| 1 | """Standalone Proton intra-kernel profiling for the fused-triton kernel. |
| 2 | |
| 3 | Runs ONLY the FMMS Triton kernel (stage 1) under Proton instrumentation, |
| 4 | skipping _local_reduce (stage 2) which uses torch.compile and would conflict |
nothing calls this directly
no test coverage detected