Fix compiler warnings. (#7974)

- Remove unused parameters. There are still many warnings that are not yet
addressed. Currently, the warnings in dmlc-core dominate the error log.
- Remove `distributed` parameter from metric.
- Fixes some warnings about signed comparison.

cmake/Utils.cmake

@@ -144,6 +144,15 @@ function(xgboost_set_cuda_flags target)
     set_property(TARGET ${target} PROPERTY CUDA_ARCHITECTURES ${CMAKE_CUDA_ARCHITECTURES})
   endif (CMAKE_VERSION VERSION_GREATER_EQUAL "3.18")
 
+  if (FORCE_COLORED_OUTPUT)
+    if (FORCE_COLORED_OUTPUT AND (CMAKE_GENERATOR STREQUAL "Ninja") AND
+        ((CMAKE_CXX_COMPILER_ID STREQUAL "GNU") OR
+          (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")))
+      target_compile_options(${target} PRIVATE
+        $<$<COMPILE_LANGUAGE:CUDA>:-Xcompiler=-fdiagnostics-color=always>)
+    endif()
+  endif (FORCE_COLORED_OUTPUT)
+
   if (USE_DEVICE_DEBUG)
     target_compile_options(${target} PRIVATE
       $<$<AND:$<CONFIG:DEBUG>,$<COMPILE_LANGUAGE:CUDA>>:-G;-src-in-ptx>)

include/xgboost/gbm.h

@@ -68,8 +68,8 @@ class GradientBooster : public Model, public Configurable {
    * \param layer_end   End of booster layer. 0 means do not limit trees.
    * \param out         Output gradient booster
    */
-  virtual void Slice(int32_t layer_begin, int32_t layer_end, int32_t step,
+  virtual void Slice(int32_t /*layer_begin*/, int32_t /*layer_end*/, int32_t /*step*/,
-                     GradientBooster *out, bool* out_of_bound) const {
+                     GradientBooster* /*out*/, bool* /*out_of_bound*/) const {
     LOG(FATAL) << "Slice is not supported by current booster.";
   }
   /*!

include/xgboost/json_io.h

@@ -89,7 +89,7 @@ class JsonReader {
     } else if (got == 0) {
       msg += "\\0\"";
     } else {
-      msg += (got <= 127 ? std::string{got} : std::to_string(got)) + " \"";  // NOLINT
+      msg += (got <= static_cast<char>(127) ? std::string{got} : std::to_string(got)) + " \"";
     }
     Error(msg);
   }

include/xgboost/linalg.h

@@ -317,7 +317,8 @@ class TensorView {
   }
 
   template <size_t old_dim, size_t new_dim, int32_t D, typename Index>
-  LINALG_HD size_t MakeSliceDim(size_t new_shape[D], size_t new_stride[D], Index i) const {
+  LINALG_HD size_t MakeSliceDim(DMLC_ATTRIBUTE_UNUSED size_t new_shape[D],
+                                DMLC_ATTRIBUTE_UNUSED size_t new_stride[D], Index i) const {
     static_assert(old_dim < kDim, "");
     return stride_[old_dim] * i;
   }

include/xgboost/metric.h

@@ -57,12 +57,8 @@ class Metric : public Configurable {
    * \brief evaluate a specific metric
    * \param preds prediction
    * \param info information, including label etc.
-   * \param distributed whether a call to Allreduce is needed to gather
-   *        the average statistics across all the node,
-   *        this is only supported by some metrics
    */
-  virtual double Eval(const HostDeviceVector<bst_float> &preds,
+  virtual double Eval(const HostDeviceVector<bst_float>& preds, const MetaInfo& info) = 0;
-                      const MetaInfo &info, bool distributed) = 0;
   /*! \return name of metric */
   virtual const char* Name() const = 0;
   /*! \brief virtual destructor */

include/xgboost/objective.h

@@ -103,8 +103,10 @@ class ObjFunction : public Configurable {
    * \param prediction Model prediction after transformation.
    * \param p_tree Tree that needs to be updated.
    */
-  virtual void UpdateTreeLeaf(HostDeviceVector<bst_node_t> const& position, MetaInfo const& info,
+  virtual void UpdateTreeLeaf(HostDeviceVector<bst_node_t> const& /*position*/,
-                              HostDeviceVector<float> const& prediction, RegTree* p_tree) const {}
+                              MetaInfo const& /*info*/,
+                              HostDeviceVector<float> const& /*prediction*/,
+                              RegTree* /*p_tree*/) const {}
 
   /*!
    * \brief Create an objective function according to name.

src/common/hist_util.h

@@ -171,14 +171,14 @@ inline HistogramCuts SketchOnDMatrix(DMatrix* m, int32_t max_bins, int32_t n_thr
 
   if (!use_sorted) {
     HostSketchContainer container(max_bins, m->Info(), reduced, HostSketchContainer::UseGroup(info),
-                                  hessian, n_threads);
+                                  n_threads);
     for (auto const& page : m->GetBatches<SparsePage>()) {
       container.PushRowPage(page, info, hessian);
     }
     container.MakeCuts(&out);
   } else {
-    SortedSketchContainer container{
+    SortedSketchContainer container{max_bins, m->Info(), reduced,
-        max_bins, m->Info(), reduced, HostSketchContainer::UseGroup(info), hessian, n_threads};
+                                    HostSketchContainer::UseGroup(info), n_threads};
     for (auto const& page : m->GetBatches<SortedCSCPage>()) {
       container.PushColPage(page, info, hessian);
     }

src/common/partition_builder.h

@@ -168,8 +168,8 @@ class PartitionBuilder {
     const size_t n_left  = child_nodes_sizes.first;
     const size_t n_right = child_nodes_sizes.second;
 
-    SetNLeftElems(node_in_set, range.begin(), range.end(), n_left);
+    SetNLeftElems(node_in_set, range.begin(), n_left);
-    SetNRightElems(node_in_set, range.begin(), range.end(), n_right);
+    SetNRightElems(node_in_set, range.begin(), n_right);
   }
 
   /**
@@ -188,8 +188,7 @@ class PartitionBuilder {
    */
   template <typename Pred>
   void PartitionRange(const size_t node_in_set, const size_t nid, common::Range1d range,
-                      bst_feature_t fidx, common::RowSetCollection* p_row_set_collection,
+                      common::RowSetCollection* p_row_set_collection, Pred pred) {
-                      Pred pred) {
     auto& row_set_collection = *p_row_set_collection;
     const size_t* p_ridx = row_set_collection[nid].begin;
     common::Span<const size_t> ridx(p_ridx + range.begin(), p_ridx + range.end());
@@ -200,8 +199,8 @@ class PartitionBuilder {
     const size_t n_left = child_nodes_sizes.first;
     const size_t n_right = child_nodes_sizes.second;
 
-    this->SetNLeftElems(node_in_set, range.begin(), range.end(), n_left);
+    this->SetNLeftElems(node_in_set, range.begin(), n_left);
-    this->SetNRightElems(node_in_set, range.begin(), range.end(), n_right);
+    this->SetNRightElems(node_in_set, range.begin(), n_right);
   }
 
   // allocate thread local memory, should be called for each specific task
@@ -223,12 +222,12 @@ class PartitionBuilder {
     return { mem_blocks_.at(task_idx)->Right(), end - begin };
   }
 
-  void SetNLeftElems(int nid, size_t begin, size_t end, size_t n_left) {
+  void SetNLeftElems(int nid, size_t begin, size_t n_left) {
     size_t task_idx = GetTaskIdx(nid, begin);
     mem_blocks_.at(task_idx)->n_left = n_left;
   }
 
-  void SetNRightElems(int nid, size_t begin, size_t end, size_t n_right) {
+  void SetNRightElems(int nid, size_t begin, size_t n_right) {
     size_t task_idx = GetTaskIdx(nid, begin);
     mem_blocks_.at(task_idx)->n_right = n_right;
   }

src/common/quantile.cc

@@ -543,7 +543,7 @@ template class SketchContainerImpl<WXQuantileSketch<float, float>>;
 
 HostSketchContainer::HostSketchContainer(int32_t max_bins, MetaInfo const &info,
                                          std::vector<size_t> columns_size, bool use_group,
-                                         Span<float const> hessian, int32_t n_threads)
+                                         int32_t n_threads)
     : SketchContainerImpl{columns_size, max_bins, info.feature_types.ConstHostSpan(), use_group,
                           n_threads} {
   monitor_.Init(__func__);

src/common/quantile.h

@@ -774,7 +774,7 @@ class HostSketchContainer : public SketchContainerImpl<WQuantileSketch<float, fl
 
  public:
   HostSketchContainer(int32_t max_bins, MetaInfo const &info, std::vector<size_t> columns_size,
-                      bool use_group, Span<float const> hessian, int32_t n_threads);
+                      bool use_group, int32_t n_threads);
 };
 
 /**
@@ -868,7 +868,7 @@ class SortedSketchContainer : public SketchContainerImpl<WXQuantileSketch<float,
  public:
   explicit SortedSketchContainer(int32_t max_bins, MetaInfo const &info,
                                  std::vector<size_t> columns_size, bool use_group,
-                                 Span<float const> hessian, int32_t n_threads)
+                                 int32_t n_threads)
       : SketchContainerImpl{columns_size, max_bins, info.feature_types.ConstHostSpan(), use_group,
                             n_threads} {
     monitor_.Init(__func__);

src/data/iterative_device_dmatrix.cu

@@ -163,8 +163,7 @@ void IterativeDeviceDMatrix::Initialize(DataIterHandle iter_handle, float missin
 
 BatchSet<EllpackPage> IterativeDeviceDMatrix::GetEllpackBatches(const BatchParam& param) {
   CHECK(page_);
-  auto begin_iter =
+  auto begin_iter = BatchIterator<EllpackPage>(new SimpleBatchIteratorImpl<EllpackPage>(page_));
-      BatchIterator<EllpackPage>(new SimpleBatchIteratorImpl<EllpackPage>(page_));
   return BatchSet<EllpackPage>(begin_iter);
 }
 }  // namespace data

src/data/iterative_device_dmatrix.h

@@ -45,8 +45,8 @@ class IterativeDeviceDMatrix : public DMatrix {
 
   bool EllpackExists() const override { return true; }
   bool SparsePageExists() const override { return false; }
-  DMatrix *Slice(common::Span<int32_t const> ridxs) override {
+  DMatrix *Slice(common::Span<int32_t const>) override {
-    LOG(FATAL) << "Slicing DMatrix is not supported for Device DMatrix.";
+    LOG(FATAL) << "Slicing DMatrix is not supported for Quantile DMatrix.";
     return nullptr;
   }
   BatchSet<SparsePage> GetRowBatches() override {

src/data/proxy_dmatrix.h

@@ -84,7 +84,7 @@ class DMatrixProxy : public DMatrix {
   bool SingleColBlock() const override { return true; }
   bool EllpackExists() const override { return true; }
   bool SparsePageExists() const override { return false; }
-  DMatrix *Slice(common::Span<int32_t const> ridxs) override {
+  DMatrix* Slice(common::Span<int32_t const> /*ridxs*/) override {
     LOG(FATAL) << "Slicing DMatrix is not supported for Proxy DMatrix.";
     return nullptr;
   }
@@ -100,7 +100,7 @@ class DMatrixProxy : public DMatrix {
     LOG(FATAL) << "Not implemented.";
     return BatchSet<SortedCSCPage>(BatchIterator<SortedCSCPage>(nullptr));
   }
-  BatchSet<EllpackPage> GetEllpackBatches(const BatchParam& param) override {
+  BatchSet<EllpackPage> GetEllpackBatches(const BatchParam&) override {
     LOG(FATAL) << "Not implemented.";
     return BatchSet<EllpackPage>(BatchIterator<EllpackPage>(nullptr));
   }

src/gbm/gbtree.cc

@@ -218,7 +218,7 @@ void CopyGradient(HostDeviceVector<GradientPair> const* in_gpair, int32_t n_thre
 }
 
 void GBTree::UpdateTreeLeaf(DMatrix const* p_fmat, HostDeviceVector<float> const& predictions,
-                            ObjFunction const* obj, size_t gidx,
+                            ObjFunction const* obj,
                             std::vector<std::unique_ptr<RegTree>>* p_trees) {
   CHECK(!updaters_.empty());
   if (!updaters_.back()->HasNodePosition()) {
@@ -257,7 +257,7 @@ void GBTree::DoBoost(DMatrix* p_fmat, HostDeviceVector<GradientPair>* in_gpair,
   if (ngroup == 1) {
     std::vector<std::unique_ptr<RegTree>> ret;
     BoostNewTrees(in_gpair, p_fmat, 0, &ret);
-    UpdateTreeLeaf(p_fmat, predt->predictions, obj, 0, &ret);
+    UpdateTreeLeaf(p_fmat, predt->predictions, obj, &ret);
     const size_t num_new_trees = ret.size();
     new_trees.push_back(std::move(ret));
     auto v_predt = out.Slice(linalg::All(), 0);
@@ -274,7 +274,7 @@ void GBTree::DoBoost(DMatrix* p_fmat, HostDeviceVector<GradientPair>* in_gpair,
       CopyGradient(in_gpair, ctx_->Threads(), ngroup, gid, &tmp);
       std::vector<std::unique_ptr<RegTree>> ret;
       BoostNewTrees(&tmp, p_fmat, gid, &ret);
-      UpdateTreeLeaf(p_fmat, predt->predictions, obj, gid, &ret);
+      UpdateTreeLeaf(p_fmat, predt->predictions, obj, &ret);
       const size_t num_new_trees = ret.size();
       new_trees.push_back(std::move(ret));
       auto v_predt = out.Slice(linalg::All(), gid);
@@ -289,7 +289,7 @@ void GBTree::DoBoost(DMatrix* p_fmat, HostDeviceVector<GradientPair>* in_gpair,
   }
 
   monitor_.Stop("BoostNewTrees");
-  this->CommitModel(std::move(new_trees), p_fmat, predt);
+  this->CommitModel(std::move(new_trees));
 }
 
 void GBTree::InitUpdater(Args const& cfg) {
@@ -378,9 +378,7 @@ void GBTree::BoostNewTrees(HostDeviceVector<GradientPair>* gpair, DMatrix* p_fma
   }
 }
 
-void GBTree::CommitModel(std::vector<std::vector<std::unique_ptr<RegTree>>>&& new_trees,
+void GBTree::CommitModel(std::vector<std::vector<std::unique_ptr<RegTree>>>&& new_trees) {
-                         DMatrix* m,
-                         PredictionCacheEntry* predts) {
   monitor_.Start("CommitModel");
   for (uint32_t gid = 0; gid < model_.learner_model_param->num_output_group; ++gid) {
     model_.CommitModel(std::move(new_trees[gid]), gid);
@@ -490,15 +488,14 @@ void GBTree::Slice(int32_t layer_begin, int32_t layer_end, int32_t step,
            "want to update a portion of trees.";
   }
 
-  *out_of_bound = detail::SliceTrees(
+  *out_of_bound = detail::SliceTrees(layer_begin, layer_end, step, this->model_, layer_trees,
-      layer_begin, layer_end, step, this->model_, tparam_, layer_trees,
+                                     [&](auto const& in_it, auto const& out_it) {
-      [&](auto const &in_it, auto const &out_it) {
+                                       auto new_tree =
-        auto new_tree =
+                                           std::make_unique<RegTree>(*this->model_.trees.at(in_it));
-            std::make_unique<RegTree>(*this->model_.trees.at(in_it));
+                                       bst_group_t group = this->model_.tree_info[in_it];
-        bst_group_t group = this->model_.tree_info[in_it];
+                                       out_trees.at(out_it) = std::move(new_tree);
-        out_trees.at(out_it) = std::move(new_tree);
+                                       out_trees_info.at(out_it) = group;
-        out_trees_info.at(out_it) = group;
+                                     });
-      });
 }
 
 void GBTree::PredictBatch(DMatrix* p_fmat,
@@ -674,11 +671,10 @@ class Dart : public GBTree {
     auto p_dart = dynamic_cast<Dart*>(out);
     CHECK(p_dart);
     CHECK(p_dart->weight_drop_.empty());
-    detail::SliceTrees(
+    detail::SliceTrees(layer_begin, layer_end, step, model_, this->LayerTrees(),
-        layer_begin, layer_end, step, model_, tparam_, this->LayerTrees(),
+                       [&](auto const& in_it, auto const&) {
-        [&](auto const& in_it, auto const&) {
+                         p_dart->weight_drop_.push_back(this->weight_drop_.at(in_it));
-          p_dart->weight_drop_.push_back(this->weight_drop_.at(in_it));
+                       });
-        });
   }
 
   void SaveModel(Json *p_out) const override {
@@ -901,9 +897,7 @@ class Dart : public GBTree {
 
  protected:
   // commit new trees all at once
-  void
+  void CommitModel(std::vector<std::vector<std::unique_ptr<RegTree>>>&& new_trees) override {
-  CommitModel(std::vector<std::vector<std::unique_ptr<RegTree>>>&& new_trees,
-              DMatrix*, PredictionCacheEntry*) override {
     int num_new_trees = 0;
     for (uint32_t gid = 0; gid < model_.learner_model_param->num_output_group; ++gid) {
       num_new_trees += new_trees[gid].size();

src/gbm/gbtree.h

@@ -162,9 +162,8 @@ inline std::pair<uint32_t, uint32_t> LayerToTree(gbm::GBTreeModel const &model,
 
 // Call fn for each pair of input output tree.  Return true if index is out of bound.
 template <typename Func>
-inline bool SliceTrees(int32_t layer_begin, int32_t layer_end, int32_t step,
+bool SliceTrees(int32_t layer_begin, int32_t layer_end, int32_t step, GBTreeModel const& model,
-                       GBTreeModel const &model, GBTreeTrainParam const &tparam,
+                uint32_t layer_trees, Func fn) {
-                       uint32_t layer_trees, Func fn) {
   uint32_t tree_begin, tree_end;
   std::tie(tree_begin, tree_end) = detail::LayerToTree(model, layer_begin, layer_end);
   if (tree_end > model.trees.size()) {
@@ -206,8 +205,7 @@ class GBTree : public GradientBooster {
    * \brief Optionally update the leaf value.
    */
   void UpdateTreeLeaf(DMatrix const* p_fmat, HostDeviceVector<float> const& predictions,
-                      ObjFunction const* obj, size_t gidx,
+                      ObjFunction const* obj, std::vector<std::unique_ptr<RegTree>>* p_trees);
-                      std::vector<std::unique_ptr<RegTree>>* p_trees);
 
   /*! \brief Carry out one iteration of boosting */
   void DoBoost(DMatrix* p_fmat, HostDeviceVector<GradientPair>* in_gpair,
@@ -325,7 +323,7 @@ class GBTree : public GradientBooster {
     };
 
     if (importance_type == "weight") {
-      add_score([&](auto const &p_tree, bst_node_t, bst_feature_t split) {
+      add_score([&](auto const&, bst_node_t, bst_feature_t split) {
         gain_map[split] = split_counts[split];
       });
     } else if (importance_type == "gain" || importance_type == "total_gain") {
@@ -423,9 +421,7 @@ class GBTree : public GradientBooster {
                                                  DMatrix* f_dmat = nullptr) const;
 
   // commit new trees all at once
-  virtual void CommitModel(std::vector<std::vector<std::unique_ptr<RegTree>>>&& new_trees,
+  virtual void CommitModel(std::vector<std::vector<std::unique_ptr<RegTree>>>&& new_trees);
-                           DMatrix* m,
-                           PredictionCacheEntry* predts);
 
   // --- data structure ---
   GBTreeModel model_;

src/learner.cc

@@ -1234,8 +1234,7 @@ class LearnerImpl : public LearnerIO {
 
       obj_->EvalTransform(&out);
       for (auto& ev : metrics_) {
-        os << '\t' << data_names[i] << '-' << ev->Name() << ':'
+        os << '\t' << data_names[i] << '-' << ev->Name() << ':' << ev->Eval(out, m->Info());
-           << ev->Eval(out, m->Info(), tparam_.dsplit == DataSplitMode::kRow);
       }
     }
 

src/metric/auc.cc

@@ -254,8 +254,7 @@ std::pair<double, uint32_t> RankingAUC(std::vector<float> const &predts,
 
 template <typename Curve>
 class EvalAUC : public Metric {
-  double Eval(const HostDeviceVector<bst_float> &preds, const MetaInfo &info,
+  double Eval(const HostDeviceVector<bst_float> &preds, const MetaInfo &info) override {
-              bool distributed) override {
     double auc {0};
     if (tparam_->gpu_id != GenericParameter::kCpuId) {
       preds.SetDevice(tparam_->gpu_id);

src/metric/auc.cu

@@ -312,10 +312,8 @@ void SegmentedReduceAUC(common::Span<size_t const> d_unique_idx,
  * up each class in all kernels.
  */
 template <bool scale, typename Fn>
-double GPUMultiClassAUCOVR(common::Span<float const> predts,
+double GPUMultiClassAUCOVR(MetaInfo const &info, int32_t device, common::Span<uint32_t> d_class_ptr,
-                           MetaInfo const &info, int32_t device,
+                           size_t n_classes, std::shared_ptr<DeviceAUCCache> cache, Fn area_fn) {
-                           common::Span<uint32_t> d_class_ptr, size_t n_classes,
-                           std::shared_ptr<DeviceAUCCache> cache, Fn area_fn) {
   dh::safe_cuda(cudaSetDevice(device));
   /**
    * Sorted idx
@@ -478,8 +476,7 @@ double GPUMultiClassROCAUC(common::Span<float const> predts,
                               double tp, size_t /*class_id*/) {
     return TrapezoidArea(fp_prev, fp, tp_prev, tp);
   };
-  return GPUMultiClassAUCOVR<true>(predts, info, device, dh::ToSpan(class_ptr),
+  return GPUMultiClassAUCOVR<true>(info, device, dh::ToSpan(class_ptr), n_classes, cache, fn);
-                                   n_classes, cache, fn);
 }
 
 namespace {
@@ -704,8 +701,7 @@ double GPUMultiClassPRAUC(common::Span<float const> predts,
     return detail::CalcDeltaPRAUC(fp_prev, fp, tp_prev, tp,
                                   d_totals[class_id].first);
   };
-  return GPUMultiClassAUCOVR<false>(predts, info, device, d_class_ptr,
+  return GPUMultiClassAUCOVR<false>(info, device, d_class_ptr, n_classes, cache, fn);
-                                    n_classes, cache, fn);
 }
 
 template <typename Fn>

src/metric/elementwise_metric.cu

@@ -178,8 +178,7 @@ class PseudoErrorLoss : public Metric {
     out["pseudo_huber_param"] = ToJson(param_);
   }
 
-  double Eval(const HostDeviceVector<bst_float>& preds, const MetaInfo& info,
+  double Eval(const HostDeviceVector<bst_float>& preds, const MetaInfo& info) override {
-              bool distributed) override {
     CHECK_EQ(info.labels.Shape(0), info.num_row_);
     auto labels = info.labels.View(tparam_->gpu_id);
     preds.SetDevice(tparam_->gpu_id);
@@ -197,7 +196,7 @@ class PseudoErrorLoss : public Metric {
           return std::make_tuple(v, wt);
         });
     double dat[2]{result.Residue(), result.Weights()};
-    if (distributed) {
+    if (rabit::IsDistributed()) {
       rabit::Allreduce<rabit::op::Sum>(dat, 2);
     }
     return EvalRowMAPE::GetFinal(dat[0], dat[1]);
@@ -342,8 +341,7 @@ struct EvalEWiseBase : public Metric {
   EvalEWiseBase() = default;
   explicit EvalEWiseBase(char const* policy_param) : policy_{policy_param} {}
 
-  double Eval(HostDeviceVector<bst_float> const& preds, const MetaInfo& info,
+  double Eval(HostDeviceVector<bst_float> const& preds, const MetaInfo& info) override {
-              bool distributed) override {
     CHECK_EQ(preds.Size(), info.labels.Size())
         << "label and prediction size not match, "
         << "hint: use merror or mlogloss for multi-class classification";
@@ -367,10 +365,7 @@ struct EvalEWiseBase : public Metric {
         });
 
     double dat[2]{result.Residue(), result.Weights()};
-
+    rabit::Allreduce<rabit::op::Sum>(dat, 2);
-    if (distributed) {
-      rabit::Allreduce<rabit::op::Sum>(dat, 2);
-    }
     return Policy::GetFinal(dat[0], dat[1]);
   }
 

src/metric/multiclass_metric.cu

@@ -167,8 +167,7 @@ class MultiClassMetricsReduction {
  */
 template<typename Derived>
 struct EvalMClassBase : public Metric {
-  double Eval(const HostDeviceVector<float> &preds, const MetaInfo &info,
+  double Eval(const HostDeviceVector<float> &preds, const MetaInfo &info) override {
-              bool distributed) override {
     if (info.labels.Size() == 0) {
       CHECK_EQ(preds.Size(), 0);
     } else {
@@ -186,9 +185,7 @@ struct EvalMClassBase : public Metric {
       dat[0] = result.Residue();
       dat[1] = result.Weights();
     }
-    if (distributed) {
+    rabit::Allreduce<rabit::op::Sum>(dat, 2);
-      rabit::Allreduce<rabit::op::Sum>(dat, 2);
-    }
     return Derived::GetFinal(dat[0], dat[1]);
   }
   /*!

src/metric/rank_metric.cc

@@ -102,9 +102,8 @@ struct EvalAMS : public Metric {
     name_ = os.str();
   }
 
-  double Eval(const HostDeviceVector<bst_float> &preds, const MetaInfo &info,
+  double Eval(const HostDeviceVector<bst_float>& preds, const MetaInfo& info) override {
-              bool distributed) override {
+    CHECK(!rabit::IsDistributed()) << "metric AMS do not support distributed evaluation";
-    CHECK(!distributed) << "metric AMS do not support distributed evaluation";
     using namespace std;  // NOLINT(*)
 
     const auto ndata = static_cast<bst_omp_uint>(info.labels.Size());
@@ -161,8 +160,7 @@ struct EvalRank : public Metric, public EvalRankConfig {
   std::unique_ptr<xgboost::Metric> rank_gpu_;
 
  public:
-  double Eval(const HostDeviceVector<bst_float> &preds, const MetaInfo &info,
+  double Eval(const HostDeviceVector<bst_float>& preds, const MetaInfo& info) override {
-              bool distributed) override {
     CHECK_EQ(preds.Size(), info.labels.Size())
         << "label size predict size not match";
 
@@ -185,7 +183,7 @@ struct EvalRank : public Metric, public EvalRankConfig {
         rank_gpu_.reset(GPUMetric::CreateGPUMetric(this->Name(), tparam_));
       }
       if (rank_gpu_) {
-        sum_metric = rank_gpu_->Eval(preds, info, distributed);
+        sum_metric = rank_gpu_->Eval(preds, info);
       }
     }
 
@@ -218,7 +216,7 @@ struct EvalRank : public Metric, public EvalRankConfig {
       exc.Rethrow();
     }
 
-    if (distributed) {
+    if (rabit::IsDistributed()) {
       double dat[2]{sum_metric, static_cast<double>(ngroups)};
       // approximately estimate the metric using mean
       rabit::Allreduce<rabit::op::Sum>(dat, 2);
@@ -342,9 +340,8 @@ struct EvalMAP : public EvalRank {
 struct EvalCox : public Metric {
  public:
   EvalCox() = default;
-  double Eval(const HostDeviceVector<bst_float> &preds, const MetaInfo &info,
+  double Eval(const HostDeviceVector<bst_float>& preds, const MetaInfo& info) override {
-              bool distributed) override {
+    CHECK(!rabit::IsDistributed()) << "Cox metric does not support distributed evaluation";
-    CHECK(!distributed) << "Cox metric does not support distributed evaluation";
     using namespace std;  // NOLINT(*)
 
     const auto ndata = static_cast<bst_omp_uint>(info.labels.Size());

src/metric/rank_metric.cu

@@ -29,8 +29,7 @@ DMLC_REGISTRY_FILE_TAG(rank_metric_gpu);
 template <typename EvalMetricT>
 struct EvalRankGpu : public GPUMetric, public EvalRankConfig {
  public:
-  double Eval(const HostDeviceVector<bst_float> &preds, const MetaInfo &info,
+  double Eval(const HostDeviceVector<bst_float> &preds, const MetaInfo &info) override {
-              bool distributed) override {
     // Sanity check is done by the caller
     std::vector<unsigned> tgptr(2, 0);
     tgptr[1] = static_cast<unsigned>(preds.Size());

src/metric/survival_metric.cu

@@ -206,20 +206,15 @@ template <typename Policy> struct EvalEWiseSurvivalBase : public Metric {
     CHECK(tparam_);
   }
 
-  double Eval(const HostDeviceVector<float> &preds, const MetaInfo &info,
+  double Eval(const HostDeviceVector<float>& preds, const MetaInfo& info) override {
-              bool distributed) override {
     CHECK_EQ(preds.Size(), info.labels_lower_bound_.Size());
     CHECK_EQ(preds.Size(), info.labels_upper_bound_.Size());
     CHECK(tparam_);
-    auto result =
+    auto result = reducer_.Reduce(*tparam_, info.weights_, info.labels_lower_bound_,
-        reducer_.Reduce(*tparam_, info.weights_, info.labels_lower_bound_,
+                                  info.labels_upper_bound_, preds);
-                        info.labels_upper_bound_, preds);
 
-    double dat[2] {result.Residue(), result.Weights()};
+    double dat[2]{result.Residue(), result.Weights()};
-
+    rabit::Allreduce<rabit::op::Sum>(dat, 2);
-    if (distributed) {
-      rabit::Allreduce<rabit::op::Sum>(dat, 2);
-    }
     return Policy::GetFinal(dat[0], dat[1]);
   }
 
@@ -240,10 +235,9 @@ struct AFTNLogLikDispatcher : public Metric {
     return "aft-nloglik";
   }
 
-  double Eval(const HostDeviceVector<bst_float> &preds, const MetaInfo &info,
+  double Eval(const HostDeviceVector<bst_float>& preds, const MetaInfo& info) override {
-              bool distributed) override {
     CHECK(metric_) << "AFT metric must be configured first, with distribution type and scale";
-    return metric_->Eval(preds, info, distributed);
+    return metric_->Eval(preds, info);
   }
 
   void Configure(const Args& args) override {

src/tree/gpu_hist/row_partitioner.cuh

@@ -116,7 +116,7 @@ class RowPartitioner {
     Segment segment = ridx_segments_.at(nidx);  // rows belongs to node nidx
     auto d_ridx = ridx_.CurrentSpan();
     auto d_position = position_.CurrentSpan();
-    if (left_counts_.size() <= nidx) {
+    if (left_counts_.size() <= static_cast<size_t>(nidx)) {
       left_counts_.resize((nidx * 2) + 1);
       thrust::fill(left_counts_.begin(), left_counts_.end(), 0);
     }

src/tree/hist/evaluate_splits.h

@@ -203,8 +203,8 @@ class HistEvaluator {
   // Returns the sum of gradients corresponding to the data points that contains
   // a non-missing value for the particular feature fid.
   template <int d_step>
-  GradStats EnumerateSplit(common::HistogramCuts const &cut, common::Span<size_t const> sorted_idx,
+  GradStats EnumerateSplit(common::HistogramCuts const &cut, const common::GHistRow &hist,
-                           const common::GHistRow &hist, bst_feature_t fidx, bst_node_t nidx,
+                           bst_feature_t fidx, bst_node_t nidx,
                            TreeEvaluator::SplitEvaluator<TrainParam> const &evaluator,
                            SplitEntry *p_best) const {
     static_assert(d_step == +1 || d_step == -1, "Invalid step.");
@@ -333,9 +333,9 @@ class HistEvaluator {
             EnumeratePart<-1>(cut, sorted_idx, histogram, fidx, nidx, evaluator, best);
           }
         } else {
-          auto grad_stats = EnumerateSplit<+1>(cut, {}, histogram, fidx, nidx, evaluator, best);
+          auto grad_stats = EnumerateSplit<+1>(cut, histogram, fidx, nidx, evaluator, best);
           if (SplitContainsMissingValues(grad_stats, snode_[nidx])) {
-            EnumerateSplit<-1>(cut, {}, histogram, fidx, nidx, evaluator, best);
+            EnumerateSplit<-1>(cut, histogram, fidx, nidx, evaluator, best);
           }
         }
       }
@@ -440,7 +440,7 @@ template <typename Partitioner, typename ExpandEntry>
 void UpdatePredictionCacheImpl(GenericParameter const *ctx, RegTree const *p_last_tree,
                                std::vector<Partitioner> const &partitioner,
                                HistEvaluator<ExpandEntry> const &hist_evaluator,
-                               TrainParam const &param, linalg::VectorView<float> out_preds) {
+                               linalg::VectorView<float> out_preds) {
   CHECK_GT(out_preds.Size(), 0U);
 
   CHECK(p_last_tree);

src/tree/updater_approx.cc

@@ -116,7 +116,7 @@ class GloablApproxBuilder {
     // Caching prediction seems redundant for approx tree method, as sketching takes up
     // majority of training time.
     CHECK_EQ(out_preds.Size(), data->Info().num_row_);
-    UpdatePredictionCacheImpl(ctx_, p_last_tree_, partitioner_, evaluator_, param_, out_preds);
+    UpdatePredictionCacheImpl(ctx_, p_last_tree_, partitioner_, evaluator_, out_preds);
     monitor_->Stop(__func__);
   }
 

src/tree/updater_approx.h

@@ -83,7 +83,7 @@ class ApproxRowPartitioner {
       const size_t task_id = partition_builder_.GetTaskIdx(node_in_set, r.begin());
       partition_builder_.AllocateForTask(task_id);
       partition_builder_.PartitionRange(
-          node_in_set, nid, r, fidx, &row_set_collection_, [&](size_t row_id) {
+          node_in_set, nid, r, &row_set_collection_, [&](size_t row_id) {
             auto cut_value = SearchCutValue(row_id, fidx, index, cut_ptrs, cut_values);
             if (std::isnan(cut_value)) {
               return candidate.split.DefaultLeft();

src/tree/updater_gpu_hist.cu

@@ -563,7 +563,7 @@ struct GPUHistMakerDevice {
     // when processing a large batch
     this->AllReduceHist(hist_nidx.at(0), reducer, hist_nidx.size());
 
-    for (int i = 0; i < subtraction_nidx.size(); i++) {
+    for (size_t i = 0; i < subtraction_nidx.size(); i++) {
       auto build_hist_nidx = hist_nidx.at(i);
       auto subtraction_trick_nidx = subtraction_nidx.at(i);
       auto parent_nidx = candidates.at(i).nid;

src/tree/updater_quantile_hist.cc

@@ -257,7 +257,7 @@ bool QuantileHistMaker::Builder::UpdatePredictionCache(DMatrix const *data,
   }
   monitor_->Start(__func__);
   CHECK_EQ(out_preds.Size(), data->Info().num_row_);
-  UpdatePredictionCacheImpl(ctx_, p_last_tree_, partitioner_, *evaluator_, param_, out_preds);
+  UpdatePredictionCacheImpl(ctx_, p_last_tree_, partitioner_, *evaluator_, out_preds);
   monitor_->Stop(__func__);
   return true;
 }

tests/cpp/common/test_device_helpers.cu

@@ -67,7 +67,7 @@ TEST(SegmentedUnique, Basic) {
   CHECK_EQ(n_uniques, 5);
 
   std::vector<float> values_sol{0.1f, 0.2f, 0.3f, 0.62448811531066895f, 0.4f};
-  for (auto i = 0 ; i < values_sol.size(); i ++) {
+  for (size_t i = 0 ; i < values_sol.size(); i ++) {
     ASSERT_EQ(d_vals_out[i], values_sol[i]);
   }
 
@@ -84,7 +84,7 @@ TEST(SegmentedUnique, Basic) {
       d_segs_out.data().get(), d_vals_out.data().get(),
       thrust::equal_to<float>{});
   ASSERT_EQ(n_uniques, values.size());
-  for (auto i = 0 ; i < values.size(); i ++) {
+  for (size_t i = 0 ; i < values.size(); i ++) {
     ASSERT_EQ(d_vals_out[i], values[i]);
   }
 }

tests/cpp/common/test_linalg.cc

@@ -315,10 +315,10 @@ TEST(Linalg, Popc) {
 TEST(Linalg, Stack) {
   Tensor<float, 3> l{{2, 3, 4}, kCpuId};
   ElementWiseTransformHost(l.View(kCpuId), omp_get_max_threads(),
-                           [=](size_t i, float v) { return i; });
+                           [=](size_t i, float) { return i; });
   Tensor<float, 3> r_0{{2, 3, 4}, kCpuId};
   ElementWiseTransformHost(r_0.View(kCpuId), omp_get_max_threads(),
-                           [=](size_t i, float v) { return i; });
+                           [=](size_t i, float) { return i; });
 
   Stack(&l, r_0);
 

tests/cpp/common/test_partition_builder.cc

@@ -50,8 +50,8 @@ TEST(PartitionBuilder, BasicTest) {
         right[i] = left_total + value_right++;
       }
 
-      builder.SetNLeftElems(nid, begin, end, n_left);
+      builder.SetNLeftElems(nid, begin, n_left);
-      builder.SetNRightElems(nid, begin, end, n_right);
+      builder.SetNRightElems(nid, begin, n_right);
     }
   }
   builder.CalculateRowOffsets();

tests/cpp/common/test_quantile.cc

@@ -77,7 +77,7 @@ void TestDistributedQuantile(size_t rows, size_t cols) {
   std::vector<float> hessian(rows, 1.0);
   auto hess = Span<float const>{hessian};
 
-  ContainerType<use_column> sketch_distributed(n_bins, m->Info(), column_size, false, hess,
+  ContainerType<use_column> sketch_distributed(n_bins, m->Info(), column_size, false,
                                                OmpGetNumThreads(0));
 
   if (use_column) {
@@ -98,7 +98,7 @@ void TestDistributedQuantile(size_t rows, size_t cols) {
   CHECK_EQ(rabit::GetWorldSize(), 1);
   std::for_each(column_size.begin(), column_size.end(), [=](auto& size) { size *= world; });
   m->Info().num_row_ = world * rows;
-  ContainerType<use_column> sketch_on_single_node(n_bins, m->Info(), column_size, false, hess,
+  ContainerType<use_column> sketch_on_single_node(n_bins, m->Info(), column_size, false,
                                                   OmpGetNumThreads(0));
   m->Info().num_row_ = rows;
 
@@ -190,7 +190,7 @@ TEST(Quantile, SameOnAllWorkers) {
 
   constexpr size_t kRows = 1000, kCols = 100;
   RunWithSeedsAndBins(
-      kRows, [=](int32_t seed, size_t n_bins, MetaInfo const &info) {
+      kRows, [=](int32_t seed, size_t n_bins, MetaInfo const&) {
         auto rank = rabit::GetRank();
         HostDeviceVector<float> storage;
         std::vector<FeatureType> ft(kCols);

tests/cpp/common/test_span.h

@@ -36,7 +36,7 @@ struct TestTestStatus {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     SPAN_ASSERT_TRUE(false, status_);
   }
 };
@@ -49,7 +49,7 @@ struct TestAssignment {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     Span<float> s1;
 
     float arr[] = {3, 4, 5};
@@ -71,7 +71,7 @@ struct TestBeginEnd {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     float arr[16];
     InitializeRange(arr, arr + 16);
 
@@ -93,7 +93,7 @@ struct TestRBeginREnd {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     float arr[16];
     InitializeRange(arr, arr + 16);
 
@@ -121,7 +121,7 @@ struct TestObservers {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     // empty
     {
       float *arr = nullptr;
@@ -148,7 +148,7 @@ struct TestCompare {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     float lhs_arr[16], rhs_arr[16];
     InitializeRange(lhs_arr, lhs_arr + 16);
     InitializeRange(rhs_arr, rhs_arr + 16);
@@ -178,7 +178,7 @@ struct TestIterConstruct {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index.
     Span<float>::iterator it1;
     Span<float>::iterator it2;
     SPAN_ASSERT_TRUE(it1 == it2, status_);
@@ -197,7 +197,7 @@ struct TestIterRef {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     float arr[16];
     InitializeRange(arr, arr + 16);
 
@@ -215,7 +215,7 @@ struct TestIterCalculate {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     float arr[16];
     InitializeRange(arr, arr + 16);
 
@@ -278,7 +278,7 @@ struct TestAsBytes {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     float arr[16];
     InitializeRange(arr, arr + 16);
 
@@ -313,7 +313,7 @@ struct TestAsWritableBytes {
   XGBOOST_DEVICE void operator()() {
     this->operator()(0);
   }
-  XGBOOST_DEVICE void operator()(int _idx) {
+  XGBOOST_DEVICE void operator()(size_t) {  // size_t for CUDA index
     float arr[16];
     InitializeRange(arr, arr + 16);
 

tests/cpp/common/test_threading_utils.cc

@@ -34,9 +34,8 @@ TEST(ParallelFor2d, Test) {
 
   // working space is matrix of size (kDim1 x kDim2)
   std::vector<int> matrix(kDim1 * kDim2, 0);
-  BlockedSpace2d space(kDim1, [&](size_t i) {
+  BlockedSpace2d space(
-      return kDim2;
+      kDim1, [&](size_t) { return kDim2; }, kGrainSize);
-  }, kGrainSize);
 
   auto old = omp_get_max_threads();
   omp_set_num_threads(4);

tests/cpp/helpers.cc

@@ -167,7 +167,7 @@ double GetMultiMetricEval(xgboost::Metric* metric,
   info.weights_.HostVector() = weights;
   info.group_ptr_ = groups;
 
-  return metric->Eval(preds, info, false);
+  return metric->Eval(preds, info);
 }
 
 namespace xgboost {
@@ -653,8 +653,6 @@ class RMMAllocator {};
 
 void DeleteRMMResource(RMMAllocator* r) {}
 
-RMMAllocatorPtr SetUpRMMResourceForCppTests(int argc, char** argv) {
+RMMAllocatorPtr SetUpRMMResourceForCppTests(int, char**) { return {nullptr, DeleteRMMResource}; }
-  return {nullptr, DeleteRMMResource};
-}
 #endif  // !defined(XGBOOST_USE_RMM) || XGBOOST_USE_RMM != 1
 }  // namespace xgboost

tests/cpp/helpers.cu

@@ -29,9 +29,7 @@ int CudaArrayIterForTest::Next() {
 }
 
 
-std::shared_ptr<DMatrix> RandomDataGenerator::GenerateDeviceDMatrix(bool with_label,
+std::shared_ptr<DMatrix> RandomDataGenerator::GenerateDeviceDMatrix() {
-                                                                    bool float_label,
-                                                                    size_t classes) {
   CudaArrayIterForTest iter{this->sparsity_, this->rows_, this->cols_, 1};
   auto m = std::make_shared<data::IterativeDeviceDMatrix>(
       &iter, iter.Proxy(), Reset, Next, std::numeric_limits<float>::quiet_NaN(),

tests/cpp/helpers.h

@@ -296,9 +296,7 @@ class RandomDataGenerator {
                                            bool float_label = true,
                                            size_t classes = 1) const;
 #if defined(XGBOOST_USE_CUDA)
-  std::shared_ptr<DMatrix> GenerateDeviceDMatrix(bool with_label = false,
+  std::shared_ptr<DMatrix> GenerateDeviceDMatrix();
-                                                 bool float_label = true,
-                                                 size_t classes = 1);
 #endif
 };
 

tests/cpp/metric/test_auc.cc

@@ -22,10 +22,10 @@ TEST(Metric, DeclareUnifiedTest(BinaryAUC)) {
   // Invalid dataset
   MetaInfo info;
   info.labels = linalg::Tensor<float, 2>{{0.0f, 0.0f}, {2}, -1};
-  float auc = metric->Eval({1, 1}, info, false);
+  float auc = metric->Eval({1, 1}, info);
   ASSERT_TRUE(std::isnan(auc));
   *info.labels.Data() = HostDeviceVector<float>{};
-  auc = metric->Eval(HostDeviceVector<float>{}, info, false);
+  auc = metric->Eval(HostDeviceVector<float>{}, info);
   ASSERT_TRUE(std::isnan(auc));
 
   EXPECT_NEAR(GetMetricEval(metric, {0, 1, 0, 1}, {0, 1, 0, 1}), 1.0f, 1e-10);

tests/cpp/metric/test_elementwise_metric.cc

@@ -36,9 +36,9 @@ inline void CheckDeterministicMetricElementWise(StringView name, int32_t device)
     h_labels[i] = dist(&lcg);
   }
 
-  auto result = metric->Eval(predts, info, false);
+  auto result = metric->Eval(predts, info);
   for (size_t i = 0; i < 8; ++i) {
-    ASSERT_EQ(metric->Eval(predts, info, false), result);
+    ASSERT_EQ(metric->Eval(predts, info), result);
   }
 }
 }  // anonymous namespace

tests/cpp/metric/test_multiclass_metric.cc

@@ -35,9 +35,9 @@ inline void CheckDeterministicMetricMultiClass(StringView name, int32_t device)
     }
   }
 
-  auto result = metric->Eval(predts, info, false);
+  auto result = metric->Eval(predts, info);
   for (size_t i = 0; i < 8; ++i) {
-    ASSERT_EQ(metric->Eval(predts, info, false), result);
+    ASSERT_EQ(metric->Eval(predts, info), result);
   }
 }
 }  // namespace xgboost

tests/cpp/metric/test_survival_metric.cu

@@ -40,9 +40,9 @@ inline void CheckDeterministicMetricElementWise(StringView name, int32_t device)
     h_upper[i] = 10;
   }
 
-  auto result = metric->Eval(predts, info, false);
+  auto result = metric->Eval(predts, info);
   for (size_t i = 0; i < 8; ++i) {
-    ASSERT_EQ(metric->Eval(predts, info, false), result);
+    ASSERT_EQ(metric->Eval(predts, info), result);
   }
 }
 }  // anonymous namespace
@@ -72,7 +72,7 @@ TEST(Metric, DeclareUnifiedTest(AFTNegLogLik)) {
     std::unique_ptr<Metric> metric(Metric::Create("aft-nloglik", &lparam));
     metric->Configure({ {"aft_loss_distribution", test_case.dist_type},
                         {"aft_loss_distribution_scale", "1.0"} });
-    EXPECT_NEAR(metric->Eval(preds, info, false), test_case.reference_value, 1e-4);
+    EXPECT_NEAR(metric->Eval(preds, info), test_case.reference_value, 1e-4);
   }
 }
 
@@ -87,15 +87,15 @@ TEST(Metric, DeclareUnifiedTest(IntervalRegressionAccuracy)) {
   HostDeviceVector<bst_float> preds(4, std::log(60.0f));
 
   std::unique_ptr<Metric> metric(Metric::Create("interval-regression-accuracy", &lparam));
-  EXPECT_FLOAT_EQ(metric->Eval(preds, info, false), 0.75f);
+  EXPECT_FLOAT_EQ(metric->Eval(preds, info), 0.75f);
   info.labels_lower_bound_.HostVector()[2] = 70.0f;
-  EXPECT_FLOAT_EQ(metric->Eval(preds, info, false), 0.50f);
+  EXPECT_FLOAT_EQ(metric->Eval(preds, info), 0.50f);
   info.labels_upper_bound_.HostVector()[2] = std::numeric_limits<bst_float>::infinity();
-  EXPECT_FLOAT_EQ(metric->Eval(preds, info, false), 0.50f);
+  EXPECT_FLOAT_EQ(metric->Eval(preds, info), 0.50f);
   info.labels_upper_bound_.HostVector()[3] = std::numeric_limits<bst_float>::infinity();
-  EXPECT_FLOAT_EQ(metric->Eval(preds, info, false), 0.50f);
+  EXPECT_FLOAT_EQ(metric->Eval(preds, info), 0.50f);
   info.labels_lower_bound_.HostVector()[0] = 70.0f;
-  EXPECT_FLOAT_EQ(metric->Eval(preds, info, false), 0.25f);
+  EXPECT_FLOAT_EQ(metric->Eval(preds, info), 0.25f);
 
   CheckDeterministicMetricElementWise(StringView{"interval-regression-accuracy"}, GPUIDX);
 }

tests/cpp/objective/test_ranking_obj_gpu.cu

@@ -170,7 +170,7 @@ TEST(Objective, NDCGLambdaWeightComputerTest) {
   EXPECT_EQ(hgroup_dcgs.size(), segment_label_sorter->GetNumGroups());
   std::vector<float> hsorted_labels(segment_label_sorter->GetNumItems());
   dh::CopyDeviceSpanToVector(&hsorted_labels, segment_label_sorter->GetItemsSpan());
-  for (auto i = 0; i < hgroup_dcgs.size(); ++i) {
+  for (size_t i = 0; i < hgroup_dcgs.size(); ++i) {
     // Compute group DCG value on CPU and compare
     auto gbegin = hgroups[i];
     auto gend = hgroups[i + 1];
@@ -244,7 +244,7 @@ TEST(Objective, ComputeAndCompareMAPStatsTest) {
   std::vector<uint32_t> hgroups(segment_label_sorter->GetNumGroups() + 1);
   dh::CopyDeviceSpanToVector(&hgroups, segment_label_sorter->GetGroupsSpan());
 
-  for (auto i = 0; i < hgroups.size() - 1; ++i) {
+  for (size_t i = 0; i < hgroups.size() - 1; ++i) {
     auto gbegin = hgroups[i];
     auto gend = hgroups[i + 1];
     std::vector<xgboost::obj::ListEntry> lst_entry;

tests/cpp/predictor/test_gpu_predictor.cu

@@ -66,10 +66,7 @@ TEST(GPUPredictor, EllpackBasic) {
   size_t constexpr kCols {8};
   for (size_t bins = 2; bins < 258; bins += 16) {
     size_t rows = bins * 16;
-    auto p_m = RandomDataGenerator{rows, kCols, 0.0}
+    auto p_m = RandomDataGenerator{rows, kCols, 0.0}.Bins(bins).Device(0).GenerateDeviceDMatrix();
-         .Bins(bins)
-         .Device(0)
-         .GenerateDeviceDMatrix(true);
     ASSERT_FALSE(p_m->PageExists<SparsePage>());
     TestPredictionFromGradientIndex<EllpackPage>("gpu_predictor", rows, kCols, p_m);
     TestPredictionFromGradientIndex<EllpackPage>("gpu_predictor", bins, kCols, p_m);
@@ -78,10 +75,8 @@ TEST(GPUPredictor, EllpackBasic) {
 
 TEST(GPUPredictor, EllpackTraining) {
   size_t constexpr kRows { 128 }, kCols { 16 }, kBins { 64 };
-  auto p_ellpack = RandomDataGenerator{kRows, kCols, 0.0}
+  auto p_ellpack =
-       .Bins(kBins)
+      RandomDataGenerator{kRows, kCols, 0.0}.Bins(kBins).Device(0).GenerateDeviceDMatrix();
-       .Device(0)
-       .GenerateDeviceDMatrix(true);
   HostDeviceVector<float> storage(kRows * kCols);
   auto columnar = RandomDataGenerator{kRows, kCols, 0.0}
        .Device(0)

tests/cpp/tree/test_constraints.cu

@@ -94,8 +94,8 @@ TEST(GPUFeatureInteractionConstraint, Init) {
     tree::TrainParam param = GetParameter();
     param.interaction_constraints = R"([[0, 1, 3], [3, 5, 6]])";
     FConstraintWrapper constraints(param, kFeatures);
-    std::vector<int32_t> h_sets {0, 0, 0, 1, 1, 1};
+    std::vector<bst_feature_t> h_sets {0, 0, 0, 1, 1, 1};
-    std::vector<int32_t> h_sets_ptr {0, 1, 2, 2, 4, 4, 5, 6};
+    std::vector<size_t> h_sets_ptr {0, 1, 2, 2, 4, 4, 5, 6};
     auto d_sets = constraints.GetDSets();
     ASSERT_EQ(h_sets.size(), d_sets.size());
     auto d_sets_ptr = constraints.GetDSetsPtr();

tests/cpp/tree/test_gpu_hist.cu

@@ -242,7 +242,7 @@ void TestHistogramIndexImpl() {
   int constexpr kNRows = 1000, kNCols = 10;
 
   // Build 2 matrices and build a histogram maker with that
-  
+
   GenericParameter generic_param(CreateEmptyGenericParam(0));
   tree::GPUHistMaker hist_maker{&generic_param,ObjInfo{ObjInfo::kRegression}},
       hist_maker_ext{&generic_param,ObjInfo{ObjInfo::kRegression}};
@@ -346,7 +346,7 @@ TEST(GpuHist, UniformSampling) {
   // Make sure the predictions are the same.
   auto preds_h = preds.ConstHostVector();
   auto preds_sampling_h = preds_sampling.ConstHostVector();
-  for (int i = 0; i < kRows; i++) {
+  for (size_t i = 0; i < kRows; i++) {
     EXPECT_NEAR(preds_h[i], preds_sampling_h[i], 1e-8);
   }
 }
@@ -376,7 +376,7 @@ TEST(GpuHist, GradientBasedSampling) {
   // Make sure the predictions are the same.
   auto preds_h = preds.ConstHostVector();
   auto preds_sampling_h = preds_sampling.ConstHostVector();
-  for (int i = 0; i < kRows; i++) {
+  for (size_t i = 0; i < kRows; i++) {
     EXPECT_NEAR(preds_h[i], preds_sampling_h[i], 1e-3);
   }
 }
@@ -409,7 +409,7 @@ TEST(GpuHist, ExternalMemory) {
   // Make sure the predictions are the same.
   auto preds_h = preds.ConstHostVector();
   auto preds_ext_h = preds_ext.ConstHostVector();
-  for (int i = 0; i < kRows; i++) {
+  for (size_t i = 0; i < kRows; i++) {
     EXPECT_NEAR(preds_h[i], preds_ext_h[i], 1e-6);
   }
 }
@@ -451,7 +451,7 @@ TEST(GpuHist, ExternalMemoryWithSampling) {
   // Make sure the predictions are the same.
   auto preds_h = preds.ConstHostVector();
   auto preds_ext_h = preds_ext.ConstHostVector();
-  for (int i = 0; i < kRows; i++) {
+  for (size_t i = 0; i < kRows; i++) {
     ASSERT_NEAR(preds_h[i], preds_ext_h[i], 1e-3);
   }
 }