Why is branch prediction faster than no branch?

Question

Inspired by this question: Why is a sorted array faster to process than an unsorted array?

I wrote my own branch prediction experiment:

public class BranchPrediction { public static void main(final String[] args) { long start; long sum = 0; /* No branch */ start = System.nanoTime(); sum = 0; for (long i = 0; i < 10000000000L; ++i) sum += i; System.out.println(System.nanoTime() - start); System.out.println(sum); /* With branch */ start = System.nanoTime(); sum = 0; for (long i = 0; i < 10000000000L; ++i) if (i >= 0) sum += i; System.out.println(System.nanoTime() - start); System.out.println(sum); /* No branch (again) */ start = System.nanoTime(); sum = 0; for (long i = 0; i < 10000000000L; ++i) sum += i; System.out.println(System.nanoTime() - start); System.out.println(sum); /* With branch (again) */ start = System.nanoTime(); sum = 0; for (long i = 0; i < 10000000000L; ++i) if (i >= 0) sum += i; System.out.println(System.nanoTime() - start); System.out.println(sum); } } 

The result confuses me: in accordance with the output of the program, a loop with a branch is reliably faster than branch loops.

Output Example:

 7949691477 -5340232226128654848 6947699555 -5340232226128654848 7920972795 -5340232226128654848 7055459799 -5340232226128654848 

Why is this so?

Edit:

• Disassembled class shows that the Java compiler did not optimize (skip) anything ( https://gist.github.com/HouzuoGuo/5692424 )
• Java comparison method used by the author Why is it faster to process a sorted array than an unsorted array? - this is the same as mine.
• The device is the i7 core for Intel, running Linux 3.2 64-bit and Oracle JVM 1.7 64-bit
• When I increase the number of iterations of a loop, a loop with a branch starts several SECONDS faster than a non-single one.

Answer 1

After starting the same experiment on other computers (servers and Intel workstations), I can conclude that the phenomenon that I experienced is specific to this notebook processor (Intel i7 Q740M).

==== 6 months later editing ====

Check it out: http://eli.thegreenplace.net/2013/12/03/intel-i7-loop-performance-anomaly/

Answer 2

Keep in mind that the JVM optimizes execution internally, and there are caches on your computer that speed up computing. Since you have such a powerful processor (many independent cores), this is not strange. Also note that there is code that runs under Java code that maps to your computer’s machine code. Just enter the code you can optimize to let the JVM worry about it.

EDIT: Machines and equipment, such as heavy loads, work with greater efficiency. Especially the cache.