[clang-tidy] replenish cppcoreguidelines-narrowing-conversions (#56904)

paddle/fluid/framework/block_desc.cc

@@ -143,18 +143,19 @@ void BlockDesc::PrependAllocatedOp(std::unique_ptr<OpDesc> &&op_desc) {
 
 OpDesc *BlockDesc::InsertOp(size_t index) {
   need_update_ = true;
-  auto it = ops_.begin() + index;
+  auto it = ops_.begin() + index;  // NOLINT
   std::unique_ptr<OpDesc> new_op(new OpDesc(this));
   it = ops_.insert(it, std::move(new_op));
   return (*it).get();
 }
 
 void BlockDesc::RemoveOp(size_t s, size_t e) {
-  if (ops_.begin() + s >= ops_.end() || ops_.begin() + e > ops_.end()) {
+  if (ops_.begin() + s >= ops_.end() ||  // NOLINT
+      ops_.begin() + e > ops_.end()) {   // NOLINT
     return;
   }
   need_update_ = true;
-  ops_.erase(ops_.begin() + s, ops_.begin() + e);
+  ops_.erase(ops_.begin() + s, ops_.begin() + e);  // NOLINT
 }
 
 void BlockDesc::RemoveOpInternal(const OpDesc *op_desc) {

paddle/fluid/framework/custom_operator.cc

@@ -653,8 +653,8 @@ static void RunDefaultInferDtypeFunc(
       if (detail::IsDuplicableVar(pair.first)) {
         size_t size = ctx->InputSize(pair.first);
         for (size_t i = 0; i < size; ++i) {
-          auto dtype = ctx->GetInputDataType(pair.first, i);
-          ctx->SetOutputDataType(pair.second, dtype, i);
+          auto dtype = ctx->GetInputDataType(pair.first, static_cast<int>(i));
+          ctx->SetOutputDataType(pair.second, dtype, static_cast<int>(i));
         }
       } else {
         auto dtype = ctx->GetInputDataType(pair.first);
@@ -681,7 +681,7 @@ static void RunInferDtypeFunc(
       std::vector<DataType> vec_custom_dtype;
       if (ctx->HasInput(in_name)) {  // general inputs
         for (size_t i = 0; i < ctx->InputSize(in_name); ++i) {
-          auto dtype = ctx->GetInputDataType(in_name, i);
+          auto dtype = ctx->GetInputDataType(in_name, static_cast<int>(i));
           vec_custom_dtype.emplace_back(
               paddle::framework::TransToPhiDataType(dtype));
         }
@@ -799,8 +799,8 @@ static void RunInferDtypeFunc(
       if (ctx->HasOutput(out_name)) {
         size_t size = ctx->InputSize(in_name);
         for (size_t i = 0; i < size; ++i) {
-          auto dtype = ctx->GetInputDataType(in_name, i);
-          ctx->SetOutputDataType(out_name, dtype, i);
+          auto dtype = ctx->GetInputDataType(in_name, static_cast<int>(i));
+          ctx->SetOutputDataType(out_name, dtype, static_cast<int>(i));
         }
       } else {
         PADDLE_ENFORCE(

paddle/fluid/framework/data_feed.cc

@@ -45,7 +45,9 @@ class BufferedLineFileReader {
   class FILEReader {
    public:
     explicit FILEReader(FILE* fp) : fp_(fp) {}
-    int read(char* buf, int len) { return fread(buf, sizeof(char), len, fp_); }
+    int read(char* buf, int len) {
+      return static_cast<int>(fread(buf, sizeof(char), len, fp_));
+    }
 
    private:
     FILE* fp_;
@@ -644,10 +646,11 @@ void MultiSlotDataFeed::Init(
   use_slots_.clear();
   use_slots_is_dense_.clear();
   for (size_t i = 0; i < all_slot_num; ++i) {
-    const auto& slot = multi_slot_desc.slots(i);
+    const auto& slot = multi_slot_desc.slots(static_cast<int>(i));
     all_slots_[i] = slot.name();
     all_slots_type_[i] = slot.type();
-    use_slots_index_[i] = slot.is_used() ? use_slots_.size() : -1;
+    use_slots_index_[i] =
+        static_cast<int>(slot.is_used() ? use_slots_.size() : -1);
     total_dims_without_inductive_[i] = 1;
     inductive_shape_index_[i] = -1;
     if (slot.is_used()) {
@@ -1048,10 +1051,11 @@ void MultiSlotInMemoryDataFeed::Init(
   use_slots_is_dense_.clear();
   slot_conf_.resize(all_slot_num);
   for (size_t i = 0; i < all_slot_num; ++i) {
-    const auto& slot = multi_slot_desc.slots(i);
+    const auto& slot = multi_slot_desc.slots(static_cast<int>(i));
     all_slots_[i] = slot.name();
     all_slots_type_[i] = slot.type();
-    use_slots_index_[i] = slot.is_used() ? use_slots_.size() : -1;
+    use_slots_index_[i] =
+        static_cast<int>(slot.is_used() ? use_slots_.size() : -1);
 
     slot_conf_[i].name = slot.name();
     slot_conf_[i].type = slot.type();
@@ -1839,21 +1843,21 @@ void PaddleBoxDataFeed::GetRankOffset(const std::vector<PvInstance>& pv_vec,
   int max_rank = 3;  // the value is setting
   int row = ins_number;
   int col = max_rank * 2 + 1;
-  int pv_num = pv_vec.size();
+  int pv_num = static_cast<int>(pv_vec.size());
 
   std::vector<int> rank_offset_mat(row * col, -1);
   rank_offset_mat.shrink_to_fit();
 
   for (int i = 0; i < pv_num; i++) {
     auto pv_ins = pv_vec[i];
-    int ad_num = pv_ins->ads.size();
+    int ad_num = static_cast<int>(pv_ins->ads.size());
     int index_start = index;
     for (int j = 0; j < ad_num; ++j) {
       auto ins = pv_ins->ads[j];
       int rank = -1;
       if ((ins->cmatch == 222 || ins->cmatch == 223) &&
           ins->rank <= static_cast<uint32_t>(max_rank) && ins->rank != 0) {
-        rank = ins->rank;
+        rank = static_cast<int>(ins->rank);
       }
 
       rank_offset_mat[index * col] = rank;
@@ -1864,12 +1868,13 @@ void PaddleBoxDataFeed::GetRankOffset(const std::vector<PvInstance>& pv_vec,
           if ((cur_ins->cmatch == 222 || cur_ins->cmatch == 223) &&
               cur_ins->rank <= static_cast<uint32_t>(max_rank) &&
               cur_ins->rank != 0) {
-            fast_rank = cur_ins->rank;
+            fast_rank = static_cast<int>(cur_ins->rank);
           }
 
           if (fast_rank > 0) {
             int m = fast_rank - 1;
-            rank_offset_mat[index * col + 2 * m + 1] = cur_ins->rank;
+            rank_offset_mat[index * col + 2 * m + 1] =
+                static_cast<int>(cur_ins->rank);
             rank_offset_mat[index * col + 2 * m + 2] = index_start + k;
           }
         }
@@ -2035,7 +2040,7 @@ void SlotRecordInMemoryDataFeed::Init(const DataFeedDesc& data_feed_desc) {
   float_total_dims_size_ = 0;
   float_total_dims_without_inductives_.clear();
   for (size_t i = 0; i < all_slot_num; ++i) {
-    const auto& slot = multi_slot_desc.slots(i);
+    const auto& slot = multi_slot_desc.slots(static_cast<int>(i));
     all_slots_[i] = slot.name();
 
     AllSlotInfo& all_slot = all_slots_info_[i];
@@ -2046,7 +2051,7 @@ void SlotRecordInMemoryDataFeed::Init(const DataFeedDesc& data_feed_desc) {
 
     if (slot.is_used()) {
       UsedSlotInfo& info = used_slots_info_[use_slot_size_];
-      info.idx = i;
+      info.idx = static_cast<int>(i);
       info.slot = slot.name();
       info.type = slot.type();
       info.dense = slot.is_dense();
@@ -2423,20 +2428,20 @@ bool SlotRecordInMemoryDataFeed::ParseOneInstance(const std::string& line,
   slot_uint64_feasigns.resize(uint64_use_slot_size_);
 
   if (parse_ins_id_) {
-    int num = strtol(&str[pos], &endptr, 10);
+    int num = static_cast<int>(strtol(&str[pos], &endptr, 10));
     CHECK(num == 1);  // NOLINT
-    pos = endptr - str + 1;
+    pos = static_cast<int>(endptr - str + 1);
     size_t len = 0;
     while (str[pos + len] != ' ') {
       ++len;
     }
     rec->ins_id_ = std::string(str + pos, len);
-    pos += len + 1;
+    pos += static_cast<int>(len + 1);
   }
   if (parse_logkey_) {
-    int num = strtol(&str[pos], &endptr, 10);
+    int num = static_cast<int>(strtol(&str[pos], &endptr, 10));
     CHECK(num == 1);  // NOLINT
-    pos = endptr - str + 1;
+    pos = static_cast<int>(endptr - str + 1);
     size_t len = 0;
     while (str[pos + len] != ' ') {
       ++len;
@@ -2452,14 +2457,14 @@ bool SlotRecordInMemoryDataFeed::ParseOneInstance(const std::string& line,
     rec->search_id = search_id;
     rec->cmatch = cmatch;
     rec->rank = rank;
-    pos += len + 1;
+    pos += static_cast<int>(len + 1);
   }
 
   int float_total_slot_num = 0;
   int uint64_total_slot_num = 0;
 
   for (auto& info : all_slots_info_) {
-    int num = strtol(&str[pos], &endptr, 10);
+    int num = static_cast<int>(strtol(&str[pos], &endptr, 10));
     PADDLE_ENFORCE(num,
                    "The number of ids can not be zero, you need padding "
                    "it in data generator; or if there is something wrong with "
@@ -2488,7 +2493,7 @@ bool SlotRecordInMemoryDataFeed::ParseOneInstance(const std::string& line,
           ++uint64_total_slot_num;
         }
       }
-      pos = endptr - str;
+      pos = static_cast<int>(endptr - str);
     } else {
       for (int j = 0; j <= num; ++j) {
         // pos = line.find_first_of(' ', pos + 1);
@@ -2565,7 +2570,7 @@ void SlotRecordInMemoryDataFeed::PutToFeedVec(const SlotRecord* ins_vec,
         batch_fea.resize(total_instance + fea_num);
         memcpy(
             &batch_fea[total_instance], slot_values, sizeof(float) * fea_num);
-        total_instance += fea_num;
+        total_instance += static_cast<int>(fea_num);
         slot_offset.push_back(total_instance);
       }
 
@@ -2588,7 +2593,7 @@ void SlotRecordInMemoryDataFeed::PutToFeedVec(const SlotRecord* ins_vec,
           memcpy(&batch_fea[total_instance],
                  slot_values,
                  sizeof(uint64_t) * fea_num);
-          total_instance += fea_num;
+          total_instance += static_cast<int>(fea_num);
         }
         if (fea_num == 0) {
           batch_fea.resize(total_instance + fea_num);

paddle/fluid/framework/data_set.cc

@@ -295,7 +295,7 @@ static void compute_batch_num(const int64_t ins_num,
   int thread_batch_num = batch_size * thread_num;
   // less data
   if (static_cast<int64_t>(thread_batch_num) > ins_num) {
-    compute_left_batch_num(ins_num, thread_num, offset, 0);
+    compute_left_batch_num(static_cast<int>(ins_num), thread_num, offset, 0);
     return;
   }
 
@@ -1271,7 +1271,7 @@ void MultiSlotDataset::PreprocessInstance() {
     input_channel_->Close();
     std::vector<PvInstance> pv_data;
     input_channel_->ReadAll(input_records_);
-    int all_records_num = input_records_.size();
+    int all_records_num = static_cast<int>(input_records_.size());
     std::vector<Record*> all_records;
     all_records.reserve(all_records_num);
     for (int index = 0; index < all_records_num; ++index) {
@@ -1333,7 +1333,7 @@ void MultiSlotDataset::GenerateLocalTablesUnlock(int table_id,
       local_map_tables = fleet_ptr_->GetLocalTable();
   local_map_tables.resize(shard_num);
   // read thread
-  int channel_num = multi_output_channel_.size();
+  int channel_num = static_cast<int>(multi_output_channel_.size());
   if (read_thread_num < channel_num) {
     read_thread_num = channel_num;
   }
@@ -1361,7 +1361,8 @@ void MultiSlotDataset::GenerateLocalTablesUnlock(int table_id,
         this->multi_output_channel_[i]->ReadAll(vec_data);
         for (auto& item : vec_data) {
           for (auto& feature : item.uint64_feasigns_) {
-            int shard = feature.sign().uint64_feasign_ % shard_num;
+            int shard =
+                static_cast<int>(feature.sign().uint64_feasign_ % shard_num);
             task_keys[shard].push_back(feature.sign().uint64_feasign_);
           }
         }
@@ -1634,11 +1635,11 @@ void MultiSlotDataset::PreprocessChannel(
   int out_channel_size = 0;
   if (cur_channel_ == 0) {
     for (auto& item : multi_output_channel_) {
-      out_channel_size += item->Size();
+      out_channel_size += static_cast<int>(item->Size());
     }
   } else {
     for (auto& item : multi_consume_channel_) {
-      out_channel_size += item->Size();
+      out_channel_size += static_cast<int>(item->Size());
     }
   }
   VLOG(2) << "DatasetImpl<T>::SlotsShuffle() begin with input channel size: "
@@ -1724,14 +1725,14 @@ void MultiSlotDataset::PreprocessChannel(
       if (!item) {
         continue;
       }
-      end_size += item->Size();
+      end_size += static_cast<int>(item->Size());
     }
   } else {
     for (auto& item : multi_consume_channel_) {
       if (!item) {
         continue;
       }
-      end_size += item->Size();
+      end_size += static_cast<int>(item->Size());
     }
   }
   CHECK(input_channel_->Size() == 0)

paddle/fluid/framework/details/fetch_async_op_handle.cc

@@ -174,7 +174,7 @@ void FetchAsyncOpHandle::FetchMergedLodTensor(
 
   // for 0D tensor, can't concat eath tensor. So stack 0D and concat 1+D tensor
   if (rank == 0) {
-    int src_lodtensor_size = src_lodtensors.size();
+    int src_lodtensor_size = static_cast<int>(src_lodtensors.size());
     new_dim = phi::make_ddim(std::vector<int>({src_lodtensor_size}));
   } else {
     bool find_first_dims = false;
@@ -211,7 +211,7 @@ void FetchAsyncOpHandle::FetchMergedLodTensor(
     if (rank == 0) {
       end = begin + 1;
     } else {
-      end = begin + src->dims()[0];
+      end = static_cast<int>(begin + src->dims()[0]);
     }
 
     if (end == begin) {

paddle/fluid/framework/details/fused_all_reduce_op_handle.cc

@@ -401,7 +401,8 @@ void FusedAllReduceOpHandle::GetDTypeAndNumel(
             "The size of grad tensors of fused_all_reduce_op_handle  "
             "must be > 0, but got %d.",
             len));
-    *numel += phi::Alignment(len * size_of_dtype, places_[0]) / size_of_dtype;
+    *numel += static_cast<int64_t>(
+        phi::Alignment(len * size_of_dtype, places_[0]) / size_of_dtype);
   }
 }
 

paddle/fluid/framework/details/fused_broadcast_op_handle.cc

@@ -49,8 +49,9 @@ void FusedBroadcastOpHandle::RunImpl() {
   for (size_t i = 0; i < in_var_handles.size(); ++i) {
     BroadcastOneVar(
         *in_var_handles[i],
-        std::vector<VarHandle *>(out_var_handles.begin() + i * place_num,
-                                 out_var_handles.begin() + (i + 1) * place_num),
+        std::vector<VarHandle *>(
+            out_var_handles.begin() + i * place_num,         // NOLINT
+            out_var_handles.begin() + (i + 1) * place_num),  // NOLINT
         local_exec_scopes_);
   }
 }

paddle/fluid/framework/details/gather_op_handle.cc

@@ -129,7 +129,7 @@ void GatherOpHandle::RunImpl() {
       out_var_handle->place(), [in_tensors, out_tensor, &dev_ctx, t_out_p] {
         int s = 0, e = 0;
         for (const auto &in_tensor : in_tensors) {
-          e += in_tensor.dims()[0];
+          e += static_cast<int>(in_tensor.dims()[0]);
           auto sub_out = out_tensor->Slice(s, e);
           paddle::framework::TensorCopy(in_tensor, t_out_p, *dev_ctx, &sub_out);
           s = e;

paddle/fluid/framework/details/gather_op_handle_test.cc

@@ -134,12 +134,12 @@ struct TestGatherOpHandle {
   }
 
   void TestGatherSelectedRows(size_t output_scope_idx) {
-    int height = kDims[0] * 2;
+    int height = static_cast<int>(kDims[0] * 2);
     std::vector<int64_t> rows{0, 1, 2, 3, 3, 0, 14, 7, 3, 1,
                               2, 4, 6, 3, 1, 1, 1,  1, 3, 7};
     std::vector<float> send_vector(phi::product(kDims));
     for (size_t k = 0; k < send_vector.size(); ++k) {
-      send_vector[k] = k;
+      send_vector[k] = static_cast<float>(k);
     }
 
     for (size_t input_scope_idx = 0; input_scope_idx < gpu_list_.size();

paddle/fluid/framework/details/scale_loss_grad_op_handle.cc

@@ -33,7 +33,7 @@ ScaleLossGradOpHandle::ScaleLossGradOpHandle(ir::Node *node,
                                              platform::DeviceContext *dev_ctx,
                                              proto::VarType::Type dtype)
     : OpHandleBase(node),
-      coeff_(static_cast<float>(1.0 / num_dev)),
+      coeff_(static_cast<float>(1.0 / num_dev)),  // NOLINT
       scope_(scope),
       place_(place),
       out_dtype_(dtype) {

paddle/fluid/framework/details/scope_buffered_monitor.cc

@@ -169,14 +169,15 @@ void ScopeBufferedMonitor::Apply(const std::function<void()> &callback,
   if (VLOG_IS_ON(8)) {
     for (size_t idx = 0; idx < gpu_memory_size_per_gpu.size(); ++idx) {
       VLOG(8) << "history local exec scopes contains "
-              << string::HumanReadableSize(gpu_memory_size_per_gpu.at(idx))
+              << string::HumanReadableSize(
+                     gpu_memory_size_per_gpu.at(idx))  // NOLINT
               << " in " << places_.at(idx);
     }
   }
 
   if (FLAGS_local_exe_sub_scope_limit > 0) {
     for (size_t idx = 0; idx < gpu_memory_size_per_gpu.size(); ++idx) {
-      if (gpu_memory_size_per_gpu.at(idx) / kMB >=
+      if (gpu_memory_size_per_gpu.at(idx) / kMB >=  // NOLINT
           FLAGS_local_exe_sub_scope_limit) {
         platform::DeviceContextPool::Instance().Get(places_.at(idx))->Wait();
         local_exec_scopes_.at(idx)->DropKids();