add commits and fix maskdisabled

example/32_batched_gemm_scale_softmax_gemm/batched_multihead_attention_backward_fp16.cpp

@@ -75,7 +75,7 @@ static constexpr auto MaskingSpec =
     ck::tensor_operation::device::MaskingSpecialization::MaskOutUpperTriangle;
 #else
 static constexpr auto MaskingSpec =
-    ck::tensor_operation::device::MaskingSpecialization::MaskOutUpperTriangle;
+    ck::tensor_operation::device::MaskingSpecialization::MaskDisabled;
 #endif
 
 static constexpr auto TensorSpecQ = ck::tensor_operation::device::TensorSpecialization::Default;

include/ck/tensor_operation/gpu/grid/gridwise_batched_multihead_attention_backward_xdl_cshuffle_v1.hpp

@@ -1267,7 +1267,8 @@ struct GridwiseBatchedMultiheadAttentionBackward_Xdl_CShuffle
                 k_grid_desc_n0_k_n1,
                 make_multi_index(0, o_block_data_idx_on_grid, 0),
                 b1_element_op,
-                Gemm1::b_block_desc_bk0_n_bk1,
+                Gemm1::b_block_desc_bk0_n_bk1, // there n actually is k, k is N, so name can be
+                                               // b_block_desc_bn0_k_bn1
                 make_multi_index(0, 0, 0),
                 tensor_operation::element_wise::PassThrough{});