Optimization/buildhist/hist util (#8218)

* BuildHistKernel optimization

Co-authored-by: dmitry.razdoburdin <drazdobu@jfldaal005.jf.intel.com>

src/common/hist_util.cc

@@ -192,11 +192,14 @@ void BuildHistKernel(const std::vector<GradientPair> &gpair,
     }
     const BinIdxType *gr_index_local = gradient_index + icol_start;
 
+    // The trick with pgh_t buffer helps the compiler to generate faster binary.
+    const float pgh_t[] = {pgh[idx_gh], pgh[idx_gh + 1]};
     for (size_t j = 0; j < row_size; ++j) {
       const uint32_t idx_bin = two * (static_cast<uint32_t>(gr_index_local[j]) +
                                       (any_missing ? 0 : offsets[j]));
-      hist_data[idx_bin] += pgh[idx_gh];
-      hist_data[idx_bin + 1] += pgh[idx_gh + 1];
+      auto hist_local = hist_data + idx_bin;
+      *(hist_local)     += pgh_t[0];
+      *(hist_local + 1) += pgh_t[1];
     }
   }
 }