. We present the first independent, cross-architecture evaluation of CuTile against established approaches such as cuBLAS, Triton, WMMA, and raw SIMT on three NVIDIA GPUs spanning Hopper and Blackwell: H100 NVL, B200, and RTX PRO 6000 Blackwell Server Edition. We benchmark representative AI workloads, including GEMM, fused multi-head attention, and end-to-end LLM inference in BF16/FP16 precision, to assess both performance and portability. Our results show that CuTile effectiveness is strongly workload- and architecture-dependent. On datacenter-class Blackwell (B200), CuTile achieves up to 1007 TFLOP/s for fused attention, outperforming FlashAttention-2 by 2.5x while requiring only 60 lines of Python kernel code. For GEMM, CuTile reaches 52-79% of cuBLAS performance in 22 lines of code (versus 123 for WMMA), making it a practical replacement for hand-written CUDA kernels but not yet for vendor-optimized libraries. However, the same CuTile attention kernel achieves only 53% of FlashAttention-2 throughput on RTX PRO 6000 (sm_120), exposing significant cross-architecture optimization gaps. In contrast, Triton sustains 62-101% of cuBLAS performance across all tested platforms without architecture-specific tuning, demonstrating substantially stronger portability.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。