removed global shuffle apis

include/ck/tensor_operation/gpu/device/impl/device_batched_multihead_attention_forward_xdl_cshuffle.hpp

@@ -39,7 +39,6 @@ template <typename GridwiseGemm,
           typename B1GridDesc_BK0_N_BK1,
           typename CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
           typename ZGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5,
-          typename ZShuffleGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_M4_N4_N5,
           typename LSEGridDescriptor_M,
           typename Block2CTileMap,
           typename ComputeBasePtrOfStridedBatch,
@@ -71,8 +70,6 @@ __global__ void
                 c_grid_desc_mblock_mperblock_nblock_nperblock,
             const ZGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5
                 z_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
-            const ZShuffleGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_M4_N4_N5
-                z_grid_shuffle_desc_m0_n0_m1_n1_m2_n2_m3_n3_m4_n4_n5,
             const LSEGridDescriptor_M lse_grid_desc_m,
             const Block2CTileMap block_2_ctile_map,
             const index_t batch_count,
@@ -130,7 +127,6 @@ __global__ void
                 b1_grid_desc_bk0_n_bk1,
                 c_grid_desc_mblock_mperblock_nblock_nperblock,
                 z_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
-                z_grid_shuffle_desc_m0_n0_m1_n1_m2_n2_m3_n3_m4_n4_n5,
                 lse_grid_desc_m,
                 block_2_ctile_map,
                 c0_matrix_mask,
@@ -163,7 +159,6 @@ __global__ void
             b1_grid_desc_bk0_n_bk1,
             c_grid_desc_mblock_mperblock_nblock_nperblock,
             z_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
-            z_grid_shuffle_desc_m0_n0_m1_n1_m2_n2_m3_n3_m4_n4_n5,
             lse_grid_desc_m,
             block_2_ctile_map,
             c0_matrix_mask,
@@ -653,10 +648,6 @@ struct DeviceBatchedMultiheadAttentionForward_Xdl_CShuffle
             z_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_ =
                 GridwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5(z_grid_desc_m_n_);
 
-            z_grid_shuffle_desc_m0_n0_m1_n1_m2_n2_m3_n3_m4_n4_n5_ =
-                GridwiseGemm::MakeCShuffleGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_M4_N4_N5(
-                    z_grid_desc_m_n_);
-
             if(p_lse_grid == nullptr)
             {
                 is_lse_storing_ = false;
@@ -706,9 +697,6 @@ struct DeviceBatchedMultiheadAttentionForward_Xdl_CShuffle
         typename GridwiseGemm::ZGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5
             z_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_;
 
-        typename GridwiseGemm::ZShuffleGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_M4_N4_N5
-            z_grid_shuffle_desc_m0_n0_m1_n1_m2_n2_m3_n3_m4_n4_n5_;
-
         // block-to-c-tile map
         typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
 
@@ -765,9 +753,8 @@ struct DeviceBatchedMultiheadAttentionForward_Xdl_CShuffle
 
             float ave_time = 0;
 
-            auto launch_kernel = [&](auto has_main_k_block_loop_,
-                                     auto is_dropout_,
-                                     auto is_lse_storing_) {
+            auto launch_kernel =
+                [&](auto has_main_k_block_loop_, auto is_dropout_, auto is_lse_storing_) {
                     const auto kernel = kernel_batched_multiheadattention_forward_xdl_cshuffle<
                         GridwiseGemm,
                         ADataType, // TODO: distiguish A/B datatype
@@ -785,7 +772,6 @@ struct DeviceBatchedMultiheadAttentionForward_Xdl_CShuffle
                         DeviceOp::B1GridDesc_BK0_N_BK1,
                         typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
                         typename GridwiseGemm::ZGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5,
-                    typename GridwiseGemm::ZShuffleGridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_M4_N4_N5,
                         DeviceOp::LSEGridDesc_M,
                         typename GridwiseGemm::DefaultBlock2CTileMap,
                         ComputeBasePtrOfStridedBatch,
@@ -817,7 +803,6 @@ struct DeviceBatchedMultiheadAttentionForward_Xdl_CShuffle
                         arg.b1_grid_desc_bk0_n_bk1_,
                         arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
                         arg.z_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_,
-                    arg.z_grid_shuffle_desc_m0_n0_m1_n1_m2_n2_m3_n3_m4_n4_n5_,
                         arg.lse_grid_desc_m_,
                         arg.block_2_ctile_map_,
                         arg.batch_count_,