change the check for vector load

example/52_flash_atten_bias/grouped_mutihead_attention_bias_infer.cpp

@@ -120,7 +120,7 @@ using DeviceGemmInstance =
         8,
         8,
         true,
-        4,
+        1,
         S<16, 16, 1>, // B1BlockTransfer
         S<0, 2, 1>,
         S<0, 2, 1>,

include/ck/tensor_operation/gpu/device/impl/device_batched_mha_fwd_xdl_cshuffle_v2.hpp

@@ -1028,8 +1028,7 @@ struct DeviceBatchedMultiheadAttentionForward_Xdl_CShuffle_V2
         {
             if(arg.d0_n_length_stride_[1] == 1)
             {
-                if(!(arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector == 0 ||
-                     Transform::matrix_padder.PadN))
+                if(arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
                     return false;
             }
             else if(Acc0BiasTransferSrcScalarPerVector != 1)

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

@@ -744,8 +744,7 @@ struct DeviceBatchedMultiheadAttentionInfer_Xdl_CShuffle
         {
             if(arg.d0_n_length_stride_[1] == 1)
             {
-                if(!(arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector == 0 ||
-                     Transform::matrix_padder.PadN))
+                if(arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
                     return false;
             }
             else if(Acc0BiasTransferSrcScalarPerVector != 1)

include/ck/tensor_operation/gpu/device/impl/device_grouped_mha_fwd_xdl_cshuffle_v2.hpp

@@ -1104,9 +1104,7 @@ struct DeviceGroupedMultiheadAttentionForward_Xdl_CShuffle_V2
             {
                 if(device_arg.d0_n_length_stride_[1] == 1)
                 {
-                    if(!(device_arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector ==
-                             0 ||
-                         Transform::matrix_padder.PadN))
+                    if(device_arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
                         return false;
                 }
                 else if(Acc0BiasTransferSrcScalarPerVector != 1)

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

@@ -788,9 +788,7 @@ struct DeviceGroupedMultiheadAttentionInfer_Xdl_CShuffle
 
                 if(device_arg.d0_n_length_stride_[1] == 1)
                 {
-                    if(!(device_arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector ==
-                             0 ||
-                         Transform::matrix_padder.PadN))
+                    if(device_arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
                     {
                         return false;
                     }