add status report

src/allreduce_base.h

@@ -144,6 +144,16 @@ class AllreduceBase : public IEngine {
   virtual void InitAfterException(void) {
     utils::Error("InitAfterException: not implemented");
   }
+  /*! 
+   * \brief report current status to the job tracker 
+   * depending on the job tracker we are in
+   */
+  inline void ReportStatus(void) const {
+    if (hadoop_mode != 0) {
+      fprintf(stderr, "reporter:status:Rabit Phase[%03d] Operation %03d\n",
+              version_number, seq_counter);
+    }
+  }
 
  protected:
   /*! \brief enumeration of possible returning results from Try functions */
@@ -284,6 +294,10 @@ class AllreduceBase : public IEngine {
    */
   ReturnType TryBroadcast(void *sendrecvbuf_, size_t size, int root);
   //---- data structure related to model ----
+  // call sequence counter, records how many calls we made so far
+  // from last call to CheckPoint, LoadCheckPoint
+  int seq_counter;
+  // version number of model
   int version_number;
   // whether the job is running in hadoop
   int hadoop_mode;

src/allreduce_robust.cc

@@ -645,6 +645,7 @@ bool AllreduceRobust::RecoverExec(void *buf, size_t size, int flag, int seqno) {
   // request
   ActionSummary req(flag, seqno);
   while (true) {
+    this->ReportStatus();
     // action
     ActionSummary act = req;    
     // get the reduced action

src/allreduce_robust.h

@@ -402,9 +402,6 @@ class AllreduceRobust : public AllreduceBase {
                                                  size_t out_index)
                                );
   //---- recovery data structure ----
-  // call sequence counter, records how many calls we made so far
-  // from last call to CheckPoint, LoadCheckPoint
-  int seq_counter;
   // the round of result buffer, used to mode the result
   int result_buffer_round;
   // result buffer of all reduce