Support gamma in GPU_Hist. (#4874)

* Just prevent building the tree instead of using an explicit pruner.

src/tree/updater_gpu_hist.cu

@@ -72,8 +72,9 @@ struct ExpandEntry {
     if (split.left_sum.GetHess() == 0 || split.right_sum.GetHess() == 0) {
       return false;
     }
-    if (param.max_depth > 0 && depth == param.max_depth) return false;
+    if (split.loss_chg < param.min_split_loss) { return false; }
-    if (param.max_leaves > 0 && num_leaves == param.max_leaves) return false;
+    if (param.max_depth > 0 && depth == param.max_depth) {return false; }
+    if (param.max_leaves > 0 && num_leaves == param.max_leaves) { return false; }
     return true;
   }
 

tests/cpp/tree/test_gpu_hist.cu

@@ -133,7 +133,7 @@ void TestBuildHist(bool use_shared_memory_histograms) {
   auto page = BuildEllpackPage(kNRows, kNCols);
   DeviceShard<GradientSumT> shard(0, page.get(), kNRows, param, kNCols, kNCols);
   shard.InitHistogram();
-  
+
   xgboost::SimpleLCG gen;
   xgboost::SimpleRealUniformDistribution<bst_float> dist(0.0f, 1.0f);
   std::vector<GradientPair> h_gpair(kNRows);
@@ -336,5 +336,66 @@ TEST(GpuHist, TestHistogramIndex) {
   TestHistogramIndexImpl();
 }
 
+// gamma is an alias of min_split_loss
+int32_t TestMinSplitLoss(DMatrix* dmat, float gamma, HostDeviceVector<GradientPair>* gpair) {
+  Args args {
+    {"max_depth", "1"},
+    {"max_leaves", "0"},
+
+    // Disable all other parameters.
+    {"colsample_bynode", "1"},
+    {"colsample_bylevel", "1"},
+    {"colsample_bytree", "1"},
+    {"min_child_weight", "0.01"},
+    {"reg_alpha", "0"},
+    {"reg_lambda", "0"},
+    {"max_delta_step", "0"},
+
+    // test gamma
+    {"gamma", std::to_string(gamma)}
+  };
+
+  tree::GPUHistMakerSpecialised<GradientPairPrecise> hist_maker;
+  GenericParameter generic_param(CreateEmptyGenericParam(0));
+  hist_maker.Configure(args, &generic_param);
+
+  RegTree tree;
+  hist_maker.Update(gpair, dmat, {&tree});
+
+  auto n_nodes = tree.NumExtraNodes();
+  return n_nodes;
+}
+
+TEST(GpuHist, MinSplitLoss) {
+  constexpr size_t kRows = 32;
+  constexpr size_t kCols = 16;
+  constexpr float kSparsity = 0.6;
+  auto dmat = CreateDMatrix(kRows, kCols, kSparsity, 3);
+
+  xgboost::SimpleLCG gen;
+  xgboost::SimpleRealUniformDistribution<bst_float> dist(0.0f, 1.0f);
+  std::vector<GradientPair> h_gpair(kRows);
+  for (auto &gpair : h_gpair) {
+    bst_float grad = dist(&gen);
+    bst_float hess = dist(&gen);
+    gpair = GradientPair(grad, hess);
+  }
+  HostDeviceVector<GradientPair> gpair(h_gpair);
+
+  {
+    int32_t n_nodes = TestMinSplitLoss((*dmat).get(), 0.01, &gpair);
+    // This is not strictly verified, meaning the numeber `2` is whatever GPU_Hist retured
+    // when writing this test, and only used for testing larger gamma (below) does prevent
+    // building tree.
+    ASSERT_EQ(n_nodes, 2);
+  }
+  {
+    int32_t n_nodes = TestMinSplitLoss((*dmat).get(), 100.0, &gpair);
+    // No new nodes with gamma == 100.
+    ASSERT_EQ(n_nodes, static_cast<decltype(n_nodes)>(0));
+  }
+  delete dmat;
+}
+
 }  // namespace tree
 }  // namespace xgboost