frost(sdpa): THD execute with zero host reads on SM100 — device-built metadata, plan-time envelope grid, CUDA-graph capturable (issue #552) - #606

frost(sdpa): THD execute with zero host reads on SM100 — device-built metadata, plan-time envelope grid, CUDA-graph capturable (issue #552)#606

vedaanta merged 7 commits into

NVIDIA:developNVIDIA/cudnn-frontend:developfrom

vedaanta:vagarwalla/frost-thd-552-envelopevedaanta/cudnn-frontend:vagarwalla/frost-thd-552-envelopeCopy head branch name to clipboard