flash-linear-attention

Profile and optimize FLA Triton-Ascend kernels using NPU traces, with guidance for UB capacity, memory movement, launch limits, and numerical correctness.

Select and run correctness coverage for FLA kernels and modules, including gradients, dispatch boundaries, and Triton addressing changes.

Define supported inputs, numerical budgets, routing, and validation before designing an FLA kernel or numerical change.

Change FLA backend registration, dispatch, or verifiers while preserving routing and call contracts.

Modify or review KDA gates, chunk and recurrent kernels, backend routes, and their correctness coverage.

Profile and optimize FLA kernels on NVIDIA GPUs, with same-hardware benchmarks and targeted Nsight Compute analysis.

Iterate on FLA kernel performance with a frozen correctness gate, a measured baseline, and reproducible candidate comparisons.

Prepare or update an FLA pull request with focused scope, verified evidence, and the current repository template.

Split oversized FLA pull requests into focused changes, reuse existing PRs, and create or maintain verified GitHub native PR stacks.

Design and optimize FLA Gluon kernels using explicit layouts, memory movement, and scheduling, including ports from Triton and improvements to existing Gluon implementations.