fla-org/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.

fla-org/flash-linear-attention

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

fla-org/flash-linear-attention

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

fla-org/flash-linear-attention

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

fla-org/flash-linear-attention

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

fla-org/flash-linear-attention

Port an FLA Triton kernel to Gluon when explicit layouts, asynchronous transfers, or scheduling can address a measured bottleneck.

fla-org/flash-linear-attention

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

fla-org/flash-linear-attention

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

fla-org/flash-linear-attention

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

fla-org/flash-linear-attention