[Open Babel] Java performance penalty?

[Miguel <mt...@mt...>]

Geoff wrote:
> (begin rant) I'll also admit a significant bias. One reason I code in
> C++ is that I hate the performance penalty of Java. As I'm doing some
> conformer searches of a good number of molecules, I'd prefer to see
> *fast* coordinate generation and molecular mechanics. (end rant)

Geoff,

I believe that the performance penalty for working in Java is neglig

I have quite a bit of experience with systems-level programming, and
performance is extremely important to me.

A year ago, in preparing to write the 3D graphics engine, I did some
serious benchmarking to to compare Java performance with C performance.
The benchmarks emulated the CPU-intensive core operations of a software
graphics z-buffer implementations. One of the things I was particularly
interested in was determining the impact of using
array-indexing-with-bounds-checking vs. pointers.

My general findings were:
 - Overall, the performance penalty for programming
   in Java was about 10%
 - the worst benchmark was about 20%
 - others were only 0-5% slower

I also tried compiling with the gcj Gnu Compiler for Java that generates
machine instructions. I found:
 - there was no runtime performance difference between the
   Sun hot-spot runtime compiler and precompiled gcj code
 - turning off array bounds checking (an option with gcj)
   in the context of a realistic benchmark had no
   measurable performance impact.

Other comments:
 - these tests did not involve the Java class libraries
   (which may be slow)
 - I suspect that Java makes it easy for inexperienced
   programmers to build systems with relatively poor
   performance because of heap memory allocation/garbage
   collection.


But, I honestly don't think that there is any significant performance
impact for writing systems in Java.


Miguel