Is there any plan to support distributed shared memory & 2SM thread block cluster in Example 77 Blackwell FMHA?

[bdh0404]

The current code implements only 1 SM case, without distributed shared memory capability. Using distributed smem with thread block cluster tile in Seq-Q direction can reduce the L2 Read traffic on K and V matrices.