Comments on: Why is C so slow? Java vs. C benchmark.

By: Daniel

Daniel — Fri, 15 Feb 2008 03:42:40 +0000

@ahmet
I know its not a good benchmark. This post was actually a little tongue-in-cheek. People frequently benchmark C/C++ vs Java in just this manor in order to defame Java. I used the same invalid benchmark, and found that the results were actually reversed on my system.

By: ahmet

ahmet — Sat, 09 Feb 2008 23:07:52 +0000

try with java 1.6 update 4. you will possibly get better results with Java. Still, just looping is not really a good benchmark.

By: Daniel

Daniel — Tue, 11 Dec 2007 20:57:55 +0000

In c, long is 32 bit, in Java, int is 32 bit. That’s why the difference.
If I use 64 bit (long) in Java, I get slightly different results.

By: Aviad Ben Dov

Aviad Ben Dov — Tue, 11 Dec 2007 20:44:17 +0000

I just noticed something, not that it should matter [much]: In the C version you used “long” for the loops and in the Java version you used “int”. Do you get different results when you change these?

By: Daniel

Daniel — Sun, 09 Dec 2007 07:52:47 +0000

I was doing a base example just to counteract the common myth that “C is always faster than Java.” The truth of the matter is that no micro-benchmark will capture all cases. The important thing to point out is that Java programs will benefit from the latest VM, where a C program would have to be recompiled to enjoy optimization for a particular platform.

By: Aviad Ben Dov

Aviad Ben Dov — Sun, 09 Dec 2007 05:47:46 +0000

Try to see if the compilation could be better optimized. Remember that the JVM enjoys the added benefit of being able to (sometimes) compile the bytecode (using JIT) to a more localized assembly code, unlike GCC which usually compiles for the (reasonably) lowest common denominator to allow binary portability between different computers (for example, allowing the binary you compiled to still execute on a Pentium 3 machine).