From 3650114b2126c5c4fec2fa6282a61f2faa610cf2 Mon Sep 17 00:00:00 2001
From: Jan Kieseler <jkiesele@cern.ch>
Date: Tue, 27 Aug 2024 09:36:30 +0200
Subject: [PATCH] better seeding

---
 bind/G4Calo.py | 28 +++++++++++++++++-----------
 1 file changed, 17 insertions(+), 11 deletions(-)

diff --git a/bind/G4Calo.py b/bind/G4Calo.py
index 68a90ec..7f5a7a0 100644
--- a/bind/G4Calo.py
+++ b/bind/G4Calo.py
@@ -154,7 +154,6 @@ def _run_mini_batch(
     seed = int(batch_seed + counter)
         
     print(f"Running mini batch with seed {seed}")
-    
     G4System.init(cw, seed) #counter gives random seed offset
 
     df = G4System.run_batch(nEvents, particleSpec, minEnergy_GeV, maxEnergy_GeV,"")
@@ -189,15 +188,21 @@ def run_batch(
     
     nevents = [nEventsPerCore if i < nCores - 1 else nEventsLastCore for i in range(nCores)]
 
-    batch_seed = int(time.time()) #this is seconds
-    # make sure that the seed is different for next run. This is not perfect but should be good enough
-    # given the course takes a week this should be sufficient (10^7 seconds ~ 4 months)
-    time.sleep(1) 
-    batch_seed = batch_seed % 10000000
-    print(f"Batch seed: {batch_seed}")
+    #the used seeds are stored in the home directory in a file called .seeds.txt
+    # check if file exists, if so, read last seed. If not create it
+    if os.path.exists(os.path.expanduser("~/.g4calo_seeds.txt")):
+        with open(os.path.expanduser("~/.g4calo_seeds.txt"), "r") as f:
+            seed = int(f.read())
+    else:
+        seed = 0
+    with open(os.path.expanduser("~/.g4calo_seeds.txt"), "w") as f:
+        f.write(str(seed + nCores))
+
+    seed += 1
+    print(f"Batch seed: {seed}")
     #use a multiprocessing pool to run the mini batches in parallel
     with multiprocessing.Pool(nCores) as pool:
-        dfs = pool.starmap(_run_mini_batch_silent, [(gd, nevents[i], particleSpec, minEnergy_GeV, maxEnergy_GeV, batch_seed, i) for i in range(nCores)])
+        dfs = pool.starmap(_run_mini_batch_silent, [(gd, nevents[i], particleSpec, minEnergy_GeV, maxEnergy_GeV, seed, i) for i in range(nCores)])
 
     alldf = pd.concat(dfs)
     alldf.reset_index(drop=True, inplace=True)
@@ -212,18 +217,19 @@ def run_batch(
 
 def _fill_event(gd : GeometryDescriptor, 
                  particleSpec: str,
-                 energy: float):
+                 energy: float,
+                 seed: int = -1):
     
     from  G4Calo import __G4System
     G4System = __G4System()
-    G4System.init(gd, -1)
+    G4System.init(gd, seed)
     G4System.run_batch(1, particleSpec, energy, energy,"")
     return gd
         
 def display_event(gd : GeometryDescriptor, 
                  particleSpec: str,
                  energy: float,
-                 logE = False, renderer=None):
+                 logE = False, renderer=None, seed = -1):
     
     #run _fill_event in forked mode using 1-core multiprocessing to avoid G4 singletons to interfere
     with multiprocessing.Pool(1) as pool: