Profiled a run, turned off flocking behaviour in skylarks

Searching for neighbours turned out to consume ~80% of processing time...

Makefile

@@ -23,16 +23,16 @@ docs:
 
 profile:
 	# run and profile an example simulation
-	if [ -d "example_results" ]; then rm -r example_results; fi
-	./runprofile.jl -o example_results
+	if [ -d "profile_results" ]; then rm -r profile_results; fi
+	./runprofile.jl -o profile_results
 
 container:
 	#TODO create a Singularity container
-	echo "Not yet implemented (#43)"
+	echo "Not yet implemented (#43)."
 
 release:
 	echo "Not yet implemented."
 
 install:
 	#TODO install Julia and/or package dependencies?
-	echo "Not yet implemented."
+	echo "Not yet implemented (#67)."

runprofile.jl

@@ -11,11 +11,11 @@ Profile.clear()
 
 @profile simulate()
 
-open("profile_flat.txt", "w") do s
+open("profile_results/profile_flat.txt", "w") do s
     Profile.print(IOContext(s, :displaysize=>(300,145)), format=:flat, mincount=10, sortedby=:count)
 end
 
-open("profile_tree.txt", "w") do s
+open("profile_results/profile_tree.txt", "w") do s
     Profile.print(IOContext(s, :displaysize=>(300,300)), mincount=10)
 end
 

src/nature/species/skylark.jl

@@ -88,9 +88,14 @@ end
 Non-breeding adults move around with other individuals and check for migration.
 """
 @phase Skylark nonbreeding begin
-    #XXX is flocking behaviour important? It may be quite computationally expensive...
+    #FIXME searching for neighbours is *the* performance bottleneck!!
+    # Two options:
+    # (a) ignore flocking, simply move randomly (scientifically OK, doesn't look as good)
+    # (b) select a random neighbour once, then keep following that individual
+
     # flocking behaviour - follow a random neighbour or move randomly
-    neighbours = @neighbours(self.visionrange) #XXX check for the closest neighbour(s)?
+    neighbours = ()
+    #neighbours = @neighbours(self.visionrange) #XXX check for the closest neighbour(s)?
     isempty(neighbours) ?
         @walk("random", self.movementrange) :
         @follow(@rand(neighbours), 30m) #XXX magic number