diff --git a/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp b/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp
index baae1c68a2..21ff8bd3eb 100644
--- a/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp
+++ b/include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp
@@ -141,6 +141,7 @@ struct BlockwiseGemmXdlops_pipeline_bpreshuffle_v1<BlockGemmPipelineScheduler::I
 
     using Base::AMmaKStride;
     using Base::BMmaKStride;
+    using Base::MWaves;
 
     static constexpr index_t PrefetchStages  = 2;
     static constexpr index_t PrefillStages   = 1;
@@ -182,9 +183,9 @@ struct BlockwiseGemmXdlops_pipeline_bpreshuffle_v1<BlockGemmPipelineScheduler::I
 
     __device__ static constexpr auto HotLoopScheduler()
     {
-        constexpr auto num_ds_read_inst_a     = HotLoopInstList::A_LDS_Read_Inst_Num * 2;
+        constexpr auto num_ds_read_inst_a     = HotLoopInstList::A_LDS_Read_Inst_Num;
         constexpr auto num_buffer_load_inst_a = HotLoopInstList::A_Buffer_Load_Inst_Num;
-        constexpr auto num_buffer_load_inst_b = HotLoopInstList::B_Buffer_Load_Inst_Num * 2;
+        constexpr auto num_buffer_load_inst_b = HotLoopInstList::B_Buffer_Load_Inst_Num * MWaves;
 
         // B global
         static_for<0, num_buffer_load_inst_b, 1>{}([&](auto i) {