onnxruntime

c36c4225 - [WebGPU EP] Fuse QMoE 1-token decode path to reduce GPU dispatches (#27998)

[WebGPU EP] Fuse QMoE 1-token decode path to reduce GPU dispatches (#27998) ### Description: ### Summary Fuse the QMoE 1-token decode path to reduce GPU dispatches from 17 (1 + k×4) to 5 (gate + fc1 + swiglu + fc2 + mix), improving token generation throughput by ~21% on Meteor Lake for the gpt-oss-20b MoE model (19 → 23 tps). ### Motivation The QMoE operator processes Mixture-of-Experts layers by selecting top-k experts (k=4) per token. In the original 1-token decode path, each expert is processed serially with 4 dispatches (gather + fc1 + swiglu + fc2 + mix), totaling 17 GPU dispatches per QMoE call. Since each dispatch has M=1, the GPU is underutilized and CPU dispatch overhead dominates. ### Approach For the 1-token path (num_rows == 1): **Gate1Token** — Select top-k experts and output an [indirect_experts](vscode-file://vscode-app/c:/Users/jiajiaqin/AppData/Local/Programs/Microsoft%20VS%20Code/ce099c1ed2/resources/app/out/vs/code/electron-browser/workbench/workbench.html) buffer mapping row index → expert index **Batched fc1 MatMulNBits** — Run a single M=k matmul with [per_row_weight_indirect](vscode-file://vscode-app/c:/Users/jiajiaqin/AppData/Local/Programs/Microsoft%20VS%20Code/ce099c1ed2/resources/app/out/vs/code/electron-browser/workbench/workbench.html) mode, where each row selects a different expert's weights via the indirect buffer **SwiGLU** — Apply activation on all k rows at once **Batched fc2 MatMulNBits** — Same per-row expert selection for the down projection **FusedFinalMix** — Accumulate all k weighted expert results into the output ### Follow-ups Fuse Batched fc1 MatMulNBits + SwiGLU Fuse Batched fc2 MatMulNBits + FusedFinalMix Finally, we only need three shaders: Gate1Token, fused Batched fc1 MatMulNBits, fused batched fc2 MatMulNBits.