diff --git a/mindspore/lite/src/runtime/kernel/arm/base/gather_base.cc b/mindspore/lite/src/runtime/kernel/arm/base/gather_base.cc
index 2cc9b986b45..e11215c342c 100644
--- a/mindspore/lite/src/runtime/kernel/arm/base/gather_base.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/base/gather_base.cc
@@ -21,7 +21,7 @@ using mindspore::lite::RET_ERROR;
 using mindspore::lite::RET_OK;
 
 namespace mindspore::kernel {
-#ifdef SERVER_INFERENCE
+#ifdef DYNAMIC_THREAD_DISTRIBUTE
 constexpr int kMinCostPerThread = 16384;
 #endif
 int GatherRun(void *cdata, int task_id, float, float) {
@@ -137,7 +137,7 @@ int GatherBaseCPUKernel::ChooseThreadCuttingstrategy() {
     return RET_OK;
   }
   int64_t block_size = 1;
-#ifdef SERVER_INFERENCE
+#ifdef DYNAMIC_THREAD_DISTRIBUTE
   auto all_bytes = static_cast<int64_t>(out_tensors_.front()->Size());
   if (all_bytes <= static_cast<int64_t>(kMinCostPerThread)) {
     block_boundary_infos_.emplace_back(BlockBoundaryInfo{0, 0, outer_size_, 0});
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/bias_fp32.cc b/mindspore/lite/src/runtime/kernel/arm/fp32/bias_fp32.cc
index d4d7140436e..602634c08d9 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/bias_fp32.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/bias_fp32.cc
@@ -28,7 +28,7 @@ using mindspore::lite::RET_OK;
 using mindspore::schema::PrimitiveType_BiasAdd;
 
 namespace mindspore::kernel {
-#ifdef SERVER_INFERENCE
+#ifdef DYNAMIC_THREAD_DISTRIBUTE
 constexpr int kMinCostPerThread = 1 << 16;
 #endif
 int BiasAddRun(void *cdata, int task_id, float lhs_scale, float rhs_scale) {
@@ -81,7 +81,7 @@ int BiasCPUKernel::ReSize() {
 int BiasCPUKernel::ChooseThreadCuttingstrategy() {
   split_points_.clear();
   int64_t block_size = 1;
-#ifdef SERVER_INFERENCE
+#ifdef DYNAMIC_THREAD_DISTRIBUTE
   block_size = MSMAX(total_num_ / op_parameter_->thread_num_, kMinCostPerThread);
   thread_num_ = MSMIN(UP_DIV(total_num_, block_size), op_parameter_->thread_num_);
 #else
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc b/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc
index f4d3da39200..44c7edac75f 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc
@@ -74,7 +74,7 @@ int MatmulFp32BaseCPUKernel::PackMatrixA() {
         reinterpret_cast<float *>(ms_context_->allocator->Malloc(matrix_a_.pack_size * sizeof(float)));
     }
   } else {
-#ifdef SERVER_INFERENCE
+#ifdef SHARING_MODEL_WEIGHT
     auto a_packed = lite::PackWeightManager::GetInstance()->GetPackedTensor(
       in_tensors()[FIRST_INPUT], static_cast<size_t>(matrix_a_.pack_size) * sizeof(float));
     matrix_a_.pack_ptr = reinterpret_cast<float *>(a_packed.second);
@@ -125,7 +125,7 @@ int MatmulFp32BaseCPUKernel::PackMatrixB() {
         reinterpret_cast<float *>(ms_context_->allocator->Malloc(matrix_b_.pack_size * sizeof(float)));
     }
   } else {
-#ifdef SERVER_INFERENCE
+#ifdef SHARING_MODEL_WEIGHT
     auto b_packed = lite::PackWeightManager::GetInstance()->GetPackedTensor(
       in_tensors()[SECOND_INPUT], static_cast<size_t>(matrix_b_.pack_size) * sizeof(float));
     matrix_b_.pack_ptr = reinterpret_cast<float *>(b_packed.second);
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.h b/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.h
index bc77c79e107..15f50c11abf 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.h
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.h
@@ -19,7 +19,7 @@
 
 #include <vector>
 #include "src/inner_kernel.h"
-#ifdef SERVER_INFERENCE
+#ifdef SHARING_MODEL_WEIGHT
 #include "src/pack_weight_manager.h"
 #endif
 #include "nnacl/matmul_parameter.h"
diff --git a/mindspore/lite/src/runtime/kernel/arm/int8/div_int8.cc b/mindspore/lite/src/runtime/kernel/arm/int8/div_int8.cc
index 6e0c03a2078..3e14937fc62 100644
--- a/mindspore/lite/src/runtime/kernel/arm/int8/div_int8.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/int8/div_int8.cc
@@ -158,11 +158,28 @@ int DivInt8CPUKernel::Run() {
 
   if (broadcast_) {
     ArithmeticParameter tile_para;
-    tile_para.ndim_ = out_tensors_.at(0)->shape().size();
-    for (size_t i = 0; i < tile_para.ndim_; i++) {
-      tile_para.in_shape0_[i] = in_tensors_.at(0)->DimensionSize(i);
-      tile_para.in_shape1_[i] = in_tensors_.at(1)->DimensionSize(i);
-      tile_para.out_shape_[i] = out_tensors_.at(0)->DimensionSize(i);
+    auto out_shape = out_tensors_[FIRST_INPUT]->shape();
+    tile_para.ndim_ = out_shape.size();
+    auto in_shape0 = in_tensors_[FIRST_INPUT]->shape();
+    MS_CHECK_TRUE_MSG(out_shape.size() >= in_shape0.size(), RET_ERROR,
+                      "Sub first-input shape size is larger than out.");
+    for (size_t i = 0; i < out_shape.size() - in_shape0.size(); ++i) {
+      tile_para.in_shape0_[i] = 1;
+    }
+    for (size_t i = 0; i < in_shape0.size(); ++i) {
+      tile_para.in_shape0_[i + out_shape.size() - in_shape0.size()] = in_shape0[i];
+    }
+    auto in_shape1 = in_tensors_[SECOND_INPUT]->shape();
+    MS_CHECK_TRUE_MSG(out_shape.size() >= in_shape1.size(), RET_ERROR,
+                      "Sub second-input shape size is larger than out.");
+    for (size_t i = 0; i < out_shape.size() - in_shape1.size(); ++i) {
+      tile_para.in_shape1_[i] = 1;
+    }
+    for (size_t i = 0; i < in_shape1.size(); ++i) {
+      tile_para.in_shape1_[i + out_shape.size() - in_shape1.size()] = in_shape1[i];
+    }
+    for (size_t i = 0; i < out_shape.size(); ++i) {
+      tile_para.out_shape_[i] = out_shape[i];
     }
     tile0_data_ = static_cast<int8_t *>(ms_context_->allocator->Malloc(out_tensors_.at(0)->Size()));
     tile1_data_ = static_cast<int8_t *>(ms_context_->allocator->Malloc(out_tensors_.at(0)->Size()));