Thread local memory allocation for BuildHist (#6358)

* thread mem locality

* fix apply

* cleanup

* fix lint

* fix tests

* simple try

* fix

* fix

* apply comments

* fix comments

* fix

* apply simple comment

Co-authored-by: ShvetsKS <kirill.shvets@intel.com>

tests/cpp/common/test_hist_util.cc

@@ -35,7 +35,7 @@ void ParallelGHistBuilderReset() {
   for(size_t inode = 0; inode < kNodesExtended; inode++) {
     collection.AddHistRow(inode);
   }
-
+  collection.AllocateAllData();
   ParallelGHistBuilder<GradientSumT> hist_builder;
   hist_builder.Init(kBins);
   std::vector<GHistRow<GradientSumT>> target_hist(kNodes);
@@ -91,7 +91,7 @@ void ParallelGHistBuilderReduceHist(){
   for(size_t inode = 0; inode < kNodes; inode++) {
     collection.AddHistRow(inode);
   }
-
+  collection.AllocateAllData();
   ParallelGHistBuilder<GradientSumT> hist_builder;
   hist_builder.Init(kBins);
   std::vector<GHistRow<GradientSumT>> target_hist(kNodes);

tests/cpp/common/test_partition_builder.cc

@@ -32,6 +32,8 @@ TEST(PartitionBuilder, BasicTest) {
     for(size_t j = 0; j < tasks[nid]; ++j) {
       size_t begin = kBlockSize*j;
       size_t end = kBlockSize*(j+1);
+      const size_t id = builder.GetTaskIdx(nid, begin);
+      builder.AllocateForTask(id);
 
       auto left  = builder.GetLeftBuffer(nid, begin, end);
       auto right = builder.GetRightBuffer(nid, begin, end);