Java vs. C/C++ Benchmark Settings

Thump, thump - up on the soap box (again)..

I keep running into benchmarks where C/C++, and specifically GCC, is often under-represented because the executables are not built correctly. I rarely see this (really poor build methodology) in real-world code, eg: Open Source make files. Conversely, it is even more rare to find the -X series switches applied to Java code in the real-world, as they are for many benchmarks.

Kind of makes you wonder about the underlying intentions of most 'Java vs. C++' authors, or maybe they are just plain and simply ignorant in this regard. Nonetheless, you would think they would want accurate results no matter how they turned out.

GCC may well be even more portable than (or at least as reliably portable as) Java and since GCC covers so much ground a developer needs to pay at least a little attention to the compiler switches when optimizing the build for a platform. One of the great things about GCC is that you can use it just about anywhere and get great results, one of the hassles is that you have to be cognizant of build options. The Intel optimization switch selection was pretty straight forward because, of course, Intel icc builds for Intel compatible architectures only.

I would have liked to provide some longer run-times for some of the tests, but in each of these cases, one or several of the Java runtimes would exhaust available memory on my system, so I stuck with the original test durations. As I mentioned elsewhere, I had to spend a lot of time experimenting with the various 'X' options for most of the Java runtimes on a test-by-test bases, and this was one of the reasons why.

THUMP - off the soapbox.

This website is hosted for free by . Get your own Free Website now!