SemiAnalysis_· @SemiAnalysis_ · X·· 5 小时前AI 评分
AMD 抢先在上游 vLLM 落地 MegaKernels 解码支持，早于 NVIDIA
AMD 已在上游 vLLM 落地 MegaKernels 解码支持，早于 NVIDIA，这是首个仅支持 ROCm 而尚不支持 CUDA 的 vLLM 重大功能。MegaKernels 通过将 kernel 启动次数减少 10x 来服务高交互性解码场景；CUDA 侧的上游 vLLM megakernel 支持目前连 draft PR 都还没有。
ALERT🚨: AMD JUST LANDED SUPPORT FOR MEGAKERNEL DECODE IN UPSTREAM VLLM WELL BEFORE NVIDIA. This marks one of the first times that a major vLLM feature is only supported on ROCm and not on CUDA yet.
MegaKernels is a technique that is massively important for high-interactivity decoding by reducing the # of kernel launches by 10x.
This marks an❗️important milestone❗️for ROCm vLLM to land major ROCm-exclusive features. CUDA vLLM doesn't even have draft PRs for upstream vLLM megakernel support yet.
来源：SemiAnalysis_ · x.com