Added no-gc benchmark results

Author: A-Herzog

.gitignore

@@ -1,4 +1,5 @@
 .vscode/settings.json
 TinyFastSimulator-Java.jar
 images/Benchmark-LoadBalancer.xlsx
-images/Benchmark-noLoadBalancer.xlsx
+images/Benchmark-noLoadBalancer.xlsx
+images/Benchmark-LoadBalancer-Epsilon.xlsx

README.md

@@ -48,7 +48,7 @@ Usage example: `./TinyFastSimulator-Benchmark.sh 32 100000000 results.txt`
 
 ## Benchmark results
 
-The following benchmarks were performed on a system with two CPU of the type AMD Epyc 7281 (Zen+ architecture, 16 physical / 32 logical cores each).
+The following benchmarks were performed on a system with two CPUs of the type AMD Epyc 7281 (Zen+ architecture, 16 physical / 32 logical cores each).
 
 ### No load balancer
 
@@ -62,6 +62,14 @@ The following benchmarks were performed on a system with two CPU of the type AMD
 
 ![Benchmark - load balancer - relative performance](images/Benchmark4.png)
 
+### Load balancer + No garbage collection ("epsilon collector")
+
+![Benchmark - load balancer - epsilon collector - absolute performance](images/Benchmark5.png)
+
+![Benchmark - load balancer - epsilon collector - relative performance](images/Benchmark6.png)
+
+(The JVM options `-XX:+UnlockExperimentalVMOptions -XX:+UseEpsilonGC` were used to active the epsilon collector.)
+
 Even so the benchmark probably runs completely from the cache and hardly accesses the RAM, the SMT winnings are nevertheless very high.
 
 The number of arrivals was 100,000,000 in the single thread case, 200,000,000 when using two threads etc. In this way the simulation runtime are (approximately) constant over all numbers of threads.