LoopBack

Смотрю микробенчмарк:
https://github.com/kuksenko/quantum/blob/master/src/main/java/quantum/selfstudy4/LoopBack.java

-----

Запускаю на 
$ java -version
java version "1.8.0-ea"
Java(TM) SE Runtime Environment (build 1.8.0-ea-b105)
Java HotSpot(TM) 64-Bit Server VM (build 25.0-b47, mixed mode)

Intel Core i7-2600 3.40GHz

-----

Результаты

Benchmark                    Mode Thr    Cnt  Sec         Mean   Mean error    Units     Mean/Mean_forward_1K
q.s.LoopBack.backward1K      avgt   1     10    5        0.348        0.009  nsec/op        1.14 +- 0.030
q.s.LoopBack.forward1K       avgt   1     10    5        0.306        0.001  nsec/op        1    

q.s.LoopBack.backward32K     avgt   1     10    5        0.338        0.007  nsec/op        1.13 +- 0.024
q.s.LoopBack.forward32K      avgt   1     10    5        0.298        0.002  nsec/op        1

q.s.LoopBack.backward4M      avgt   1     10    5        0.457        0.014  nsec/op        1.06 +- 0.053
q.s.LoopBack.forward4M       avgt   1     10    5        0.433        0.017  nsec/op        1


Замечание:

Обратный цикл в среднем работает медленнее. 

Смотрим ассемблер.

java -XX:-TieredCompilation -XX:+UnlockDiagnosticVMOptions -XX:+PrintCompilation -XX:+PrintInlining -XX:+PrintAssembly -jar target/microbenchmarks.jar "quantum.sel.*.Lo.*"  -i 10 > out.txt

------------------------------------------------------------------------------------------------------------------

ForwardSum:


  0x0000000002ad4a00: add    0x10(%rbx,%r13,4),%r14d
  0x0000000002ad4a05: movslq %r13d,%r8
  0x0000000002ad4a08: add    0x14(%rbx,%r8,4),%r14d
  0x0000000002ad4a0d: add    0x18(%rbx,%r8,4),%r14d
  0x0000000002ad4a12: add    0x1c(%rbx,%r8,4),%r14d
  0x0000000002ad4a17: add    0x20(%rbx,%r8,4),%r14d
  0x0000000002ad4a1c: add    0x24(%rbx,%r8,4),%r14d
  0x0000000002ad4a21: add    0x28(%rbx,%r8,4),%r14d
  0x0000000002ad4a26: add    0x2c(%rbx,%r8,4),%r14d
  0x0000000002ad4a2b: add    0x30(%rbx,%r8,4),%r14d
  0x0000000002ad4a30: add    0x34(%rbx,%r8,4),%r14d
  0x0000000002ad4a35: add    0x38(%rbx,%r8,4),%r14d
  0x0000000002ad4a3a: add    0x3c(%rbx,%r8,4),%r14d
  0x0000000002ad4a3f: add    0x40(%rbx,%r8,4),%r14d
  0x0000000002ad4a44: add    0x44(%rbx,%r8,4),%r14d
  0x0000000002ad4a49: add    0x48(%rbx,%r8,4),%r14d
  0x0000000002ad4a4e: add    0x4c(%rbx,%r8,4),%r14d  ;*iadd
                                                ; - quantum.selfstudy4.LoopBack::forwardSum@14 (line 51)

  0x0000000002ad4a53: add    $0x10,%r13d        ;*iinc
                                                ; - quantum.selfstudy4.LoopBack::forwardSum@16 (line 50)

  0x0000000002ad4a57: cmp    %r11d,%r13d
  0x0000000002ad4a5a: jl     0x0000000002ad4a00  ;*if_icmpge
                                                ; - quantum.selfstudy4.LoopBack::forwardSum@7 (line 50)
                                                
                                                
------------------------------------------------------------------------------------------------------------------

BackwardSum

  0x000000000299e895: add    0x10(%rdx,%rcx,4),%eax  ;*iadd
                                                ; - quantum.selfstudy4.LoopBack::backwardSum@15 (line 59)

  0x000000000299e899: dec    %ecx               ;*iinc
                                                ; - quantum.selfstudy4.LoopBack::backwardSum@17 (line 58)

  0x000000000299e89b: cmp    %r11d,%ecx
  0x000000000299e89e: jg     0x000000000299e895  ;*iflt
                                                ; - quantum.selfstudy4.LoopBack::backwardSum@8 (line 58)