[backport] Fix column split race condition. (dmlc#10572)

src/tree/common_row_partitioner.h

@@ -36,10 +36,11 @@ class ColumnSplitHelper {
                     common::PartitionBuilder<kPartitionBlockSize>* partition_builder,
                     common::RowSetCollection* row_set_collection)
       : partition_builder_{partition_builder}, row_set_collection_{row_set_collection} {
-    decision_storage_.resize(num_row);
-    decision_bits_ = BitVector(common::Span<BitVector::value_type>(decision_storage_));
-    missing_storage_.resize(num_row);
-    missing_bits_ = BitVector(common::Span<BitVector::value_type>(missing_storage_));
+    auto n_bytes = BitVector::ComputeStorageSize(num_row);
+    decision_storage_.resize(n_bytes);
+    decision_bits_ = BitVector{common::Span<BitVector::value_type>{decision_storage_}};
+    missing_storage_.resize(n_bytes);
+    missing_bits_ = BitVector{common::Span<BitVector::value_type>{missing_storage_}};
   }
 
   template <typename BinIdxType, bool any_missing, bool any_cat, typename ExpandEntry>
@@ -51,14 +52,43 @@ class ColumnSplitHelper {
     // we first collect all the decisions and whether the feature is missing into bit vectors.
     std::fill(decision_storage_.begin(), decision_storage_.end(), 0);
     std::fill(missing_storage_.begin(), missing_storage_.end(), 0);
-    common::ParallelFor2d(space, n_threads, [&](size_t node_in_set, common::Range1d r) {
-      const int32_t nid = nodes[node_in_set].nid;
+
+    this->tloc_decision_.resize(decision_storage_.size() * n_threads);
+    this->tloc_missing_.resize(decision_storage_.size() * n_threads);
+    std::fill_n(this->tloc_decision_.data(), this->tloc_decision_.size(), 0);
+    std::fill_n(this->tloc_missing_.data(), this->tloc_missing_.size(), 0);
+
+    // Make thread-local storage.
+    using T = decltype(decision_storage_)::value_type;
+    auto make_tloc = [&](std::vector<T>& storage, std::int32_t tidx) {
+      auto span = common::Span<T>{storage};
+      auto n = decision_storage_.size();
+      auto bitvec = BitVector{span.subspan(n * tidx, n)};
+      return bitvec;
+    };
+
+    common::ParallelFor2d(space, n_threads, [&](std::size_t node_in_set, common::Range1d r) {
+      bst_node_t const nid = nodes[node_in_set].nid;
+      auto tidx = omp_get_thread_num();
+      auto decision = make_tloc(this->tloc_decision_, tidx);
+      auto missing = make_tloc(this->tloc_missing_, tidx);
       bst_bin_t split_cond = column_matrix.IsInitialized() ? split_conditions[node_in_set] : 0;
       partition_builder_->MaskRows<BinIdxType, any_missing, any_cat>(
           node_in_set, nodes, r, split_cond, gmat, column_matrix, *p_tree,
-          (*row_set_collection_)[nid].begin, &decision_bits_, &missing_bits_);
+          (*row_set_collection_)[nid].begin, &decision, &missing);
     });
 
+    // Reduce thread local
+    auto decision = make_tloc(this->tloc_decision_, 0);
+    auto missing = make_tloc(this->tloc_missing_, 0);
+    for (std::int32_t tidx = 1; tidx < n_threads; ++tidx) {
+      decision |= make_tloc(this->tloc_decision_, tidx);
+      missing |= make_tloc(this->tloc_missing_, tidx);
+    }
+    CHECK_EQ(decision_storage_.size(), decision.NumValues());
+    std::copy_n(decision.Data(), decision_storage_.size(), decision_storage_.data());
+    std::copy_n(missing.Data(), missing_storage_.size(), missing_storage_.data());
+
     // Then aggregate the bit vectors across all the workers.
     auto rc = collective::Success() << [&] {
       return collective::Allreduce(ctx, &decision_storage_, collective::Op::kBitwiseOR);
@@ -85,6 +115,10 @@ class ColumnSplitHelper {
   BitVector decision_bits_{};
   std::vector<BitVector::value_type> missing_storage_{};
   BitVector missing_bits_{};
+
+  std::vector<BitVector::value_type> tloc_decision_;
+  std::vector<BitVector::value_type> tloc_missing_;
+
   common::PartitionBuilder<kPartitionBlockSize>* partition_builder_;
   common::RowSetCollection* row_set_collection_;
 };

tests/cpp/tree/test_approx.cc

@@ -7,6 +7,7 @@
 #include "../../../src/tree/common_row_partitioner.h"
 #include "../collective/test_worker.h"  // for TestDistributedGlobal
 #include "../helpers.h"
+#include "test_column_split.h"  // for TestColumnSplit
 #include "test_partitioner.h"
 
 namespace xgboost::tree {
@@ -150,4 +151,26 @@ TEST(Approx, PartitionerColSplit) {
                                mid_partitioner);
   });
 }
+
+namespace {
+class TestApproxColSplit : public ::testing::TestWithParam<std::tuple<bool, float>> {
+ public:
+  void Run() {
+    auto [categorical, sparsity] = GetParam();
+    TestColumnSplit(1u, categorical, "grow_histmaker", sparsity);
+  }
+};
+}  // namespace
+
+TEST_P(TestApproxColSplit, Basic) { this->Run(); }
+
+INSTANTIATE_TEST_SUITE_P(ColumnSplit, TestApproxColSplit, ::testing::ValuesIn([]() {
+                           std::vector<std::tuple<bool, float>> params;
+                           for (auto categorical : {true, false}) {
+                             for (auto sparsity : {0.0f, 0.6f}) {
+                               params.emplace_back(categorical, sparsity);
+                             }
+                           }
+                           return params;
+                         }()));
 }  // namespace xgboost::tree

tests/cpp/tree/test_column_split.h

@@ -0,0 +1,79 @@
+/**
+ * Copyright 2023-2024, XGBoost Contributors
+ */
+#pragma once
+
+#include <xgboost/data.h>          // for FeatureType, DMatrix
+#include <xgboost/tree_model.h>    // for RegTree
+#include <xgboost/tree_updater.h>  // for TreeUpdater
+
+#include <cstddef>  // for size_t
+#include <memory>   // for shared_ptr
+#include <vector>   // for vector
+
+#include "../../../src/tree/param.h"    // for TrainParam
+#include "../collective/test_worker.h"  // for TestDistributedGlobal
+#include "../helpers.h"                 // for RandomDataGenerator
+
+namespace xgboost::tree {
+inline std::shared_ptr<DMatrix> GenerateCatDMatrix(std::size_t rows, std::size_t cols,
+                                                   float sparsity, bool categorical) {
+  if (categorical) {
+    std::vector<FeatureType> ft(cols);
+    for (size_t i = 0; i < ft.size(); ++i) {
+      ft[i] = (i % 3 == 0) ? FeatureType::kNumerical : FeatureType::kCategorical;
+    }
+    return RandomDataGenerator(rows, cols, 0.6f).Seed(3).Type(ft).MaxCategory(17).GenerateDMatrix();
+  } else {
+    return RandomDataGenerator{rows, cols, 0.6f}.Seed(3).GenerateDMatrix();
+  }
+}
+
+inline void TestColumnSplit(bst_target_t n_targets, bool categorical, std::string name,
+                            float sparsity) {
+  auto constexpr kRows = 32;
+  auto constexpr kCols = 16;
+
+  RegTree expected_tree{n_targets, static_cast<bst_feature_t>(kCols)};
+  ObjInfo task{ObjInfo::kRegression};
+  Context ctx;
+  {
+    auto p_dmat = GenerateCatDMatrix(kRows, kCols, sparsity, categorical);
+    auto gpair = GenerateRandomGradients(&ctx, kRows, n_targets);
+    std::unique_ptr<TreeUpdater> updater{TreeUpdater::Create(name, &ctx, &task)};
+    std::vector<HostDeviceVector<bst_node_t>> position(1);
+    TrainParam param;
+    param.Init(Args{});
+    updater->Configure(Args{});
+    updater->Update(&param, &gpair, p_dmat.get(), position, {&expected_tree});
+  }
+
+  auto verify = [&] {
+    Context ctx;
+    auto p_dmat = GenerateCatDMatrix(kRows, kCols, sparsity, categorical);
+    auto gpair = GenerateRandomGradients(&ctx, kRows, n_targets);
+
+    ObjInfo task{ObjInfo::kRegression};
+    std::unique_ptr<TreeUpdater> updater{TreeUpdater::Create(name, &ctx, &task)};
+    std::vector<HostDeviceVector<bst_node_t>> position(1);
+
+    std::unique_ptr<DMatrix> sliced{
+        p_dmat->SliceCol(collective::GetWorldSize(), collective::GetRank())};
+
+    RegTree tree{n_targets, static_cast<bst_feature_t>(kCols)};
+    TrainParam param;
+    param.Init(Args{});
+    updater->Configure(Args{});
+    updater->Update(&param, &gpair, sliced.get(), position, {&tree});
+
+    Json json{Object{}};
+    tree.SaveModel(&json);
+    Json expected_json{Object{}};
+    expected_tree.SaveModel(&expected_json);
+    ASSERT_EQ(json, expected_json);
+  };
+
+  auto constexpr kWorldSize = 2;
+  collective::TestDistributedGlobal(kWorldSize, [&] { verify(); });
+}
+}  // namespace xgboost::tree

tests/cpp/tree/test_histmaker.cc

@@ -1,32 +1,19 @@
 /**
- * Copyright 2019-2023 by XGBoost Contributors
+ * Copyright 2019-2024, XGBoost Contributors
  */
 #include <gtest/gtest.h>
 #include <xgboost/tree_model.h>
 #include <xgboost/tree_updater.h>
 
-#include "../../../src/tree/param.h"    // for TrainParam
-#include "../collective/test_worker.h"  // for TestDistributedGlobal
+#include "../../../src/tree/param.h"  // for TrainParam
 #include "../helpers.h"
+#include "test_column_split.h"  // for GenerateCatDMatrix
 
 namespace xgboost::tree {
-std::shared_ptr<DMatrix> GenerateDMatrix(std::size_t rows, std::size_t cols,
-                                         bool categorical = false) {
-  if (categorical) {
-    std::vector<FeatureType> ft(cols);
-    for (size_t i = 0; i < ft.size(); ++i) {
-      ft[i] = (i % 3 == 0) ? FeatureType::kNumerical : FeatureType::kCategorical;
-    }
-    return RandomDataGenerator(rows, cols, 0.6f).Seed(3).Type(ft).MaxCategory(17).GenerateDMatrix();
-  } else {
-    return RandomDataGenerator{rows, cols, 0.6f}.Seed(3).GenerateDMatrix();
-  }
-}
-
 TEST(GrowHistMaker, InteractionConstraint) {
   auto constexpr kRows = 32;
   auto constexpr kCols = 16;
-  auto p_dmat = GenerateDMatrix(kRows, kCols);
+  auto p_dmat = GenerateCatDMatrix(kRows, kCols, 0.0, false);
   Context ctx;
 
   linalg::Matrix<GradientPair> gpair({kRows}, ctx.Device());
@@ -69,62 +56,4 @@ TEST(GrowHistMaker, InteractionConstraint) {
     ASSERT_NE(tree[tree[0].RightChild()].SplitIndex(), 0);
   }
 }
-
-namespace {
-void VerifyColumnSplit(int32_t rows, bst_feature_t cols, bool categorical,
-                       RegTree const& expected_tree) {
-  Context ctx;
-  auto p_dmat = GenerateDMatrix(rows, cols, categorical);
-  linalg::Matrix<GradientPair> gpair({rows}, ctx.Device());
-  gpair.Data()->Copy(GenerateRandomGradients(rows));
-
-
-  ObjInfo task{ObjInfo::kRegression};
-  std::unique_ptr<TreeUpdater> updater{TreeUpdater::Create("grow_histmaker", &ctx, &task)};
-  std::vector<HostDeviceVector<bst_node_t>> position(1);
-
-  std::unique_ptr<DMatrix> sliced{
-      p_dmat->SliceCol(collective::GetWorldSize(), collective::GetRank())};
-
-  RegTree tree{1u, cols};
-  TrainParam param;
-  param.Init(Args{});
-  updater->Configure(Args{});
-  updater->Update(&param, &gpair, sliced.get(), position, {&tree});
-
-  Json json{Object{}};
-  tree.SaveModel(&json);
-  Json expected_json{Object{}};
-  expected_tree.SaveModel(&expected_json);
-  ASSERT_EQ(json, expected_json);
-}
-
-void TestColumnSplit(bool categorical) {
-  auto constexpr kRows = 32;
-  auto constexpr kCols = 16;
-
-  RegTree expected_tree{1u, kCols};
-  ObjInfo task{ObjInfo::kRegression};
-  {
-    Context ctx;
-    auto p_dmat = GenerateDMatrix(kRows, kCols, categorical);
-    linalg::Matrix<GradientPair> gpair({kRows}, ctx.Device());
-    gpair.Data()->Copy(GenerateRandomGradients(kRows));
-    std::unique_ptr<TreeUpdater> updater{TreeUpdater::Create("grow_histmaker", &ctx, &task)};
-    std::vector<HostDeviceVector<bst_node_t>> position(1);
-    TrainParam param;
-    param.Init(Args{});
-    updater->Configure(Args{});
-    updater->Update(&param, &gpair, p_dmat.get(), position, {&expected_tree});
-  }
-
-  auto constexpr kWorldSize = 2;
-  collective::TestDistributedGlobal(
-      kWorldSize, [&] { VerifyColumnSplit(kRows, kCols, categorical, expected_tree); });
-}
-}  // anonymous namespace
-
-TEST(GrowHistMaker, ColumnSplitNumerical) { TestColumnSplit(false); }
-
-TEST(GrowHistMaker, ColumnSplitCategorical) { TestColumnSplit(true); }
 }  // namespace xgboost::tree

tests/cpp/tree/test_quantile_hist.cc

@@ -12,9 +12,9 @@
 
 #include "../../../src/tree/common_row_partitioner.h"
 #include "../../../src/tree/hist/expand_entry.h"  // for MultiExpandEntry, CPUExpandEntry
-#include "../../../src/tree/param.h"
 #include "../collective/test_worker.h"  // for TestDistributedGlobal
 #include "../helpers.h"
+#include "test_column_split.h"  // for TestColumnSplit
 #include "test_partitioner.h"
 #include "xgboost/data.h"
 
@@ -203,57 +203,26 @@ TEST(QuantileHist, PartitionerColSplit) { TestColumnSplitPartitioner<CPUExpandEn
 TEST(QuantileHist, MultiPartitionerColSplit) { TestColumnSplitPartitioner<MultiExpandEntry>(3); }
 
 namespace {
-void VerifyColumnSplit(Context const* ctx, bst_idx_t rows, bst_feature_t cols, bst_target_t n_targets,
-                       RegTree const& expected_tree) {
-  auto Xy = RandomDataGenerator{rows, cols, 0}.GenerateDMatrix(true);
-  linalg::Matrix<GradientPair> gpair = GenerateRandomGradients(ctx, rows, n_targets);
-
-  ObjInfo task{ObjInfo::kRegression};
-  std::unique_ptr<TreeUpdater> updater{TreeUpdater::Create("grow_quantile_histmaker", ctx, &task)};
-  std::vector<HostDeviceVector<bst_node_t>> position(1);
-
-  std::unique_ptr<DMatrix> sliced{Xy->SliceCol(collective::GetWorldSize(), collective::GetRank())};
-
-  RegTree tree{n_targets, cols};
-  TrainParam param;
-  param.Init(Args{});
-  updater->Configure(Args{});
-  updater->Update(&param, &gpair, sliced.get(), position, {&tree});
-
-  Json json{Object{}};
-  tree.SaveModel(&json);
-  Json expected_json{Object{}};
-  expected_tree.SaveModel(&expected_json);
-  ASSERT_EQ(json, expected_json);
-}
-
-void TestColumnSplit(bst_target_t n_targets) {
-  auto constexpr kRows = 32;
-  auto constexpr kCols = 16;
-
-  RegTree expected_tree{n_targets, kCols};
-  ObjInfo task{ObjInfo::kRegression};
-  Context ctx;
-  {
-    auto Xy = RandomDataGenerator{kRows, kCols, 0}.GenerateDMatrix(true);
-    auto gpair = GenerateRandomGradients(&ctx, kRows, n_targets);
-    std::unique_ptr<TreeUpdater> updater{
-        TreeUpdater::Create("grow_quantile_histmaker", &ctx, &task)};
-    std::vector<HostDeviceVector<bst_node_t>> position(1);
-    TrainParam param;
-    param.Init(Args{});
-    updater->Configure(Args{});
-    updater->Update(&param, &gpair, Xy.get(), position, {&expected_tree});
+class TestHistColSplit : public ::testing::TestWithParam<std::tuple<bst_target_t, bool, float>> {
+ public:
+  void Run() {
+    auto [n_targets, categorical, sparsity] = GetParam();
+    TestColumnSplit(n_targets, categorical, "grow_quantile_histmaker", sparsity);
   }
-
-  auto constexpr kWorldSize = 2;
-  collective::TestDistributedGlobal(kWorldSize, [&] {
-    VerifyColumnSplit(&ctx, kRows, kCols, n_targets, std::cref(expected_tree));
-  });
-}
+};
 }  // anonymous namespace
 
-TEST(QuantileHist, ColumnSplit) { TestColumnSplit(1); }
-
-TEST(QuantileHist, ColumnSplitMultiTarget) { TestColumnSplit(3); }
+TEST_P(TestHistColSplit, Basic) { this->Run(); }
+
+INSTANTIATE_TEST_SUITE_P(ColumnSplit, TestHistColSplit, ::testing::ValuesIn([]() {
+                           std::vector<std::tuple<bst_target_t, bool, float>> params;
+                           for (auto categorical : {true, false}) {
+                             for (auto sparsity : {0.0f, 0.6f}) {
+                               for (bst_target_t n_targets : {1u, 3u}) {
+                                 params.emplace_back(n_targets, categorical, sparsity);
+                               }
+                             }
+                           }
+                           return params;
+                         }()));
 }  // namespace xgboost::tree