Fix the API pointer types in qloop_v1 device-op

include/ck/tensor_operation/gpu/device/impl/device_batched_mha_bwd_xdl_cshuffle_qloop_v1.hpp

@@ -1270,10 +1270,10 @@ struct DeviceBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V1
         void* p_qgrad_grid,
         void* p_kgrad_grid,
         void* p_vgrad_grid,
-        const D0DataType* p_acc0_bias,
-        const D1DataType* p_acc1_bias,
-        D0DataType* p_d0grad_grid,
-        D1DataType* p_d1grad_grid,
+        const void* p_acc0_bias,
+        const void* p_acc1_bias,
+        void* p_d0grad_grid,
+        void* p_d1grad_grid,
         const std::vector<index_t>& a_gs_ms_ks_lengths,
         const std::vector<index_t>& a_gs_ms_ks_strides,
         const std::vector<index_t>& b_gs_ns_ks_lengths,
@@ -1312,8 +1312,8 @@ struct DeviceBatchedMultiheadAttentionBackward_Qloop_Xdl_CShuffle_V1
             static_cast<OutputDataType*>(p_vgrad_grid),
             static_cast<const D0DataType*>(p_acc0_bias), // cast in struct Argument
             static_cast<const D1DataType*>(p_acc1_bias), // cast in struct Argument
-            static_cast<const D0DataType*>(p_d0grad_grid),
-            static_cast<const D1DataType*>(p_d1grad_grid),
+            static_cast<D0DataType*>(p_d0grad_grid),
+            static_cast<D1DataType*>(p_d1grad_grid),
             a_gs_ms_ks_lengths,
             a_gs_ms_ks_strides,
             b_gs_ns_ks_lengths,