Fibonacci(50): Rust slower than Go ?

I my previous blog post "Fibonnaci (50) : Java > C > C++ > D > Go > Terra (Lua) > LuaJit (Lua) ",  I compared language performance using Fibonnaci algorithm where Rust language was not included.

Rust language is being clamined to  be "a system programming language that runs blazingly fast, prevents almost all crashes*, and eliminates data races".

I thought of running same fibonnaci algorithm against Rust to see how blazingly fast is it.

You can find fib.rs source code here on lang-compare  on my github.  I compared it against fib.go and fib.c 

FIBONACCI - 50

Language C (gcc-4.9.1):
gcc-4.9 -O3 fib.c
/usr/bin/time -lp ./a.out 50
LANGUAGE C: 12586269025
real        52.87
user        52.83
sys          0.01
    557056  maximum resident set size
         0  average shared memory size
         0  average unshared data size
         0  average unshared stack size
       163  page reclaims
         0  page faults
         0  swaps
         0  block input operations
         0  block output operations
         0  messages sent
         0  messages received
         0  signals received
         0  voluntary context switches
       497  involuntary context switches

Language Go (go1.4beta1 darwin/amd64):
go build fib.go
/usr/bin/time -lp ./fib 50
LANGUAGE GO: 12586269025
real        96.56
user        96.47
sys          0.06
   1232896  maximum resident set size
         0  average shared memory size
         0  average unshared data size
         0  average unshared stack size
       300  page reclaims
         0  page faults
         0  swaps
         0  block input operations
         0  block output operations
         0  messages sent
         0  messages received
         0  signals received
      9010  voluntary context switches
      1338  involuntary context switches

Language Rust (0.13.0-nightly (45cbdec41 2014-11-07 00:02:18 +0000)):
/usr/local/bin/rustc --opt-level 3 fib.rs
/usr/bin/time -lp ./fib 50
LANGUAGE Rust: 12586269025
real       101.53
user       101.44
sys          0.01
   1040384  maximum resident set size
         0  average shared memory size
         0  average unshared data size
         0  average unshared stack size
       229  page reclaims
        46  page faults
         0  swaps
         0  block input operations
         0  block output operations
         0  messages sent
         0  messages received
         0  signals received
         1  voluntary context switches
       914  involuntary context switches

Surprisingly results are as below where CPU usage (real) is as C (52.87) > Go (96.56)  > Rust (101.53).

Language	CPU(real)	Memory(Max resident size)
C	52.87	544K
Go	96.56	1204K
Rust	101.53	1016K

NOTE:  Is there a better way to read command line argument and convert into integer?   I felt it was very painful to figure that out.

E.g.
In C language:
long n = atol(argv[1]);

In Go language:
n, _ := strconv.Atoi(os.Args[1])

In Rust language:
let args = os::args(); let args = args.as_slice(); let n :u64 = from_str::<u64>(args[1].as_slice()).unwrap();

Fibonacci(50) performance : Java > C > C++ > D > Go > Terra (Lua) > Lua-JIT (Lua)

Recently,  I have been spending some time to learn D and Go  languages.  D Lang is an evolution of C++ where as Go is being claimed to be an evolution of C but I think  it is a Google's attempt to replace dependency on the Java.

I really like the simplicity of Go where the learning curve is very lean but accepting it as a system level programming would be a tough sell.   I think it is more for Java/Python developers for building the enterprise softwares rather than using it for hardware/device-drivers/embedded programming, but you never know as  Google is very successful selling  Android on low powered smart devices.

I did some performance bench for C, C++, D and Go  languages using the Fibonacci algorithm.

Results: (see updated results at the end)

For Fibonacci(25), C++ >= Go > C >= Lua-JIT > D > Lua-Terra > Java 1.6

For Fibonacci(50), Java > C > C++ > D-ldc > D-dmd > Go > Lua-Terra > Lua-JIT

Now surprisingly,  Java out performed C/C++  for Fibonacci (50) which hurts my ego :) !!

Language	% C++ Speed	Compiler/VM	Flags
FIBONACCI-25
C++	100.0000	Apple LLVM version 6.0	-O3
GO	100.0000	go version go1.3.3 darwin/amd64
C	77.7778	Apple LLVM version 6.0	-O3
LUA	77.7778	LuaJIT 2.0.3
D	63.6364	dmd	-m64 -O  -inline -noboundscheck
D	63.6364	ldc	-m64 -O  -inline
LUA	43.7500	Terra
JAVA 1.6	43.4783	1.6.0_65-b14-462-11M4609)
FIBONACCI-50
JAVA 1.6	169.6710	1.6.0_65-b14-462-11M4609)
C	101.9846	Apple LLVM version 6.0	-O3
C++	100.0000	Apple LLVM version 6.0	-O3
D	92.6376	ldc	-m64 -O  -inline
D	81.7197	dmd	-m64 -O  -inline -noboundscheck
GO	76.7760	go version go1.3.3 darwin/amd64
LUA	43.9684	Terra
LUA	38.9649	LuaJIT 2.0.3

For Source code and results:  check out my github project.

Update:
It seems Clang on MacOS has some issue. I executed these on my Linux Virtual machine with gnu g++ and g++ is outperforming Java. My ego is intact :)

$g++ --version
g++ (GCC) 4.4.7 20120313 (Red Hat 4.4.7-4)
$g++ -O3 fib.cpp
time ./a.out 50
real 0m47.991s
user 0m47.981s
sys 0m0.000s

$java -version
java version "1.7.0_51"
$javac fib.java
$time java fib 50
real 0m51.897s
user 0m51.815s
sys 0m0.113s