← 研踪 PaperTrailarXiv:2611.09225 · v1
Fused-Softmax for efficient quantized attention
首次提交 2026-11-06 17:43
版本更新 2026-11-06 17:43
雷达观测 2026-11-09 11:19
摘要
Low-precision matrix throughput on modern GPUs is growing far faster than vector-arithmetic throughput. Attention therefore cannot realize the available GEMM performance because softmax has become the pipeline bottleneck. We introduce Fused-Softmax Attention, a training-free algorithm that reformulates and approximates softmax to fold its dominant operations into the attention GEMMs, moving work from vector units to matrix cores. Implemented for the AMD CDNA 4 architecture, our MXFP8, MXFP6 and MXFP4 kernels reach approximately 3.0, 4.0 and 4.6 PFLOP/s on an AMD Instinct MI355X GPU, yielding speedups of $2.1\times$, $2.8\times$ and $3.2\times$ over the BF16 attention. Importantly, the additional error introduced by Fused-Softmax remains below the error floor set by quantizing the GEMM operands. On Wan2.2-T2V-A14B, a hybrid MXFP6/4 schedule delivers up to $2.05\times$ end-to-end speedup while preserving the visual quality of the outputs. The kernels ship in AMD's open-source AITER library, and they powered AMD's MLPerf Inference v6.1 Wan2.2 submission, which ran $1.7\times$ faster than AMD's previous-round result on the same hardware. Our results show that redesigning softmax to utilize the matrix pipeline enables attention to capitalize on continued growth in low-precision matrix throughput.
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-06 17:43 · 最新观测版本
Fused-Softmax for efficient quantized attention
Low-precision matrix throughput on modern GPUs is growing far faster than vector-arithmetic throughput. Attention therefore cannot realize the available GEMM performance because softmax has become the pipeline bottleneck. We introduce Fused-Softmax Attention, a training-free algorithm that reformulates and approximates softmax to fold its dominant operations into the attention GEMMs, moving work from vector units to matrix cores. Implemented for the AMD CDNA 4 architecture, our MXFP8, MXFP6 and MXFP4 kernels reach approximately 3.0, 4.0 and 4.6 PFLOP/s on an AMD Instinct MI355X GPU, yielding speedups of $2.1\times$, $2.8\times$ and $3.2\times$ over the BF16 attention. Importantly, the additional error introduced by Fused-Softmax remains below the error floor set by quantizing the GEMM operands. On Wan2.2-T2V-A14B, a hybrid MXFP6/4 schedule delivers up to $2.05\times$ end-to-end speedup while preserving the visual quality of the outputs. The kernels ship in AMD's open-source AITER library, and they powered AMD's MLPerf Inference v6.1 Wan2.2 submission, which ran $1.7\times$ faster than AMD's previous-round result on the same hardware. Our results show that redesigning softmax to utilize the matrix pipeline enables attention to capitalize on continued growth in low-precision matrix throughput.
查看此版本解读 →查看此版本原文 ↗