[PWGLF] Reduce MC memory usage: reuse vectors per event and shrink every 1000 events (#13342)

Author: alcaliva

PWGLF/Tasks/Nuspex/antinucleiInJets.cxx

@@ -151,6 +151,9 @@ struct AntinucleiInJets {
   Configurable<std::string> weightsXi{"weightsXi", "", "weightsXi"};
   Configurable<std::string> weightsOmega{"weightsOmega", "", "weightsOmega"};
 
+  // Number of events
+  Configurable<int> shrinkInterval{"shrinkInterval", 1000, "variable that controls how often shrinking happens"};
+
   // Reweighting histograms
   TH1F* primaryAntiprotons;
   TH1F* primaryAntiLambda;
@@ -1616,9 +1619,23 @@ struct AntinucleiInJets {
   // Generated events
   void processJetsMCgen(GenCollisionsMc const& collisions, aod::McParticles const& mcParticles)
   {
+    // Define per-event particle containers
+    std::vector<fastjet::PseudoJet> fjParticles;
+    std::vector<TVector3> protonMomentum;
+
+    // Event counter
+    int eventCounter = 0;
+
     // Loop over all simulated collisions
     for (const auto& collision : collisions) {
 
+      // Increment event counter
+      eventCounter++;
+
+      // Clear containers at the start of the event loop
+      fjParticles.clear();
+      protonMomentum.clear();
+
       // Event counter: before event selection
       registryMC.fill(HIST("genEvents"), 0.5);
 
@@ -1629,9 +1646,7 @@ struct AntinucleiInJets {
       // Event counter: after event selection
       registryMC.fill(HIST("genEvents"), 1.5);
 
-      // Loop over all MC particles
-      std::vector<fastjet::PseudoJet> fjParticles;
-      std::vector<TVector3> protonMomentum;
+      // Loop over MC particles
       for (const auto& particle : mcParticles) {
 
         // Select physical primaries within acceptance
@@ -1644,7 +1659,7 @@ struct AntinucleiInJets {
         // Store 3-momentum vectors of antiprotons for further analysis
         if (particle.pdgCode() == PDG_t::kProtonBar) {
           TVector3 pVec(particle.px(), particle.py(), particle.pz());
-          protonMomentum.push_back(pVec);
+          protonMomentum.emplace_back(pVec);
         }
 
         // 4-momentum representation of a particle
@@ -1754,16 +1769,36 @@ struct AntinucleiInJets {
       if (isAtLeastOneJetSelected) {
         registryMC.fill(HIST("genEvents"), 3.5);
       }
+
+      // Shrink large vectors
+      if (eventCounter % shrinkInterval == 0) {
+        fjParticles.shrink_to_fit();
+        protonMomentum.shrink_to_fit();
+      }
     }
   }
   PROCESS_SWITCH(AntinucleiInJets, processJetsMCgen, "process jets mc gen", false);
 
   // Reconstructed events
   void processJetsMCrec(RecCollisionsMc const& collisions, AntiNucleiTracksMc const& mcTracks, McParticles const&)
   {
+    // Define per-event containers
+    std::vector<fastjet::PseudoJet> fjParticles;
+    std::vector<int> antiprotonTrackIndex;
+
+    // Event counter
+    int eventCounter = 0;
+
     // Loop over all reconstructed collisions
     for (const auto& collision : collisions) {
 
+      // Increment event counter
+      eventCounter++;
+
+      // Clear containers at the start of the event loop
+      fjParticles.clear();
+      antiprotonTrackIndex.clear();
+
       // Event counter: before event selection
       registryMC.fill(HIST("recEvents"), 0.5);
 
@@ -1806,8 +1841,6 @@ struct AntinucleiInJets {
 
       // Loop over reconstructed tracks
       int id(-1);
-      std::vector<fastjet::PseudoJet> fjParticles;
-      std::vector<int> antiprotonTrackIndex;
       for (auto const& track : mcTracks) {
         id++;
 
@@ -2028,6 +2061,12 @@ struct AntinucleiInJets {
       if (isAtLeastOneJetSelected) {
         registryMC.fill(HIST("recEvents"), 9.5);
       }
+
+      // Shrink large vectors
+      if (eventCounter % shrinkInterval == 0) {
+        fjParticles.shrink_to_fit();
+        antiprotonTrackIndex.shrink_to_fit();
+      }
     }
   }
   PROCESS_SWITCH(AntinucleiInJets, processJetsMCrec, "process jets MC rec", false);