我在库中看到了一些并发问题，我使用它实现了Celluloid actor。出于好奇，我在Celluloid中用the benchmarking script运行了几个不同的口译员。
Ruby 1.9.3
Calculating -------------------------------------
spawn 387 i/100ms
calls 757 i/100ms
async calls 431 i/100ms
messages 3063 i/100ms
-------------------------------------------------
spawn 3957.4 (±3.9%) i/s - 20124 in 5.093148s
calls 8054.8 (±5.8%) i/s - 40878 in 5.092786s
async calls 15709.4 (±11.0%) i/s - 77580 in 5.008169s
messages 497844.7 (±16.0%) i/s - 2383014 in 4.989021s
Ruby 2.1.1
Calculating -------------------------------------
spawn 359 i/100ms
calls 749 i/100ms
async calls 383 i/100ms
messages 3579 i/100ms
-------------------------------------------------
spawn 3823.2 (±8.1%) i/s - 19027 in 5.011383s
calls 8205.4 (±7.3%) i/s - 41195 in 5.049513s
async calls 14086.1 (±8.7%) i/s - 70089 in 5.016073s
messages 766163.5 (±19.8%) i/s - 3510999 in 4.993806s
JRuby 1.7.11使用JIT
Calculating -------------------------------------
spawn 75 i/100ms
calls 294 i/100ms
async calls 154 i/100ms
messages 4548 i/100ms
-------------------------------------------------
spawn 2521.8 (±24.3%) i/s - 11625 in 4.999000s
calls 8207.7 (±25.3%) i/s - 37632 in 5.049000s
async calls 9001.3 (±30.2%) i/s - 36652 in 4.971000s
messages 334366.4 (±14.7%) i/s - 1619088 in 4.999000s
JRuby 1.7.11使用AOT
Calculating -------------------------------------
spawn 88 i/100ms
calls 324 i/100ms
async calls 126 i/100ms
messages 2576 i/100ms
-------------------------------------------------
spawn 2198.8 (±25.5%) i/s - 9416 in 5.004000s
calls 6188.9 (±21.5%) i/s - 29160 in 5.014000s
async calls 6906.6 (±26.3%) i/s - 29862 in 4.996000s
messages 198558.0 (±20.3%) i/s - 922208 in 4.995000s
如果我正确地阅读此内容，1.9和2.1解释器产生的演员数量多于在JRuby中运行时的演员数量。我原本期望JRuby能够产生更多的演员，并且比任何一个C语言解释器运行得更快 - 但事实似乎并非如此。
我运行基准测试的方式是否缺少某些东西，或者Celluloid在启动大量演员方面有些困难？
编辑我最初使用了错误的设置来编译AOT。现在AOT的数字显得更低 - 虽然除了编译之外我没有指定任何其他选项。