Ameyn/gdn decode cutedsl kernel#2498

Open

ameynaik-hub wants to merge 9 commits intoflashinfer-ai:mainfrom

ameynaik-hub:ameyn/gdn-decode-cutedsl-kernel

Commits on Feb 4, 2026

Add Gated Delta Rule CuTe-DSL kernel for decode-phase inference
ameynaik-hub
committed

Commits on Feb 5, 2026

Fix ruff B905 linter error
ameynaik-hub
committed

Commits on Feb 7, 2026

Add bf16 h-state reference, fastmath/rsqrt/L1-bypass optimizations to CuTe-DSL kernel, and improved CuTe-DSL benchmarking/testing
ameynaik-hub
committed

Commits on Feb 13, 2026

Merge branch 'flashinfer-ai:main' into ameyn/gdn-decode-cutedsl-kernel
ameynaik-hub
authored

Commits on Feb 16, 2026

Refactor: Consolidate dtype parameters in reference implementation

ameynaik-hub
and
claude
committed