LinkedList info

java-optimizations/LinkedList.diff

@@ -0,0 +1,30 @@
+--- ../primes.java	2016-09-22 22:10:58.787516331 +0300
++++ primes.java	2016-09-22 22:11:14.919516331 +0300
+@@ -2,15 +2,15 @@
+ import java.lang.Math;
+
+ class PrimeNumbersGenerator {
+-	ArrayList<Integer> get_primes7(int n) {
+-		ArrayList<Integer> res = new ArrayList<Integer>();
++    List<Integer> get_primes7(int n) {
++        List<Integer> res = new LinkedList<Integer>();
+
+ 		if (n < 2) return res;
+ 		if (n == 2) {
+ 			res.add(2);
+ 			return res;
+ 		}
+-		ArrayList<Integer> s = new ArrayList<Integer>();
++        List<Integer> s = new ArrayList<Integer>();
+ 		for (int i = 3; i < n + 1; i += 2) {
+ 			s.add(i);
+ 		}
+@@ -46,7 +46,7 @@
+ 		long startTime = System.currentTimeMillis();
+ 		long periodTime = Long.parseLong(System.getenv("RUN_TIME"), 10) * 1000;
+
+-		ArrayList<Integer> res;
++        List<Integer> res;
+
+ 		while ((System.currentTimeMillis() - startTime) < periodTime) {
+ 			res = (new PrimeNumbersGenerator()).get_primes7(10000000);

java-optimizations/README.md

@@ -4,5 +4,6 @@ Java can be (a lot) faster than what we see it with its default interpeter setti
 * Initial heap size -- setting this larger than the default value will allow the ArrayList to grow more efficiently by pre-allocating a bigger memory block from the system. The "java" command-line argument is "-Xms500m".
 * Heap and GC tuning -- we can control the ratio between the old and young generations. The "java" command-line argument is "-XX:NewRatio=1".
 * Skip the unneeded boxing/unboxing -- this requires a more native Java implementation. See the source code of "[primes-alt.java](../primes-alt.java)".
+* Use LinkedList -- it has O(1) complexity for adding elements. Benchmark tests however don't show significant improvement under Java 8, and under Java 7 we see a slow down of 59%.
 
 These tweaks are not included in the Java benchmark test, because I wanted to test the default languages setup. I don't aim to fine-tune any of the implementations by optimizing them for the current task/algorithm. This is a generic test, not an attempt to complete the current task in the fastest possible way for each programming language.