Bring the v1 card up to main

Bring the v1 card up to main

Bring the v1 sources and card up to main

Bring the v1 sources and card up to main

Record the sm90 comparison after the float32 row-tile change

Add the CuTe decode comparison and share the backward shared-memory sizing

Add the sm90 comparison, both dtypes

Check forward agreement at sequence lengths to 32768

Range-reduced rotation and contiguous backward operands

Update the card on the release branch to the current measurements

Sync v1 sources and card to main

Sync v1 sources and card to main

Mamba-3 MIMO chunked forward and recurrent decode, native CUDA