Understand this story
On October 22, 2026, the official PyTorch blog introduced that users of AMD MI350 series GPUs can now use FlyDSL through existing operator APIs without calling custom kernels. The support covers dense and grouped GEMM, MXFP8/MXFP4 scaled GEMM, RMSNorm, TopK, and attention kernels provided via the AOTriton 0.14b series. In terms of performance, on MI355X, BF16 dense GEMM achieves a 1.10x geometric mean speedup over the faster baseline among ATen and Triton, MXFP8 and MXFP4 achieve 1.58x and 1.68x over ATen respectively, and grouped GEMM achieves 1.20x over Triton. In vLLM end-to-end A/B tests, whole-request speedups are BF16 1.13x, MXFP8 2.04x, and MXFP4 3.01x, depending on the model and concurrency. There are currently no subsequent development facts.
AI-generated from public reports
Latest development · Report
PyTorch integrates FlyDSL backend on AMD GPUsDevelopments
1 confirmed developments
PyTorch supports FlyDSL on AMD MI350 GPUs through existing operator APIs
Source:PyTorch Blog
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 22
PyTorch BlogFirst-partySelected
PyTorch integrates FlyDSL backend on AMD GPUs
The official PyTorch blog introduces that users of AMD MI350 series GPUs can now use FlyDSL through existing operator APIs, covering dense and grouped GEMM, MXFP8/MXFP4 scaled GEMM, RMSNorm, TopK, and attention kernels provided via the AOTriton 0.14b series, without calling custom kernels. On MI355X, BF16 dense GEMM achieves a 1.10x geometric mean speedup over the faster baseline among ATen and Triton, MXFP8 and MXFP4 achieve 1.58x and 1.68x over ATen respectively, and grouped GEMM achieves 1.20x over Triton. In vLLM end-to-end A/B tests, whole-request speedups are BF16 1.13x, MXFP8 2.04x, and MXFP4 3.01x, depending on the model and concurrency.