cudarc adds 18.8ms/step overhead at b=256 (GPU kernels take 0.16ms). Phase 1: raw cuLaunchKernel wrapper. Phase 2: pre-extracted raw ptrs. Phase 3: fused mega-kernels (20 launches → 3). Target: 1ms/step → 500+ sps. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>