[feature](profile) sort pipelineX task by total time #34053

be/src/runtime/runtime_state.h

@@ -569,7 +569,18 @@ public:
 
     void resize_op_id_to_local_state(int operator_size);
 
-    auto& pipeline_id_to_profile() { return _pipeline_id_to_profile; }
+    auto& pipeline_id_to_profile() {
+        for (auto& pipeline_profile : _pipeline_id_to_profile) {
+            // pipeline 0
+            //  pipeline task 0
+            //  pipeline task 1
+            //  pipleine task 2
+            //  .......
+            // sort by pipeline task total time
+            pipeline_profile->sort_children_by_total_time();
+        }
+        return _pipeline_id_to_profile;
+    }
 
     void set_task_execution_context(std::shared_ptr<TaskExecutionContext> context) {
         _task_execution_context_inited = true;

be/src/util/runtime_profile.cpp

@@ -25,6 +25,7 @@
 #include <rapidjson/stringbuffer.h>
 #include <rapidjson/writer.h>
 
+#include <algorithm>
 #include <iomanip>
 #include <iostream>
 
@@ -716,4 +717,15 @@ void RuntimeProfile::print_child_counters(const std::string& prefix,
     }
 }
 
+void RuntimeProfile::sort_children_by_total_time() {
+    std::lock_guard<std::mutex> l(_children_lock);
+    auto cmp = [](const std::pair<RuntimeProfile*, bool>& L,
+                  const std::pair<RuntimeProfile*, bool>& R) {
+        const RuntimeProfile* L_profile = L.first;
+        const RuntimeProfile* R_profile = R.first;
+        return L_profile->_counter_total_time.value() > R_profile->_counter_total_time.value();
+    };
+    std::sort(_children.begin(), _children.end(), cmp);
+}
+
 } // namespace doris

be/src/util/runtime_profile.h

@@ -301,6 +301,8 @@ public:
         std::sort(_children.begin(), _children.end(), cmp);
     }
 
+    void sort_children_by_total_time();
+
     // Merges the src profile into this one, combining counters that have an identical
     // path. Info strings from profiles are not merged. 'src' would be a const if it
     // weren't for locking.