[SYCL] Add nodes initialisation (dmlc#10269)

razdoburdin · trivialfis · web-flow · commit c7e7ce75693f · 2024-05-21T23:38:52.000+08:00
---------

Co-authored-by: Dmitry Razdoburdin &lt;&gt;
Co-authored-by: Jiaming Yuan &lt;jm.yuan@outlook.com&gt;
diff --git a/plugin/sycl/device_manager.cc b/plugin/sycl/device_manager.cc
@@ -2,11 +2,6 @@
  * Copyright 2017-2023 by Contributors
  * \file device_manager.cc
  */
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wtautological-constant-compare"
-#pragma GCC diagnostic ignored "-W#pragma-messages"
-#pragma GCC diagnostic pop
-
 #include "../sycl/device_manager.h"
 
 #include "../../src/collective/communicator-inl.h"
diff --git a/plugin/sycl/device_manager.h b/plugin/sycl/device_manager.h
@@ -12,7 +12,11 @@
 
 #include <CL/sycl.hpp>
 
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wtautological-constant-compare"
+#pragma GCC diagnostic ignored "-W#pragma-messages"
 #include "xgboost/context.h"
+#pragma GCC diagnostic pop
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/objective/multiclass_obj.cc b/plugin/sycl/objective/multiclass_obj.cc
@@ -3,19 +3,15 @@
  * \file multiclass_obj.cc
  * \brief Definition of multi-class classification objectives.
  */
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wtautological-constant-compare"
-#pragma GCC diagnostic ignored "-W#pragma-messages"
-#pragma GCC diagnostic pop
-
 #include <vector>
 #include <algorithm>
 #include <limits>
 #include <utility>
 
-#include "xgboost/parameter.h"
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wtautological-constant-compare"
+#pragma GCC diagnostic ignored "-W#pragma-messages"
+#include "xgboost/parameter.h"
 #include "xgboost/data.h"
 #include "../../src/common/math.h"
 #pragma GCC diagnostic pop
diff --git a/plugin/sycl/tree/hist_updater.cc b/plugin/sycl/tree/hist_updater.cc
@@ -8,6 +8,7 @@
 #include <oneapi/dpl/random>
 
 #include "../common/hist_util.h"
+#include "../../src/collective/allreduce.h"
 
 namespace xgboost {
 namespace sycl {
@@ -111,7 +112,6 @@ void HistUpdater<GradientSumT>::InitSampling(
 
 template<typename GradientSumT>
 void HistUpdater<GradientSumT>::InitData(
-                                Context const * ctx,
                                 const common::GHistIndexMatrix& gmat,
                                 const USMVector<GradientPair, MemoryType::on_device> &gpair,
                                 const DMatrix& fmat,
@@ -215,6 +215,101 @@ void HistUpdater<GradientSumT>::InitData(
       data_layout_ = kSparseData;
     }
   }
+
+  if (data_layout_ == kDenseDataZeroBased || data_layout_ == kDenseDataOneBased) {
+    /* specialized code for dense data:
+       choose the column that has a least positive number of discrete bins.
+       For dense data (with no missing value),
+       the sum of gradient histogram is equal to snode[nid] */
+    const std::vector<uint32_t>& row_ptr = gmat.cut.Ptrs();
+    const auto nfeature = static_cast<bst_uint>(row_ptr.size() - 1);
+    uint32_t min_nbins_per_feature = 0;
+    for (bst_uint i = 0; i < nfeature; ++i) {
+      const uint32_t nbins = row_ptr[i + 1] - row_ptr[i];
+      if (nbins > 0) {
+        if (min_nbins_per_feature == 0 || min_nbins_per_feature > nbins) {
+          min_nbins_per_feature = nbins;
+          fid_least_bins_ = i;
+        }
+      }
+    }
+    CHECK_GT(min_nbins_per_feature, 0U);
+  }
+
+  std::fill(snode_host_.begin(), snode_host_.end(),  NodeEntry<GradientSumT>(param_));
+  builder_monitor_.Stop("InitData");
+}
+
+template <typename GradientSumT>
+void HistUpdater<GradientSumT>::InitNewNode(int nid,
+                                            const common::GHistIndexMatrix& gmat,
+                                            const USMVector<GradientPair,
+                                                            MemoryType::on_device> &gpair,
+                                            const DMatrix& fmat,
+                                            const RegTree& tree) {
+  builder_monitor_.Start("InitNewNode");
+
+  snode_host_.resize(tree.NumNodes(), NodeEntry<GradientSumT>(param_));
+  {
+    if (tree[nid].IsRoot()) {
+      GradStats<GradientSumT> grad_stat;
+      if (data_layout_ == kDenseDataZeroBased || data_layout_ == kDenseDataOneBased) {
+        const std::vector<uint32_t>& row_ptr = gmat.cut.Ptrs();
+        const uint32_t ibegin = row_ptr[fid_least_bins_];
+        const uint32_t iend = row_ptr[fid_least_bins_ + 1];
+        const auto* hist = reinterpret_cast<GradStats<GradientSumT>*>(hist_[nid].Data());
+
+        std::vector<GradStats<GradientSumT>> ets(iend - ibegin);
+        qu_.memcpy(ets.data(), hist + ibegin,
+                   (iend - ibegin) * sizeof(GradStats<GradientSumT>)).wait_and_throw();
+        for (const auto& et : ets) {
+          grad_stat += et;
+        }
+      } else {
+        const common::RowSetCollection::Elem e = row_set_collection_[nid];
+        const size_t* row_idxs = e.begin;
+        const size_t size = e.Size();
+        const GradientPair* gpair_ptr = gpair.DataConst();
+
+        ::sycl::buffer<GradStats<GradientSumT>> buff(&grad_stat, 1);
+        qu_.submit([&](::sycl::handler& cgh) {
+          auto reduction = ::sycl::reduction(buff, cgh, ::sycl::plus<>());
+          cgh.parallel_for<>(::sycl::range<1>(size), reduction,
+                            [=](::sycl::item<1> pid, auto& sum) {
+            size_t i = pid.get_id(0);
+            size_t row_idx = row_idxs[i];
+            if constexpr (std::is_same<GradientPair::ValueT, GradientSumT>::value) {
+              sum += gpair_ptr[row_idx];
+            } else {
+              sum += GradStats<GradientSumT>(gpair_ptr[row_idx].GetGrad(),
+                                             gpair_ptr[row_idx].GetHess());
+            }
+          });
+        }).wait_and_throw();
+      }
+      auto rc = collective::Allreduce(
+                      ctx_, linalg::MakeVec(reinterpret_cast<GradientSumT*>(&grad_stat), 2),
+                      collective::Op::kSum);
+      SafeColl(rc);
+      snode_host_[nid].stats = grad_stat;
+    } else {
+      int parent_id = tree[nid].Parent();
+      if (tree[nid].IsLeftChild()) {
+        snode_host_[nid].stats = snode_host_[parent_id].best.left_sum;
+      } else {
+        snode_host_[nid].stats = snode_host_[parent_id].best.right_sum;
+      }
+    }
+  }
+
+  // calculating the weights
+  {
+    auto evaluator = tree_evaluator_.GetEvaluator();
+    bst_uint parentid = tree[nid].Parent();
+    snode_host_[nid].weight = evaluator.CalcWeight(parentid, snode_host_[nid].stats);
+    snode_host_[nid].root_gain = evaluator.CalcGain(parentid, snode_host_[nid].stats);
+  }
+  builder_monitor_.Stop("InitNewNode");
 }
 
 template class HistUpdater<float>;
diff --git a/plugin/sycl/tree/hist_updater.h b/plugin/sycl/tree/hist_updater.h
@@ -26,19 +26,36 @@ namespace xgboost {
 namespace sycl {
 namespace tree {
 
+// data structure
+template<typename GradType>
+struct NodeEntry {
+  /*! \brief statics for node entry */
+  GradStats<GradType> stats;
+  /*! \brief loss of this node, without split */
+  GradType root_gain;
+  /*! \brief weight calculated related to current data */
+  GradType weight;
+  /*! \brief current best solution */
+  SplitEntry<GradType> best;
+  // constructor
+  explicit NodeEntry(const xgboost::tree::TrainParam& param)
+      : root_gain(0.0f), weight(0.0f) {}
+};
+
 template<typename GradientSumT>
 class HistUpdater {
  public:
   template <MemoryType memory_type = MemoryType::shared>
   using GHistRowT = common::GHistRow<GradientSumT, memory_type>;
   using GradientPairT = xgboost::detail::GradientPairInternal<GradientSumT>;
 
-  explicit HistUpdater(::sycl::queue qu,
-                    const xgboost::tree::TrainParam& param,
-                    std::unique_ptr<TreeUpdater> pruner,
-                    FeatureInteractionConstraintHost int_constraints_,
-                    DMatrix const* fmat)
-    : qu_(qu), param_(param),
+  explicit HistUpdater(const Context* ctx,
+                       ::sycl::queue qu,
+                       const xgboost::tree::TrainParam& param,
+                       std::unique_ptr<TreeUpdater> pruner,
+                       FeatureInteractionConstraintHost int_constraints_,
+                       DMatrix const* fmat)
+    : ctx_(ctx), qu_(qu), param_(param),
       tree_evaluator_(qu, param, fmat->Info().num_col_),
       pruner_(std::move(pruner)),
       interaction_constraints_{std::move(int_constraints_)},
@@ -61,8 +78,7 @@ class HistUpdater {
                     USMVector<size_t, MemoryType::on_device>* row_indices);
 
 
-  void InitData(Context const * ctx,
-                const common::GHistIndexMatrix& gmat,
+  void InitData(const common::GHistIndexMatrix& gmat,
                 const USMVector<GradientPair, MemoryType::on_device> &gpair,
                 const DMatrix& fmat,
                 const RegTree& tree);
@@ -78,6 +94,12 @@ class HistUpdater {
                                    data_layout_ != kSparseData, hist_buffer, event_priv);
   }
 
+  void InitNewNode(int nid,
+                   const common::GHistIndexMatrix& gmat,
+                   const USMVector<GradientPair, MemoryType::on_device> &gpair,
+                   const DMatrix& fmat,
+                   const RegTree& tree);
+
   void BuildLocalHistograms(const common::GHistIndexMatrix &gmat,
                             RegTree *p_tree,
                             const USMVector<GradientPair, MemoryType::on_device> &gpair);
@@ -89,6 +111,7 @@ class HistUpdater {
                       const USMVector<GradientPair, MemoryType::on_device> &gpair);
 
   //  --data fields--
+  const Context* ctx_;
   size_t sub_group_size_;
 
   // the internal row sets
@@ -113,9 +136,16 @@ class HistUpdater {
   /*! \brief culmulative histogram of gradients. */
   common::HistCollection<GradientSumT, MemoryType::on_device> hist_;
 
+  /*! \brief TreeNode Data: statistics for each constructed node */
+  std::vector<NodeEntry<GradientSumT>> snode_host_;
+
   xgboost::common::Monitor builder_monitor_;
   xgboost::common::Monitor kernel_monitor_;
 
+  /*! \brief feature with least # of bins. to be used for dense specialization
+              of InitNewNode() */
+  uint32_t fid_least_bins_;
+
   uint64_t seed_ = 0;
 
   // key is the node id which should be calculated by Subtraction Trick, value is the node which
diff --git a/plugin/sycl/tree/param.h b/plugin/sycl/tree/param.h
@@ -49,6 +49,115 @@ struct TrainParam {
 template <typename GradType>
 using GradStats = xgboost::detail::GradientPairInternal<GradType>;
 
+/*!
+ * \brief SYCL implementation of SplitEntryContainer for device compilation.
+ *        Original structure cannot be used due 'cat_bits' field of type std::vector<uint32_t>,
+ *        which is not device-copyable
+ */
+template<typename GradientT>
+struct SplitEntryContainer {
+  /*! \brief loss change after split this node */
+  bst_float loss_chg {0.0f};
+  /*! \brief split index */
+  bst_feature_t sindex{0};
+  bst_float split_value{0.0f};
+
+
+  GradientT left_sum;
+  GradientT right_sum;
+
+
+  SplitEntryContainer() = default;
+
+
+  friend std::ostream& operator<<(std::ostream& os, SplitEntryContainer const& s) {
+    os << "loss_chg: " << s.loss_chg << ", "
+       << "split index: " << s.SplitIndex() << ", "
+       << "split value: " << s.split_value << ", "
+       << "left_sum: " << s.left_sum << ", "
+       << "right_sum: " << s.right_sum;
+    return os;
+  }
+  /*!\return feature index to split on */
+  bst_feature_t SplitIndex() const { return sindex & ((1U << 31) - 1U); }
+  /*!\return whether missing value goes to left branch */
+  bool DefaultLeft() const { return (sindex >> 31) != 0; }
+  /*!
+   * \brief decides whether we can replace current entry with the given statistics
+   *
+   *   This function gives better priority to lower index when loss_chg == new_loss_chg.
+   *   Not the best way, but helps to give consistent result during multi-thread
+   *   execution.
+   *
+   * \param new_loss_chg the loss reduction get through the split
+   * \param split_index the feature index where the split is on
+   */
+  inline bool NeedReplace(bst_float new_loss_chg, unsigned split_index) const {
+    if (::sycl::isinf(new_loss_chg)) {  // in some cases new_loss_chg can be NaN or Inf,
+                                      // for example when lambda = 0 & min_child_weight = 0
+                                      // skip value in this case
+      return false;
+    } else if (this->SplitIndex() <= split_index) {
+      return new_loss_chg > this->loss_chg;
+    } else {
+      return !(this->loss_chg > new_loss_chg);
+    }
+  }
+  /*!
+   * \brief update the split entry, replace it if e is better
+   * \param e candidate split solution
+   * \return whether the proposed split is better and can replace current split
+   */
+  inline bool Update(const SplitEntryContainer &e) {
+    if (this->NeedReplace(e.loss_chg, e.SplitIndex())) {
+      this->loss_chg = e.loss_chg;
+      this->sindex = e.sindex;
+      this->split_value = e.split_value;
+      this->left_sum = e.left_sum;
+      this->right_sum = e.right_sum;
+      return true;
+    } else {
+      return false;
+    }
+  }
+  /*!
+   * \brief update the split entry, replace it if e is better
+   * \param new_loss_chg loss reduction of new candidate
+   * \param split_index feature index to split on
+   * \param new_split_value the split point
+   * \param default_left whether the missing value goes to left
+   * \return whether the proposed split is better and can replace current split
+   */
+  bool Update(bst_float new_loss_chg, unsigned split_index,
+              bst_float new_split_value, bool default_left,
+              const GradientT &left_sum,
+              const GradientT &right_sum) {
+    if (this->NeedReplace(new_loss_chg, split_index)) {
+      this->loss_chg = new_loss_chg;
+      if (default_left) {
+        split_index |= (1U << 31);
+      }
+      this->sindex = split_index;
+      this->split_value = new_split_value;
+      this->left_sum = left_sum;
+      this->right_sum = right_sum;
+      return true;
+    } else {
+      return false;
+    }
+  }
+
+
+  /*! \brief same as update, used by AllReduce*/
+  inline static void Reduce(SplitEntryContainer &dst,         // NOLINT(*)
+                            const SplitEntryContainer &src) { // NOLINT(*)
+    dst.Update(src);
+  }
+};
+
+template<typename GradType>
+using SplitEntry = SplitEntryContainer<GradStats<GradType>>;
+
 }  // namespace tree
 }  // namespace sycl
 }  // namespace xgboost
diff --git a/tests/cpp/plugin/test_sycl_hist_updater.cc b/tests/cpp/plugin/test_sycl_hist_updater.cc
