C# Hashtable vs c++ hash_map

Question

I'm comparing the following code in C++ and C# and C# (Mono 2.4) seems to be faster. Is there anything wrong with the C++ code?

 #include <map>
 #include <string>
 #include <iostream>
 #include <ext/hash_map>
 #include <boost/any.hpp>

 int main()
 {
    //std::map<long, long> m;
    // hash_map is a little bit faster
    __gnu_cxx::hash_map<long, long> m;

    for( long i = 0; i < 1000000; ++i )
    {
        m[i]  = i;
    }

 }

And C#

 using System;
 using System.Collections;

 public int Main()
 {
     Hashtable m = new Hashtable();

     for( long i = 0; i < 1000000; ++i )
     {
        m[i]  = i;
     }

}

C# code is actually twice as fast on the same machine.

$ time ./a.out

real    0m1.028s
user    0m0.986s
sys     0m0.041s

$ time mono test.exe

real    0m0.603s
user    0m0.732s
sys     0m0.090s

Also note that initial hash map sizes may be different. E.g., if the C# is larger, and C++ version lower, you might easily see a big differences. — csl
– csl, Commented Oct 27, 2009 at 23:20
What OS are you on? Is the C++ binary stripped? Is it statically or dynamically linking the a.out file. — jabbie
– jabbie, Commented Oct 27, 2009 at 23:23
Hi. I just compiled with gcc test.cc. No switches, nothing. Didn't try with the default sizes either, but the Mono/C# default size is not very big either. Didn't try Boost.Unordered. The old fedora I'm using doesn't seem to have it. I'll look at a newer box (ok, or I'll install it :-P). — pablo
– pablo, Commented Oct 27, 2009 at 23:23

Crashworks · Accepted Answer · 2009-10-27 23:35:23Z

8

You need to compile the C++ code with compiler optimizations turned on for a fair comparison. Otherwise, you are comparing apples to debug builds — the compiler will not even try to emit fast code.

In GCC this would be the -O3 flag, to start with.

answered Oct 27, 2009 at 23:35

Crashworks

41.6k14 gold badges108 silver badges171 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

csl · Accepted Answer · 2009-10-27 23:22:06Z

4

Some thoughts:

You're comparing Mono against Boost
Which optimization (default or not) settings did you use for Mono and C++ ?
What are the initial / default hash table sizes for the C# and C++ versions ?
Have you tried comparing against other C++ hash maps, such as the sparse hash table ?
Remember not only to look at real time, but actual CPU time used.

answered Oct 27, 2009 at 23:22

csl

11.4k5 gold badges64 silver badges92 bronze badges

2 Comments

pablo Over a year ago

I'd like to know how much faster Boost is. Didn't enable any optimizations for Mono nor C++. Not sure about the table sizes, I'll have to check. Considering other hashtables is exactly what I'd like to do. Yep, but even real+user is better for Mono.

Joe Over a year ago

I'd be surprised if Boost is that much faster. I don't think there have been great improvements in hash_map implementations the last few years. (Btw, your Linux setup might have the std::tr1::unordered_map, which is another hashing container implementation...)

Collectives™ on Stack Overflow

C# Hashtable vs c++ hash_map

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related