Mojo package
mha_depth512
Modulesβ
- β
barriers: Barrier infrastructure for depth=256/512 pair-CTA SM100 attention kernels. - β
config: Configuration for pair-CTA SM100 (Blackwell) MHA kernels (depth 256/512). - β
correction_warp: Correction warp group logic for depth=256/512 pair-CTA SM100 attention. - β
dispatch: Dispatch for depth=256/512 pair-CTA SM100 (Blackwell) MHA prefill. - β
kernel: Kernel entry point for depth=256/512 pair-CTA SM100 (Blackwell) MHA prefill. - β
load_warp: TMA load warp logic for depth=256/512 pair-CTA SM100 attention. - β
mma_warp: MMA warp logic for depth=256/512 pair-CTA SM100 attention. - β
smem: Shared memory layout for depth=512 pair-CTA SM100 attention kernels. - β
softmax_warp: Softmax warp group logic for depth=256/512 pair-CTA SM100 attention.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!